USF Libraries
USF Digital Collections

Cervical cancer screening disparities in an ethnically diverse population of women residing in the United States in 1999

MISSING IMAGE

Material Information

Title:
Cervical cancer screening disparities in an ethnically diverse population of women residing in the United States in 1999 a secondary analysis of data from the 1999 behavioral risk factor surveillance system
Physical Description:
Book
Language:
English
Creator:
Morgan, Chodaesessie Wellesley-Cole
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla.
Publication Date:

Subjects

Subjects / Keywords:
American cancer society
Centers for disease control and prevention
Compliance with
Ethnic origin
Preventive health model
Primary care provider advice
Public health research
Self-reported data
Dissertations, Academic -- Public Health -- Doctoral -- USF   ( lcsh )
Genre:
government publication (state, provincial, terriorial, dependent)   ( marcgt )
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Summary:
ABSTRACT: Black American women have the highest screening rates for cervical cancer among all the ethnic groups in the United States. Even though evidence from the literature suggests that the number of deaths from cervical cancer in the United States could be reduced by preventive screening, this particular minority population still suffers disproportionately higher mortality from the disease than the other minority and majority populations in the United States.This study was proposed to investigate cancer screening disparities among different subpopulations of women residing in the United States during 1999, and to recommend public health interventions that could potentially increase cervical cancer screening rates, thereby decreasing differential mortality rates for cervical cancer among these subpopulations.The Preventive Health Model in conjunction with data from the 1999 Behavioral Risk Factor Surveillance System was used to identify the covariates of cervical cancer screening behavior in an ethnically diverse population of American women residing in the United States during the specified timeframe. Univariate, bivariate and multivariable logistic regression procedures were used to evaluate the association between each one of the independent variables and the dependent variable (compliance with the 1999 cervical screening guidelines of the American Cancer Society).
Thesis:
Thesis (Ph.D.)--University of South Florida, 2005.
Bibliography:
Includes bibliographical references.
System Details:
System requirements: World Wide Web browser and PDF reader.
System Details:
Mode of access: World Wide Web.
Statement of Responsibility:
by Chodaesessie Wellesley-Cole Morgan.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 197 pages.

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 001670340
oclc - 62285275
usfldc doi - E14-SFE0001202
usfldc handle - e14.1202
System ID:
SFS0025523:00001


This item is only available as the following downloads:


Full Text
xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam Ka
controlfield tag 001 001670340
003 fts
005 20051216093301.0
006 m||||e|||d||||||||
007 cr mnu|||uuuuu
008 051116s2005 flu sbm s000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0001202
035
(OCoLC)62285275
SFE0001202
040
FHM
c FHM
049
FHMM
090
RA425 (Online)
1 100
Morgan, Chodaesessie Wellesley-Cole
0 245
Cervical cancer screening disparities in an ethnically diverse population of women residing in the United States in 1999
h [electronic resource] :
b a secondary analysis of data from the 1999 behavioral risk factor surveillance system /
by Chodaesessie Wellesley-Cole Morgan.
260
[Tampa, Fla.] :
University of South Florida,
2005.
502
Thesis (Ph.D.)--University of South Florida, 2005.
504
Includes bibliographical references.
516
Text (Electronic thesis) in PDF format.
538
System requirements: World Wide Web browser and PDF reader.
Mode of access: World Wide Web.
500
Title from PDF of title page.
Document formatted into pages; contains 197 pages.
520
ABSTRACT: Black American women have the highest screening rates for cervical cancer among all the ethnic groups in the United States. Even though evidence from the literature suggests that the number of deaths from cervical cancer in the United States could be reduced by preventive screening, this particular minority population still suffers disproportionately higher mortality from the disease than the other minority and majority populations in the United States.This study was proposed to investigate cancer screening disparities among different subpopulations of women residing in the United States during 1999, and to recommend public health interventions that could potentially increase cervical cancer screening rates, thereby decreasing differential mortality rates for cervical cancer among these subpopulations.The Preventive Health Model in conjunction with data from the 1999 Behavioral Risk Factor Surveillance System was used to identify the covariates of cervical cancer screening behavior in an ethnically diverse population of American women residing in the United States during the specified timeframe. Univariate, bivariate and multivariable logistic regression procedures were used to evaluate the association between each one of the independent variables and the dependent variable (compliance with the 1999 cervical screening guidelines of the American Cancer Society).
590
Adviser: Dr. Robert McDermott.
653
American cancer society.
Centers for disease control and prevention.
Compliance with.
Ethnic origin.
Preventive health model.
Primary care provider advice.
Public health research.
Self-reported data.
690
Dissertations, Academic
z USF
x Public Health
Doctoral.
773
t USF Electronic Theses and Dissertations.
4 856
u http://digital.lib.usf.edu/?e14.1202



PAGE 1

Cervical Cancer Screening Disparities in an Ethnically Diverse Population of Women Residing in the United States in 1999: A S econdary Analysis of Data from the 1999 Behavioral Risk Factor Surveillance System by Chodaesessie Wellesley-Cole Morgan A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Family and Community Health College of Public Health University of South Florida Major Professor: Robert J. McDermott, Ph.D. Thomas J. Mason, Ph.D. Karen M. Perrin, Ph.D. Ellen M. Daley, Ph.D. Richard G. Roetzheim, M.D. Date of Approval: July 01, 2005 Keywords: American Cancer Society, Cent ers for Disease Control and Prevention, compliance with, ethnic origin, Preventive He alth Model, primary care provider advice, public health research, self-reported data Copyright 2005, Chodaesessie Wellesley-Cole Morgan

PAGE 2

Dedication This dissertation is dedicated to my be loved children, Yeatie and Yannick. I thank you with all my heart for your steadfast belief in my ability to stay the course. Thank you for your constant encouragement throughout th is long, long, journey. I would never have made it through without your prayers, and without your unwavering love for me.

PAGE 3

Acknowledgements I would like to acknowledge my much d early loved husband, Beale Morgan, M.D. This journey began when you suggested that I re turn to academia much earlier in my life than I had planned. I thank you for offering me such excellent advice, and for being my anchor throughout the turbulence encountered during this journey. It was an honor to work with a brill iant, dedicated, and dynamic faculty, who accepted the invitation to serve on both my doctoral and dissertation committees: Robert McDermott, Ph.D., my major professo r, I appreciate your patience, kindness and wisdom. Thank you for guiding me through th is incredible journey upon which we embarked, so many years ago, during my Master’s degree program. Thomas Mason, Ph.D., Karen (Kay) Perrin, Ph.D., Ellen Daley, Ph.D. and Richard Roetzheim, M.D., I thank you for your guida nce, and encouragement during the many years that it has taken to develop a nd complete this research project. I would also like to expre ss my gratitude to John Large, Ph.D., Kofi Marfo, Ph.D. and Patricia Romilly, M.D., me mbers of my expert panel, and to David Hogeboom, BBA, ‘my SAS guru.’ It was my pleasure to work w ith each and every one of you, and I thank you for all the assistance you offered me. Finally, I would like to ac knowledge the friends who ha ve traveled with me throughout this journey: Mirt ha Montejo Whaley, Ph.D.(c), David Austell, Ph.D., Phillip Marty, Ph.D., Rony Francois, M.D., Tracey Perez Koehlmoos, Ph.D., Robert Tabler, Ph.D.(c), Cecelia Wachdorf, Ph.D., Lt. Commander Robert Fry, Ph.D. and Arlesia Brock, Ph.D.(c). I thank you with my heart and soul for your abiding support and friendship.

PAGE 4

i Table of Contents List of Tables v Abstract vii Chapter One: Introduction 1 Statement of the Problem 1 Rationale for the Study 5 Purpose of the Study 7 Theories and Theoretical Mode ls of Health Behavior 8 The Preventive Health Model and Constructs from the Health Belief Model, the Theory of Reasoned Action, Social Cognitive Theory 11 Assumptions of the Study 13 Delimitations of the Study 14 Limitations of the Study 15 Definitions of Terms 15 Chapter Two: Literature Review 26 Etiology of Cervical Cancer 26 Development of the Papanicolaou Screening Test 27 Principles of Cancer Screening 29 Effectiveness and Cost-Effectiveness of Cervical Cancer Screening 32 Cervical Cancer Screening Guidelines of the American Cancer Society from 1999 to 2005 34 Studies of: Covariates of Cervi cal Cancer Screening Behavior: 1990 to 2005 36 Physician Recommendation and Adherence to Cancer Screening Guidelines 48 Ethnic Group and Cancer Screening Behavior 53 The Preventive Health Model 59 Overview of the Behavioral Risk Factor Surveillance System 67 Reliability and Validity of the 1999 BRFSS Data Set 68 Limitations and Strengths of the 1999 BRFSS Data Set 69 Public Health Significance of Study 70

PAGE 5

ii Chapter Three: Research Methods 73 Purpose of the Study 73 Research Questions 73 Null Hypotheses 74 Study Design 75 Selection of Study Sample 75 Selection of Variables 77 Operationalization of the Constructs of the Preventive Health Model 78 Background Factors Construct 79 Socio-demographic Variables 79 Past Screening Variables 81 Health Profile Variables 81 Representation, Social Influence and Program Factors Constructs 82 Summary of the Proposed Study 83 Data Management 84 Logistic Regression Models 86 Regression Diagnostics 90 Chapter Four: Results 93 Description of Study Sample 93 Constructs of the Preventive Health Model 96 Background Factors Construct 96 Socio-demographic Variables 96 Past Screening Behavior Variables 100 Health Profile Variables 100 Representation, Social Influence and Program Factors Constructs 101 Data Analyses 102 Frequency Analyses 102 Univariate Analyses 105 Bivariate Analyses 107 Logistic Regression Analyses 109 Regression Diagnostic Analyses 113 Summary of the Data Analyses 114 Research Question 1 114 Black Ethnic Origin 115 Child-bearing Age 115 Have Health Coverage 115 Had Mammogram 115 Had Breast Exam 115 Smoking Status 116 Body Mass Index 116 Had Hysterectomy 116 Cervical Cancer Screening Advice 116

PAGE 6

iii Chapter Four: Results, continued 116 Research Question 2 116 Research Question 3 117 Research Question 4 117 Chapter Five: Summary, Discussion, Conclusions, Recommendations 118 Summary 118 Research Question 1 119 Research Question 2 120 Research Question 3 121 Research Question 4 122 Discussion 123 Implications for Public H ealth Intervention Strategies 123 Evaluation of the Preventive Health Model 128 Conclusions 132 Limitations of the Study 132 Contribution of the Study to the Body of Knowledge 133 HPV DNA Screening, HPV Vacci ne Trials and Public Health Intervention Strategies 135 Recommendations for Future Research 141 References 145 Appendices 178 Appendix A: Panel of Experts 179 Appendix B: The Stepwise L ogistic Regression Procedure For Model Selection 180 Model A: Adjusted Odds Ratio Estimates for the Socio-demographic-(9) C ovariates of Compliance with the 1999 Cervical Screening Guidelines of the ACS 180 Model B: Adjusted Odds Ratio Estimates for the Socio-demographic-(6) C ovariates of Compliance with the 1999 Cervical Screening Guidelines of the ACS 180 Model C: Adjusted Odds Ratio Estimates for the Socio-demographic-(6) and Screening-(1) Covariates of Compliance with the 1999 Cervical Screening Guidelines of the ACS 181 Model D: Adjusted Odds Ratio Estimates for the Socio-demographic-(6) and Screening-(2) Covariates of Compliance with the 1999 Cervical Screening Guidelines of the ACS 181 Model E: Adjusted Odds Ratio Estimates for the Socio-demographic-(5) and Screening-(2) Health Profile-(1) Covariates of Compliance with the 1999 Cervical Screening Guidelines of the ACS 182

PAGE 7

iv Appendix B: The Stepwise L ogistic Regression Procedure For Model Selection, continued 182 Model F: Adjusted Odds Ratio Estimates for the Socio-demographic-(5) and Screening-(2) Health Profile-(2) Covariates of Compliance with the 1999 Cervical Screening Guidelines of the ACS 182 Model G: Adjusted Odds Ratio Estimates for the Socio-demographic-(5) and Screening-(2) Health Profile-(3) Covariates of Compliance with the 1999 Cervical Screening Guidelines of the ACS 183 Model H: Final Model of Adjusted Odds Ratio Estimates for Covariates of Compliance with the 1999 Cervical Cancer Screening Guidelines of the ACS 183 Model I: Final Model of Adjusted Odds Ratio Estimates for Covariates of Compliance with the 1999 Cervical Cancer Screening Guidelines of the ACS with Interaction Terms 184 Appendix C: Exemption Certif ication for Protocol No. 99341 185 About the Author End Page

PAGE 8

v List of Tables Table 1 Operationalization of the Constructs of the Preventive Health Model 79 Table 2 Variables Associated with Compliance with the 1999 Cervical Cancer Screening Guidelines of the ACS in the Study Sample 97 Table 3 Census Bureau Regi ons of the United States with State FIPS Codes 99 Table 4 Descriptive Statisti cs of the Study Sample 103 Table 5 Sources of Health Care Coverage for the Study Sample 105 Table 6 Summary of Selected De scriptive Unvariate Statistics Skewness and Kurtosis for Variables in the Study Sample 106 Table 7 Summary of the Correlation Analyses: Assessment of the Strength of Association betw een each of the Variables in the Study Sample 108 Table 8 Chi-Square Analyses: Measurement of the Bivariate Associations between each Independent Variable and the Dependent Variable in the Study Sample 109 Table 9 Models of Unadjusted Odds Ratio Estimates for Covariates of Compliance with the 1999 Cervical Cancer Screening Guideline of the ACS 110 Table 10 Model H: Final Model of Adjusted Odds Ratio Estimates for Covariates of Compliance with the 1999 Cervical Cancer Screening Guidelines of the ACS 112

PAGE 9

vi Table 11 Model I: Adjusted Odds Ra tio Estimates for Covariates of Compliance with the 1999 Cervical Cancer Screening Guidelines of the ACS with Interaction Terms 113 Table 12 The Modified Preventive Health Model 131

PAGE 10

vii Cervical Cancer Screening Disparities in an Ethnically Diverse Population of Women Residing in the United States in 1999: A S econdary Analysis of Data from the 1999 Behavioral Risk Factor Surveillance System Chodaesessie Wellesley-Cole Morgan ABSTRACT Black American women have the highest screening rates for cervical cancer among all the ethnic groups in the United States. Even though evidence from the literature suggests that the num ber of deaths from cervical cancer in the United States could be reduced by preventive screening, this particular minority population still suffers disproportionately higher mortality from the di sease than the other minority and majority populations in the United Stat es. This study was proposed to investigate cancer screening disparities among different subpopulations of women residing in the United States during 1999, and to recommend public health interv entions that could potentially increase cervical cancer screening rates, thereby decr easing differential mortality rates for cervical cancer among these subpopulations. The Preventive Health Model in conjunc tion with data from the 1999 Behavioral Risk Factor Surveillance System was used to identify the covariates of cervical cancer screening behavior in an et hnically diverse population of American women residing in the United States during the specified timefra me. Univariate, bivariate and multivariable logistic regression procedures were used to evaluate the a ssociation between each one of

PAGE 11

viii the independent variables and the dependent variable (compliance with the 1999 cervical screening guidelines of the American Cancer Society). One of the major findings of this stu dy was that Black, White and Hispanic American women were more similar in thei r screening behavior than dissimilar. The study also showed that the disparity in cer vical cancer screening behavior in this population is in age, rather than in ethnic origin. Black, White and Hispanic American women of child-bearing age (18-44 years) were more likely to be compliant with the 1999 cervical cancer screening guidelines of th e American Cancer Society, than Black, White and Hispanic American women who were not of child-bearing age (45 to 64 years). Implications for public health intervention studies are discussed, and recommendations made for future research in this area of cerv ical cancer screening behavior.

PAGE 12

1 CHAPTER ONE: INTRODUCTION Statement of the Problem Evidence from the literature suggests that the overall health status of the American nation has improved during the past 25 years. However, disparities in health status between minority and majority populat ions in the United States still exist (American Cancer Society, [ACS], 2004; DHS S. Race and health: Cancer management, 1999). The health status of individuals and th e health status of th e communities within which they live and work contribute to the hea lth status of the nati on as a whole. It is, therefore, imperative that grea ter attention be paid to hea lth care of minority populations because as these populations increase in size, th eir contribution to the ill-health status or the well-being status of the entire nation wi ll also increase (NIH Guide: Understanding and eliminating minority health dispar ities, 2000; Woodward & Kawachi, 2000). A number of socio-demographic, socio-cultura l, and socio-economic variables have been associated with disparities in health betw een minority and majority populations in the United States. These variables include region of residence, ethnic origin, age, gender, poverty, health behavior, access to health care geographical location and patient-primary care provider communication (ACS, 2004; DHSS. Race and health: Cancer management, 1999). Cancer is one of the diseases of wh ich some minority populations bear a disproportionate amount of the burden. Black Americans have one of the highest

PAGE 13

2 proportions of the cancer burden among all the et hnic groups in the United States. Cancer is the second leading cause of death in the United States and disparities exist in both incidence and mortality rate s between the minority and majority populations. The rates for cancer have been age-adjusted to the 2000 U.S. standard population The age-adjusted, incidence rates for all invasive cancers at all sites from 1997-2001 for Black Americans is 515.8 cases per 100,000; for Hispanic American s the rate is 351.3 cases per 100,000; for White Americans the rate is 479.6 cases pe r 100,000; for American Indians/Alaska Natives the rate is 237.7 cases per 100,000, for As ians/Pacific Islanders the rate is 336.6 cases per 100,000, and for all ethnic groups the rate is 470.3 cases per 100,000. Since Black Americans, White Americans, American Indians/Alaska Natives or Asians/Pacific Islanders are groups not mutually exclusiv e of Hispanic Americans, the observed incidence rates for Hispanic Americans are undere stimates of the true rates (Ries et al., 2004). The age-adjusted mortality rates for all invasive cancers at all sites for Black Americans is 252.5 deaths per 100,000; for Hisp anic Americans the rate is 136.5 deaths per 100,000; among White Americans the rate is 196.9 deaths per 100,000; for American Indians/Alaska Natives the rate is 134.9 d eaths per 100,000, for Asians/Pacific Islanders the rate is 122.0 deaths per 100,000, and for a ll ethnic groups the rate is 199.8 deaths per 100,000. Since Black Americans, White Americans, American Indians/Alaska Natives or Asians/Pacific Islanders are groups not mutually exclusive of Hispanic Americans, the observed mortality rates for Hispanic Ameri cans are underestimates of the true rates (Ries et al., 2004).

PAGE 14

3 Cervical cancer in the second most hi ghly incident cancer in women throughout the world (Bosch & de Sanjose, 2003; Gray & Walzer, 2004; Munoz et al., 2004; World Health Organization [WHO] 2002) It is also the second mo st common cause of cancer mortality in women worldwide (Adams, et al., 2001). Cervical cancer is an important public health problem because of the burden of the disease and the potential for prevention by means of screening (Gray & Walzer, 2004). In the United States, the probability of developing invasive cervical cancer is 0.16 (1 in 632) for women from birth to 39 years of age a nd 0.31 (1 in 322) for women age 40 to 59 years of age (National Cancer Institute [N CI] 2003). Cervical cancer is the second leading cause of cancer mortality, among women aged 20 to 39 years, and among women aged 40 to 59 years, it is the fifth leading cause of can cer mortality (Greenlee, Murry, Bolden & Wingo, 2000). The age-adjusted mortality rate for cervical cancer for wome n 20 to 54 years of age is 2.7 cases per 100,000, and the age-adjust ed mortality rate for cervical cancer women 55 to 64 years of age is 5.7 cases per 100,000 (NCI, 2005). Disparities exist in both the incidence and mortality rates for cervical cancer between the minority and majority populati ons. Hispanic American women have the highest age-adjusted incidence rates for cervical cancer with 16.2 cases per 100,000; followed by Black American women with 11.8 cases per 100,000; followed by Asians/Pacific Islanders w ith 9.5 cases per 100,000; followed by White American women with 8.9 cases pre 100,000 and American Indi ans/Alaska Natives with 6.0 cases per 100,000. The age-adjusted incidence rates for cer vical cancer for all ethnic groups is 9.3 deaths per 100, 000 (Ries et al., 2004).

PAGE 15

4 Black American women have the highest age-adjusted mortality rate for cervical cancer among all the ethnic groups with 5.6 cervical cancer d eaths per 100,000. For Hispanic American women the rate is 3.6 cervical cancer deaths per 100,000; Asians/Pacific Islanders and American Indians/Alaska Natives followed with 2.8 cervical cancer deaths per 100, 000 and White Ameri can women with 2.6 cervical cancer deaths per 100,000. The age-adjusted mortality rates fo r cervical cancer for all ethnic groups is 2.9 deaths per 100, 000 (Ries et al., 2004). Evidence from the literature suggests that the number of deaths from cervical cancer in the United States could be reduced by preventive screening, a public health intervention strategy (CDC, 1998-1999; DHHS. Race and health: Cancer management, 1999; Franco, Duarte-Franco, & Ferenczy, 2001; Holmquist, 2000; Klaes, et al. 2001; Koop, 1997; MMWR, 1997; Runowicz & Fi elds, 1999; Sasieni & Adams, 1999; Schiffman, Brinton, Devessa & Fraumeni, 1996). Data from the 1999 Behavioral Risk Factor Surveillance System show that Black American women have the highest prevalence of preventive cervical cancer screening behavior among all the ethnic groups in the United States (CDC/NCCDPHP. BRF SS: Prevalence data, 2002). Other evidence from the literature indicates that cervical cancer screeni ng rates are higher among Black American women when compared to White American women (Makuc, Fried & Kleinman, 1989; Martin, Parker, Wingo & H eath, 1996). It is well documented in the breast cancer screening lite rature that primary care provider recommendation about obtaining screening mammograms is one of th e strongest covariates of adherence to breast cancer screening guidelines (American Cancer Society [ACS], 2004; Lippert, Eaker, Vierkant & Remington, 1999; Mande lblatt & Yabroff, 2000; O’Malley, Earp,

PAGE 16

5 Hawley, Schell, Mathews & Mitchell, 2001; Roetzheim, Fox, Leake & Houn, 1996). However the relationship between adherence to cervical cancer scr eening guidelines and primary care provider recomme ndation about obtaining a smear test has not been well studied (Hiatt, Klabunde, Breen, Sw an & Ballard-Barbash, 2002). In summary, although evidence from the lite rature suggests that the overall health status of the American nation has improved during the past 25 years, disparities in health status between minority and majority populations in the United States still exist (ACS, 2004; DHSS. Race and health: Cancer manage ment, 1999). Disparities exist in both the incidence and mortality rates for cervical cancer between the minority and majority populations. Even though Black American wo men have the highest screening rates among all the ethnic groups (CDC/NCCDPH P. BRFSS: Prevalence data, 2002); and evidence from the literature suggests that the number of deaths from cervical cancer in the United States can be reduced by pr eventive screening (CDC, 1998-1999; DHHS. Race and health: Cancer management, 1999; Franco, Duarte-Franco, & Ferenczy, 2001; Holmquist, 2000; Klaes, et al. 2001; Koop, 1997; MMWR, 1997; Runowicz & Fields, 1999; Sasieni & Adams, 1999; Schiffman, Br inton, Devessa & Fraumeni, 1996); this particular minority population still suffers di sproportionately higher mortality from the disease than the other minority and majority populations in the United States (Ries et al., 2004). Rationale for the Study This study was proposed to investigat e cancer screening disparities among different subpopulations of women residing in the Un ited States during 1999, and recommend public health interventions that could potentially increase cervical cancer

PAGE 17

6 screening rates, thereby decreasing differentia l mortality rates for cervical cancer among these subpopulations (CDC, 1998-1999; DHHS. Race and health: Cancer management, 1999; Franco, Duarte-Franco, & Ferenczy, 2001; Holmquist, 2000; Klaes, et al. 2001; Koop, 1997; MMWR, 1997; Runowicz & Fi elds, 1999; Sasieni & Adams, 1999; Schiffman, Brinton, Devessa & Fraumeni, 1996). The principal investigator intends to use an appropriate theoretical model or theory of health behavior to provide an evidencebased, scientific rationale (a ) to guide the research on di fferences in cervical cancer screening behavior among the ethnically dive rse groups of women residing in the United States, and (b) to guide the ensuing discus sion on designing and de veloping appropriate interventions to increase cervi cal cancer screening in thes e groups (Glanz & Rimer, 1997; Race and health, 1999; Turnock, 1997). The Preventive Health Model has not been used to explain preventive cervical cancer screening behavior in women. The Pr eventive Health Model in conjunction with data from the 1999 Behavioral Risk Factor Su rveillance System will be used to identify the covariates of cervical can cer screening behavior in th is ethnically diverse population of American women. The principa l investigator is also interested in investigating the magnitude of the associati on between cervical cancer scr eening behavior and ethnic origin of the women in the United States, because evidence from the literature indicates that cervical cancer screening rates ar e higher among Black American women when compared to White American women (Makuc Fried & Kleinman, 1989; Martin, Parker, Wingo & Heath, 1996). One of the most powerful measures for reducing the mortality rates of cervical cancer, especially in minority populations, is by increasing cervical cancer screening

PAGE 18

7 (CDC, 1998-1999; DHHS. Race and health: Cancer management, 1999; Franco, DuarteFranco, & Ferenczy, 2001; Holmquist, 2000; Klaes, et al. 2001; Koop, 1997; MMWR, 1997; Runowicz & Fields, 1999; Sasieni & Ad ams, 1999; Schiffman, Brinton, Devessa & Fraumeni, 1996). Although the literature id entifies physician recommendation as the most powerful predictor for breast cancer sc reening (Mandelblatt & Yabroff, 2000), little is known about the effect of physician r ecommendation on cervical cancer screening behavior, especially in minority popul ations (Champion & Menon, 1997; Hiatt, Klabunde, Breen, Swan & Ballard-Barbash, 2002). It is critical, given the pu blic health orientation of this study, to investigate whether membership of an ethnic group with its distinctive social and cultural traditions could have a moderating effect on the associ ation between primary care provider advice about cervical cancer screening and complia nce with the 1999 cervi cal cancer screening guidelines of the ACS. The effect of ethni c origin as a modera ting variable in the association between primary care provider ad vice about cervical cancer screening and compliance with the 1999 cervical cancer screening guidelines of the ACS will also be investigated. Purpose of the Study The primary purpose of the study is to use the Preventive Health Model in conjunction with data from th e 1999 Behavioral Risk Factor Surveillance System to: (a) investigate the association of selecte d, significant, covariates of cervical cancer screening behavior with compliance with the 1999 cervical cancer screening guidelines of the ACS in an ethnically diverse populat ion of American wome n, (b) investigate the magnitude of the association of ethnic orig in and compliance with the 1999 cervical

PAGE 19

8 cancer screening guidelines of the ACS, (c) investigate the association between primary care provider advice about cervical cancer screening and compliance with the 1999 cervical cancer screen ing guidelines of the ACS in th is population, and (d) investigate whether the association betw een primary care provider advi ce about cervical cancer screening and compliance with the 1999 cervical cancer scre ening guidelines of the ACS is moderated by ethnic origin. The secondary purpose of this study is to examine the utility of the Preventive Health Model as a theoretical model in (a) guiding the research into cervical cancer screeni ng behavior in this ethnically diverse population of women, and (b) guiding the ensuing discussion on designing and developing appropriate intervention strategies for incr easing cervical cancer screenin g in this population (Glanz & Rimer, 1997; Race and health, 1999; Turnock, 1997). Theories and Theoretical M odels of Health Behavior Even though many theories a nd theoretical models of h ealth behavior appear to have similar explanatory or pr edictive powers, no single theo ry, or theoretical model can be used to explain all of th e processes involved in comple x, health behavioral change, such as cervical cancer screening (Jennings 1997; Prochaska, R eading & Evers, 1997; Weinstein, 1993). The Health Belief Mode l (Rosenstock, Strecher & Becker, 1988), Social Cognitive Theory (Bandura 1986), th e Theory of Reasoned Action (Fishbein & Ajzen, 1975; Ajzen & Fishbein, 1980), and the Transtheoretical Mode l (Prochaska, 1997) have all been used to try to explain why women do or do not obtain screening tests for cervical cancer (Rimer, 1996). The ensuing discussion about the applicati on of these theories and models will be in the context of cervical cancer screen ing behavior. The Health Belief Model

PAGE 20

9 hypothesizes that cervical cancer screening behavior depends on a woman’s need to avoid cervical cancer and her belief that havi ng a cervical cancer test will help to detect early symptoms of the disease if it is pres ent. The constructs of the model include a woman’s perceptions about: (a) her own suscepti bility to cervical cance r, (b) the severity of the consequences of leavi ng cervical cancer untreated, (c ) the benefits of compliance with the recommended guidelines for screening for the disease, and (d) the barriers of adhering to the recommended screening guidelines (Janz & Becker, 1984). These constructs were originally proposed to explai n an individual’s read iness to engage in a particular health action, such as cervi cal cancer screening (G lanz & Rimer, 1997). Hochbaum (1958) proposed the construct of cues to action to explain how an individual’s readiness to take part in a cervical cancer screening pr ogram, could be activated by personal factors and environmental influences Rosenstock, Strecher and Becker (1988) added the construct of self-efficacy to the Heal th Belief Model to increase its explanatory power. Bandura (1986) explained human behavior in terms of Social Cognitive Theory which posits that behavior, personal factor s (including cognition), socio-cultural and socio-demographic factors all in teract in a reciprocal manner to determine an individual’s behavior. The constructs of this theory in clude the individual wo man’s capacity to: (a) think through the conseque nces of compliance with, or non-compliance with the cervical cancer screen ing guidelines, (b) learn more a bout cervical cancer screening by observing other women (vicarious learning), (c) have the conf idence to comply with to the recommended screening guidelines, (d) ha ve the confidence to overcome any barriers

PAGE 21

10 presented, and comply with the recommende d screening guidelines, and (e) regulate her own cervical cancer screening behavior. The Theory of Reasoned Action posits that behavioral intention is the most important covariate of behavior. Therefore, in the context of cervical cancer screening behavior, a woman’s intention to comply with recommended cervical cancer screening guidelines is the most important covariate as to whether she will or will not perform this behavior. The direct covariates of a woman’ s intention to comply with the recommended screening guidelines are: (a) her attitude towards compliance with the health behavior, (b) the subjective norm associated with cervi cal cancer screening behavior and, (c) her perceptions about her own self control co mpliance with the r ecommended screening guidelines (Ajzen, 1991). The basic premise of Transtheoretical M odel is that behavi oral change is a process involving progress thr ough five stages of change wh ich are: (a) precontemplation (a woman in this stage of behavioral cha nge has not even thought about screening for cervical cancer); (b) contempl ation (a woman in this stag e has thought about getting a cancer screening test); (c) preparation (a woma n in this stage of be havioral change has made arrangements for obtaining a cervical cancer screening test); (d) action (in this stage of behavioral change a woman has been screened once for cervical cancer) and, (e) maintenance (in this final stage of behavi oral change a woman is being screened regularly for cervical cancer) (Pro chaska, et al., 1994; Rimer, 1996). In summary, the Health Belief Model (Rosenstock, Strecher & Becker, 1988), Social Cognitive Theory (Bandura 1986), th e Theory of Reasoned Action (Fishbein & Ajzen, 1975; Ajzen & Fishbein, 1980), and the Transtheoretical Mode l (Prochaska, 1997)

PAGE 22

11 have all been used individually to try to predict and understand se lf-initiated health behavior such as cervical can cer screening (Rimer, 1996). Ev en though these theories and theoretical models of health behavior appear to have similar expl anatory or predictive powers, none of them can be used individually to explain all of the processes involved in complex, health behavioral change, such as cervical cancer scr eening (Jennings, 1997; Prochaska, Reading & Evers, 1997; Weinstein, 1993). The Preventive Health Model and Constructs from the Health Belief Model, the Theory of Reasoned Action and Social Cognitive Theory The Preventive Health Model was de veloped specifically to explain and understand the covariates of preventive h ealth behavior (Myers, Ross, Jepson, Wolf, Balshem, Millner, & Leventhal, 1994). It is a theoretical model which has been derived from constructs of the Health Belief Mode l, the Theory of Reasoned Action and Social Cognitive Theory that are thought to be important in explaining and understanding preventive health behavior, such as cancer sc reening (Carver & Scheie r as cited in Myers, et al., 1994). This model has been used in st udies to determine: (a ) colorectal cancer screening behavior in men and women (M yers et al., 1994, Watts, Vernon, Myers & Tilley, 2003), and (b) prostate cancer screen ing behavior in African American men (Gwede, 2001; Myers, Wolf, Mckee, McGo ry, Burgh, Nelson, & Nelson 1996). In the context of cervical cancer scr eening behavior, the Preventive Health Model posits that a broad set of factors influence a woman’s de cision to be screened for cervical cancer (Myers et al., 1994). The first set of variables may be described as the background factors which may include age, gender, ethnic origin, educati on, and past preventive screening behavior.

PAGE 23

12 These background factors may be viewed as the socio-demographic context within which a woman makes decisions about her health care. The constr uct of background factors is derived from the Health Belief Mo del (Strecher & Rosenstock, 1997). The second set of variables may be described as representation factors. This construct is also derived from the Hea lth Belief Model and includes a woman’s perceptions about: (a) her own su sceptibility to cervical cancer, (b) the severity of the consequences of leaving cervical cancer untre ated, (c) the benefits compliance with the recommended guidelines for screening for the di sease, and (d) the ba rriers of compliance with the recommended screening guidelines including the practical convenience of compliance with the screening guidelines (Jan z & Becker, 1984; Strecher & Rosenstock, 1997). Social influence factors form the third set of variab les. These factors include a woman’s relationship with her health care professional, and the social support for cervical cancer screening beha vior provided by significant indi viduals in her life. This particular construct is derived from Social Cognitive Theory which posits that behavior, personal factors (including cogni tion), socio-cultural and so cio-demographic factors all interact in a reciprocal manner to determine an individual’s pattern of health behavior. (Baranowski, Perry & Parcel, 1997). Program factors comprise the fourth set of fact ors in the model. A woman may be exposed to educational interventions, su ch as primary care provider advice or recommendations, the purpose of which is to motivate and reinforce cervical cancer screening behavior (Myers, et al., 1994). This construct is derived from both Social Cognitive Theory and the Health Belief Model and refers to the contacts made by the

PAGE 24

13 health services system for the purpose of motivating and reinforcing a given preventive health behavior (Baranowski, Perry & Pa rcel, 1997; Strecher & Rosenstock, 1997). According to the Theory of Reasoned Action (Fishbein & Ajzen, 1975; Ajzen & Fishbein, 1980), the combination of backgr ound, representation, social influence and program factors could influence a woman’s intention towards having a cervical cancer screening test. The Preventive Health Model posits that background, psychological representation, social influen ce and program factors influe nce a woman’s intention to participate in a cervical cancer screeni ng program, and also influence a woman’s behavior in obtaining a cervi cal cancer screening test (Myers, et al., 1994). The Preventive Health Model has not b een used to examine cervical cancer screening behavior in women. The investigat or proposes to contribute to the body of knowledge by using the model to guide the research into compliance with the 1999 cervical cancer screenin g guidelines of the ACS in an ethnically diverse population of women residing in the United States. The Prev entive Health Model is a theoretical model within which covariates associated with cervical cancer screen ing behavior can be studied, so that interventions can be focu sed on those that are most likely to reduce cervical cancer mortality rates, by increasi ng cervical cancer behavior, especially among the Black American women (Glanz & Rimer, 1997; Race and health, 1999; Turnock, 1997). Assumptions of the Study The principal investigator will use data previously collected by researchers for the 1999 Behavioral Risk Factor Surveillance Syst em (BRFSS), a collaborative project of the Centers for Disease Control and Prevention (C DC), and the U.S. states and territories.

PAGE 25

14 The BRFSS is an on-going da ta collection (survey) pr ogram designed to measure behavioral risk factors in the adult population 18 years of age and over, living in households with telephones (CDC/N CCDPHP. Overview: BRFSS 1999, 2002). The research methods and data anal ysis techniques employed by the BRFSS researchers are explained briefly in Chapter 3. For the purpose of this study it is assumed that: 1. The telephone interviewers followed sta ndardized procedures when reading the questions to the study partic ipants over the telephone. 2. The questions were appropriate in design and content to elicit the required responses. 3. The questions were understood by the study pa rticipants, answered honestly, and to the best of their ability. 4. The managers checked the questionnai res for errors (CDC/NCDPHP/DASH Behavioral Risk factor Surveillance System Training Guide, 2000). Delimitations of the Study The following delimitations, under the contro l of researchers at the Centers for Disease Control and Prevention (CDC), we re imposed upon the study participants. The study participants (a) must be 18 years or ol der, (b) must reside in households with telephones, (c) must be non-inst itutionalized civilian s, (d) must be able to understand what is said to them over the telephone, a nd (e) must be underst ood over the telephone. The delimitations under principal investig ator control are th at: (a) the study sample has been selected from the populati on interviewed by the CDC researchers, and (b) the study variables have b een selected from those alrea dy chosen by the researchers at the CDC for inclusion in the 1999 BRFSS database.

PAGE 26

15 Limitations of the Study The limitations of the study are such that the results may not be generalizable to (a) adults who do not have telephones, (b) adults who have telephones but are institutionalized, (c) adults who are not ci vilians, (d) adults who cannot hear or understand what is being conveyed to them ove r the telephone, (e) adults who cannot be understood over the telephone. Definitions of Terms The terms defined below are specific to this particular study and will be expanded, where appropriate, within the context of the study. Acculturation “Process of cultural ch ange in which members of a group or one group assimilate(s) various cultural patterns from a different group” (On-line Medical Dictionary, 2004). American Indian or Alaska Native “A person having origins in any of th e original peoples of North and South America (including Central America), and who maintains tribal affiliation of community attachment” (Office of Management and Budget, [OMB], 1997). Asian American “A person having origins in any of the or iginal peoples of the Far East, Southeast Asia, or the Indian subcontinent including, fo r example Cambodia, China, India, Japan, Korea, Malaysia, Laos, Pakistan, the Philip pine Islands, Thailand and Vietnam” (OMB, 1997).

PAGE 27

16 Ancestry “A line of descent” (Webster’s New American Dictionary, 1995). Asymptomatic “Showing or causing no symptoms of disease” (Dorland’s Illustrated Medical Dictionary, 2000). Black American “A person having origins in any of the original peoples of sub-Saharan Africa” (OMB, 1997). Cancer “A general term for more than 100 diseases that are characterized by uncontrolled, abnormal growth of cells. Cancer cells can spread locally or through the bloodstream and lymphatic system to ot her parts of the body” (On-line Medical Dictionary 1997-2004). Cancer, cervix “Cancer of the entrance to the womb (uterus)” (On-line Me dical Dictionary 19972004). Carcinoma “A malignant new growth that arises fr om epithelium, found in the skin, or more commonly, the lining of body organs, for exam ple: breast, prostate, lung, stomach or bowel. Carcinomas tend to infiltrate into adjacent tissue and spread (metastasize) to distant organs, for example: to bone, liver, l ung or the brain” (On-lin e Medical Dictionary 1997-2004).

PAGE 28

17 Concept (Construct) 1. “An abstract idea or notion” 2. “An explanatory variable or principl e in a scientific system” (On-line Medical Dictionary 1997-2004). Dependent Variable “A dependent variable is a variable the value of which is dependent on the effect of other variable(s) – independent variab le(s) – in the rela tionship under study. A manifestation or outcome whose variation we seek to explain or account for by the influence of independent variables” (Last, 1995). Epidemiologist “One who specializes in epidemiol ogy” (Dorland’s Il lustrated Medical Dictionary, 2000). Epidemiology “The science concerned with the study of factors determining and influencing the frequency and distribution of disease, injur y, and other health-rela ted events and their causes in a defined human population for purpos es of establishing programs to prevent and control their development and spread. Also the sum of knowledge gained in such a study” (Dorland’s Illustrate d Medical Dictionary, 2000). Ethnic Group “A social group characterized by a dis tinctive social and cultural tradition, maintained within the group from generation to generation, a common history and origin, and a sense of identification with the group. Members of the group have distinct features

PAGE 29

18 in their way of life, shared experiences and often a common genetic heritage. These features may be reflected in their hea lth and disease experience” (Last, 1995). Gynecological “Pertaining to gynecology” (Dorland’s Illustrated Medical Dictionary, 2000). Gynecologist “A medical doctor who sp ecializes in gynecology and diseases affecting the female reproductive system” (Online Medical Dictionary, 2004). Gynecology “A branch of medicine dealing with the diagnosis and treatment of disorders affecting the female reproductive organs and genital tract” (On-lin e Medical Dictionary, 2004). Health “The state of being hale, sound, or whole, in body, mind, or soul; especially, the state of being free from phys ical disease or pain” (Online Medical Di ctionary 19972004). Health Behavior “Behaviors expressed by individuals to pr otect, maintain or promote their health status. For example, proper diet, and appropr iate exercise are activities perceived to influence health status. Life st yle is closely associated with health behavior and factors influencing life style are so cio-economic, educational and cultural” (On-line Medical Dictionary 1997-2004).

PAGE 30

19 Health Problem “A health problem is a situation or co ndition of the people (expressed in health outcome measures such mortality, morbidity, or disability) that is considered undesirable and is likely to exist in the future” (Turnock, 1997). Health Problem Analysis “Health problem analysis is a framework for analyzing health problems to identify their covariates and contributing fact ors so that interventions can be targeted rationally towards those factors most likely to reduce the level of the health problem” (Turnock, 1997). Health Status Indicators “The measurement of the health status for a given population using a variety of indices, including morbidity, mortality and av ailable health resour ces” (On-line Medical Dictionary 1997-2004). Hispanic American “A person of Cuban, Mexican, Puerto Rica n, South or Central American, or other Spanish culture or origin” (OMB, 1997). Incidence “Incidence more generally refers to the number of new events, e.g., new cases of a disease in a defined population, within a specified period of time” (Last, 1995). Human Papillomavirus “A disease caused by the human papillomavirus characterized by a wart-like growth on the genitalia (for example penis or vulva). The infection is most commonly transmitted sexually” (On-line Medical Dictionary, 1997-2003).

PAGE 31

20 Incidence Rate “The incidence rate is the rate at which new events occur in a population. The numerator is the number of new events that occur in a defined period; the denominator i.e. the population at risk of experiencing the event dur ing this period, sometimes expressed as person-time” (Last, 1995). Independent Variable “An independent variable is the characteris tic being observed or measured that is hypothesized to influence an event or manifest ation (the dependent variable) within the defined area of relationships under study; th at is the independen t variable is not influenced by the event or manifestation but may cause or contribute to variation of the event or manifestation” (Last, 1995). Inequalities (Disparities) in Health “Inequalities in health refer to the vi rtual universal phenomenon of variation in health indicators the incidence, prevalen ce, mortality, morbidity, survival, burden of disease and other adverse health conditions that exist among specific population groups in the United States etc” (Last, 1995). Intervention Studies “Epidemiological investigations desi gned t test hypothesized cause-effect relationships by modifying the supposed cau sal factor(s) in the study population” (Online Medical Dictionary, 1997-2004). Medicine “A scientifically-based discipline dedi cated to the prevention and treatment of disease and injury” (On-line Medical Dic tionary, 1997-2004).

PAGE 32

21 Minority Group “A minority group is a group characterized by a sense of sepa rate identity and awareness of status apart from a usually larger group of which it forms or is held to form a part as: (a) a body of nationals of a state forming a small, but appreciable part of the population of another usually neighbori ng state, (b) a group differing from the predominant section of a larger group in one or more characteristics (as in ethnic origin, language, culture, or religion) and as a result often subjected to differential treatment, especially discrimination, or (c) a group numerically smaller than other groups or a combination of other groups in a community but constituting the predominant element” (Webster’s Third New Inte rnational Dictionary, 1993). Model “A model is a formalized expression of a theory or the causal situation that is regarded as having generated the observed data” (Last, 1995). Moderator Variable “A factor that modifies the effect of a putative causal f actor under study. For example immunization status is an effect m odifier for the consequences of exposure to pathogenic organisms. Effect modification is detected by varying the selected effect measure for the under study across levels of another factor” (Last, 1995). Morbidity “Morbidity is any departure, subjective or objective, from a state of physiological or psychological well-being. The WHO Expert Committee on Health Statistics noted in its Sixth report (1959) that morbidity could be measured in term s of three units: (1) individuals who were ill; (2) the illnesses (p eriods or spells of illness) that these

PAGE 33

22 individuals experienced; and (3) the duration (days, weeks, etc. ) of these illnesses” (Last, 1995). Mortality Rate “The mortality rate is an estimate of the proportion of a population that dies during a specified period. The numerator is the number of indivi duals dying during the period; the denominator is the number in the population, usually estimated as the midyear population” (Last, 1995). Nurse, registered “A graduate nurse who has been legally authorized (registered ) to practice after examination by a state board of nurse examin ers or similar regulat ing authority, and who is legally entitled to use the designation RN ” (Dorland’s Illustrated Medical Dictionary, 2000). Nurse-Practitioner, advanced, registered “A registered nurse who is an expe rt in nursing practice and ensures ongoing development of expertise through clinical experience and continuing education” (Dorland’s Illustrated Me dical Dictionary, 2000). Obstetric, obstetrical “Pertaining to obstetric s” (Dorland’s Illustrated Medical Dictionary, 2000). Obstetrician “One who practices obstet rics” (Dorland’s Illustrate d Medical Dictionary, 2000). Obstetrics “A branch of medicine that deals with management of pregnancy, labor and the puerium” (Dorland’s Illustrate d Medical Dictionary, 2000).

PAGE 34

23 Pacific Islander “A person having origins in any of the or iginal peoples of Hawaii, Guam, Samoa, or other Pacific Islanders” (OMB, 1997). Papanicolaou (Pap) Smear Test “Microscopic examination of cells collected from the uterine cervix. The Pap test is used to detect cancer, changes in the uterine cervix that may lead to cancer, and noncancerous conditions, such as infection or inflammation” (On-line Medical Dictionary, 1997-2004). Predictive Value “In screening and diagnostic tests, th e probability that an individual with a positive test is a true positive (i.e., does have the disease) is referred to as the predictive value of a positive test. The predictive value of a negative test is the probability that an individual with a negative does not have the disease. The predictive value of a screening test is determined by the sensitivity and speci ficity of the test, and by the prevalence of the condition for which the te st is used” (Last, 1995). Prevalence “Prevalence is the number of events, e .g., instances of a given disease or other condition, in a given population at a designated time. When used without qualification, the term usually refers to the situation at a specified point in time (point prevalence). Point prevalence is a number not a rate” (Last, 1995). Prevalence “Rate” (Ratio) “The prevalence ratio is a proportion not a rate and is the total number of all individuals who have an attri bute or disease at a particular time (or during a particular

PAGE 35

24 period) divided by the population at risk of having the attribute or disease at this point in time or midway through the period” (Last, 1995). Preventive Health Services “Services designed for promotion of hea lth and prevention of disease” (On-line Medical Dictionary, 1997-2004). Preventive Health Behavior Preventive health behavior is any ac tivity undertaken by an individual who believes himself/herself to be healthy, for th e purpose of preventing or detecting illness in an asymptomatic state (Kasl & Cobb, 1966). Public Health “Public health is fulfilling society’s in terest in assuring c onditions in which people can be healthy” (Institute of Medicine [IOM], 1988, p. 7). Risk Factor “A risk factor is an aspect of personal behavior or lifestyle, an environmental exposure, or an inborn or inherited characte ristic, which on the basis of epidemiological evidence, is known to be associated with heal th-related condition(s) considered important to prevent “(Last, 1995). Screening “The U.S. Commission on Chronic Illne ss defined screening in 1951 as being the presumptive identification of unrecognized diseas e or defect by the application of tests, examinations or other procedures that can be applied rapidly. Scr eening tests sort out apparently well persons who probably have a disease from those who probably do not. A screening test is not intended to be diagnostic. Individua ls with positive or suspicious

PAGE 36

25 findings must be referred to their physicia ns for diagnosis and necessary treatment. Screening is an initial preventive examination only, and positive responders require a second diagnostic examination” (Last, 1995). Sensitivity and Specificity (of a screening test) 1. “Sensitivity is the proportion of truly diseased individuals in the screened population who are identified as diseased by the screening test Sensitivity is a measure of the probability that any give n case will be iden tified by the test – the true positive rate” (Last, 1995). 2. “Specificity is the proportion of trul y non-diseased individuals who are so identified by the screening test. It is a measure of the probability of correctly identifying a non-diseased i ndividual with a screening test – the true negative rate” (Last, 1995). Theory “In science, an explanation for some phenomenon which is based on observation, experimentation, and reasoning, and has b een confirmed over the course of many independent experiments. Theories are more cer tain than hypotheses, but less certain than laws” (On-line Medica l Dictionary, 1997-2004). Variable “A variable is any quantit y that varies. Any attribute, phenomenon, or event that can have different values” (Last, 1995) White American “A person having origins in any of the original peoples of Europe, the Middle East and North Africa” (OMB, 1997).

PAGE 37

26 CHAPTER TWO: LITERATURE REVIEW Etiology of Cervical Cancer Although the overall incidence rates and mortality rate s of the disease have declined in developed countries cervical cancer (can cer of the entrance to the uterus) is still a major public health problem thr oughout the world (Masood, 1997; On-line Medical Dictionary, 1997-2004). Women in le ss affluent regions of the world have higher rates of incidence and mortality, and lo wer rates of survival from this disease than do women from more affluent areas. Inci dence and mortality rates in the United States and Canada are among the lowest in the world and have declined during the last 50 years and evidence suggests that it is because of th e increased availability of cervical cancer screening programs in these countries (Fra nco, Duarte-Franco & Ferenczy, 2001). It is, therefore, imperative to ensure the ava ilability of cervical screening programs, particularly among women in lower socio-economic regions and groups, (Herrero, 1996). The most likely causative agent of cervical cancer and its precursors is the human papillomavirus (HPV) (Cuzick, 2000; Hoffman & Cavanagh, 1996; Palefsky, 2003; Rohan, Burk, & Franco, 2003). In addition to being a common sexually transmitted infection (STI), several case-c ontrol and cohort studies have shown that specific types of HPV are the causal agent for most squamous epithelial cancers of the female and male genital tracts (Munoz, & Bosch, 1992). The genital HPVs are classified as high, intermediate or low risk depending on the fr equency with which they are detected in

PAGE 38

27 cancers. The high-risk HPV genotypes are 16, 18, 31, and 45; the in termediate-risk HPV genotypes are 33, 35, 39, 45 51, 52, 56, 58, 59 and 68. HPV 6, 11, 26, 32, 40, 42, 43, 44, 53, 54, 55, 61, 66, 69, 73 and unnumbered types designated as Pap155, Pap291, CP4173, C8304, IS039, CP141, and W13b (among immune competent individuals) are almost never found in association with malignant tumo rs and are classified as low risk types (Jacobsen, et al., 2000; Lorincz, et al., 1992; Poljak, Marin, Seme & Vince, 2002). Results of several epidemiological st udies indicate that there is a strong association between HPV (types 16, 18, 31, 33, a nd 35) infection and invasive cervical cancer (Munoz, Bosch, de Sanjose, & Sh ah, 1994; Munoz, et al., 1992, Munoz, et al., 1993). A dose-response relationship has been reported between increasing estimated viral load and risk of cervical cancer (Mu noz, & Bosch, 1992). Although there is strong evidence to support the etiologi cal role of certain HPV geno types in the development of cervical cancer; evidence s uggests that these HPV genot ypes are necessary but not sufficient causes of cervical cancer, and other factors must also be present for cancer to progress to malignancy The other factors co uld be infectious agents, chemical cocarcinogens and factors relate d to the host immune respons e (Kiviat, 1996; Schiffman, 1992; Schiffman & Castle, 2003). Development of the Papanicolaou Screening Test In 1940, Dr. George Papanicolaou develope d a screening test for cancer of the uterine cervix in women. Precancerous changes, invisi ble to the naked eye, were detected in cellular debris from the cervix and fundus of the uterus, collect ed by the Papanicolaou vaginal smear technique. The cellular debris was smeared on glass slides, stained and then the cells examined for early detecti on of cancers of the ut erus (Papanicolaou &

PAGE 39

28 Traut, 1941). Studies had shown that the squa mocolumnar junction of the cervix was the area in which cancer developed most frequen tly than in any other area. The spatula technique was developed to study malignant ch anges in squamous cells from this site before they were exfoliated. A small spatula wa s used to scrape the surface of the tissues at the squamocolumnar junction. The cells a nd secretions were then placed on a glass slide, stained and examined for any malignanc ies (Ayre, 1947). Ther efore, in the late 1940s to early 1950s, the Papanicolaou smear e volved into a techni que to screen for precancerous cervical conditions, which were th en histologically confirmed and treated, with the intention of preven ting progression to invasive cervical cancer (Hoffman & Cavanagh, 1996). In April 1996, the Consensus Development Conference of Cancer of the Cervix, which was convened by the National Institutes of Health (NIH), concluded that failure to detect cervical disease in women could be due either to sampling error or to screening error (NIH, 1996). Although the panel conclude d that the conventional cervicovaginal (Papanicolaou) smear test was the best av ailable method for screening for cervical cancer, recommendations were made that ne w methods of collecting specimens and reading cervical cells were needed to reduce the number of false-ne gative results (NIH, 1996). The ThinPrep Papanicolaou test, a liquidbased cervical cytological preparation was approved by the United States Food a nd Drug Administration for use in 1996. The cervical cells are collected with a brush or similar collection instrument. The collection instrument is then rinsed in a vial of liquid preservative. The vial is sent to a laboratory where an automated thin-layer slide device prepares the slide for viewing (Carpenter &

PAGE 40

29 Davey, 1999). The ThinPrep Papanicolaou test improves the quality of cervical cancer screening by improving the detection and management of patients with cervical abnormalities, and by reducing the number of false-negative results (Linder, 1998; Linder & Zahniser, 1998; Roberts, Gurley, Thurloe, Bowditch & Laverty, 1997). Since its initial publication in 1989, the Beth esda System has been widely used by commercial and academic cytopathology laborator ies in the United States for reporting cervical/vaginal cytological diagnoses (Kurman, Malkas ian, Sedlis & Solomon, 1991; National Cancer Institute [NCI], 1989). Th e terminology for reporting the results of cervical cytology in the Bethesda System also included information on HPV as part of the cytological criteria (Rosetti, Gerli, Saa b, & Drano, 2000). In 2001 the Bethesda System terminology for reporting the results of cer vical cytology was updated to reflect the advances in the biological pe rspectives of cervical cancer and cervical cancer screening technology. In light of these advances, th e panel from the Bethesda 2001 Workshop concluded that cervical cytology is prim arily a screening test, and secondarily a diagnostic test (Solomon et al., 2002). Principles of Cancer Screening For screening to be implemented as public health policy, the disease has to be an important health problem (Miller, 1995). The main objective of screening for a disease is to reduce the mortality or burden of suffering from the disease in an individual, or in a population (Bjorge, Trope & Engeland, 1999; MacLean, 1996; Smith, 1999). However, several governing principles ha ve to be taken into consid eration before a screening program can be initiated within a given population:

PAGE 41

30 1. The biology and natural history of the disease sh ould be known. 2. The cancer should exist for a long time in an asymptoma tic, preclinical phase. 3. The disease should have a high prev alence and incidence in the given population. 4. The disease should have serious clinical consequences measured in mortality, morbidity and costs. 5. The mortality rate from the diseas e in the population of the screened individuals should be significantly lowe r in comparison to the mortality rate from the disease in an equivalent popul ation of unscreened individuals (Clark, 1995; Clark & Reintgen, 1996; Schi feling, Horton & Tafelski, 1997). Cervical cancer is the se cond leading cause of cancer mortality, among women aged 20 to 39 years, and among women aged 40 to 59 years, it is the fifth leading cause of cancer mortality in the United States (Greenlee, Murry, Bolden & Wingo, 2000), therefore, it is a disease of significant public health im portance. According to the governing principles of cancer screening (Clark & Reintg en, 1996) cervical cancer is amenable to screening for the following reasons. 1. The biology and natural history of the disease is known (Ayre, 1947). 2. The cancer exists in a slow-growing pr eclinical phase for a long period of time (Vilos, 1998). 3. Hispanic American women have the highe st age-adjusted incidence rates for cervical cancer with 16.2 cases pe r 100,000; followed by Black American women with 11.8 cases per 100,000. The ag e-adjusted incidence rates for

PAGE 42

31 cervical cancer for all ethnic groups is 9.3 deaths per 100, 000 (Ries et al., 2004). 4. Black American women have the highest age-adjusted mortality rate for cervical cancer among all th e ethnic groups with 5.6 cervical cancer deaths per 100,000. For Hispanic American wo men the rate is 3.6 cervical cancer deaths per 100,000 The age-ad justed mortality rates fo r cervical can cer for all ethnic groups is 2.9 deaths per 100, 000 (Ries et al., 2004). Helms and Melnikow (1999) reported that cost estimates for a preventive Papnicolaou smear test ranged from $23.79 to $123.48. Treatment for curable invasive cervical cancer was estimated at $12,893. For incurable invasive cervical cancer treatment cost was estimated at $128,927. 5. Evidence from the literature suggests that the number of deaths from cervical cancer in the United States could be reduced by preventive screening (CDC, 1998-1999; DHHS. Race and health: Ca ncer management, 1999; Franco, Duarte-Franco, & Ferenczy, 2001; Holm quist, 2000; Klaes, et al. 2001; Koop, 1997; MMWR, 1997; Runowicz & Fields, 1999; Sasieni & Adams, 1999; Schiffman, Brinton, Deve ssa & Fraumeni, 1996). There are many benefits to screening such as a reduction in the mortality and morbidity rates of cervical cancer, and savings in the overall cost of health care (Clark & Reintgen, 1996). However, there are also potenti ally harmful side eff ects associated with the invasive screening test which should be taken into consideration. These side effects include the inconvenience, anxiet y and discomfort associated with screening test and, the expense incurred in terms of time and finan ces. There are also potential side effects

PAGE 43

32 related to the screening test results. A fals e-positive test result co uld cause anxiety and incur unnecessary, expensive diagno stic evaluations such as a diagnostic screening and/or colposcopy. A false-negative result may lead to a false sense of security, with ensuing clinical symptoms of cervical cance r being dismissed because of the previous negative result. The consequences of these act ions could prove to be a fatal (Clark & Reintgen, 1996). Effectiveness and Cost-Effectivene ss of Cervical Cancer Screening For a screening program to be judged effec tive, the cervical cancer mortality rate for the screened population should be signi ficantly lower than for the unscreened population, and the stage distri bution of detected cancers in the screened population should be shifted towards lowe r stage cancers compared to higher stage cancers detected in the unscreened population (Clark & Rein tgen, 1996). Staging is the process of describing the spread of cancer from the site of origin, and is essential in determining the choice of therapy and assessing pr ognosis. If cancer cells are pr esent only in the layer of cells in which they have developed, and have not spread, the stage is in situ If the cancer cells have spread beyond the original layer of tissue, the cancer is referred to as being invasive (ACS, 2004). As the use of the Papanicolaou cytological screening has become more prevalent throughout the United States, and other countries such as th e United Kingdom and India, there has been a shift towards diagnosis of earlier clinical stages in patients with cervical cancer (Camilleri-Ferrante & Day, 1997; Ch ao, Becker, Jordan, Darling, Gilliland & Key, 1996; Gibson, Spiegelhalter, Jayant, Ra o, Nene & Dale, 1994; Shingleton, et al., 1996). In countries such as the United St ates and others that have implemented

PAGE 44

33 widespread screening programs, there has also been a reduction in the mortality from, invasive cervical cancer (Franco, Duarte -Franco, & Ferenczy, 2001; Holmquist, 2000; Klaes, et al. 2001; MMWR, 1997; Runowicz, & Fields, 1999; Sasieni & Adams, 1999). Cost-effectiveness analysis defines outcom es in natural units, such as cancers detected, survival rates, years of life gained, or tumor response to tr eatment. To be costeffective the cost of a screening progra m (screening costs, diagnostic evaluations, treatment cost of detected cancers, and value of lives lost to cancer deaths) should be less than the costs incurred by the unscreened population (diagnostic ev aluations, treatment cost of detected cancer and value of years lost to cancer deaths) (Clark & Reintgen, 1996). Evidence for the effectiv eness of Papanicolaou cytolo gical screening has been shown by epidemiological studies indicating that the risk of i nvasive cervical cancer is greater among women who have not been scre ened or who have been screened on an irregular basis, and that the ri sk increases with time since th e last normal smear, or with lower frequency of screening. Surveillance statis tics from different regions also indicate that the rates of cervical cancer incidence and mortality have decreased following the introduction of screening in the Scandinavian countries, Canada, and in the United States, with reductions in the inciden ce and mortality rates being pro portional to the coverage of the screening programs (Franco, Duar te-Franco & Ferenczy, 2001). Although no prospective, randomized, contro lled trial of screening for cer vical cancer has been done, the data from observational trials in many countries th roughout the world documenting the decreasing mortality rate for cervical cance r is evidence that screening for this tumor is both effective and cost-effective (Adami Ponten, Sparen, Bergstrom, Gustafsson & Friberg, 1994; Bocciolone, La Vecchia, Le vi, Lucchini & Frances chi, 1993; Gompel,

PAGE 45

34 1998; Helms & Melnikow, 1999; Hristova & Hakama, 1997; Ku, 1999; Mandelblatt, 1997; Sato, Matunaga, Tsuji, Yajima & Sa saki, 1999; Smith, Mettlin, Davis & Eyre, 2000; Suba, Nguyen, Nguyen & Raab, 2001; Shingleton, Patrick, Johnston & Smith, 1995). Cervical Cancer Screening Guidelines of the American Cancer Society from 1999 to 2005 The American Cancer Society (ACS) scre ening guidelines for early detection of cancer in asymptomatic people are assessed each year in the event that novel scientific evidence indicates a change, clarification or re-formulation of any of the existing screening guidelines. The guide lines are evaluated every five years whether or not there is new scientific evidence to imply that any change be made to the recommendations (ACS 2003; ACS 2004). From 1999 to 2002 the ACS recommended that “all women who are or have been sexually active or w ho are 18 and older should have an annual Papanicolaou test to check for changes in the uterine cervix and pelvic examination to check for changes in the uterus and ovaries. After three or more consecutive satisfactory examinations with satisfactory normal findi ngs, the Papanicolaou test may be performed less frequently…” (ACS, 1999, p. 31; ACS, 2000, p. 34; ACS, 2001, p. 35; ACS, 2002, p. 19). Since 2003, new cervical cancer screening guidelines, based on recent advances in cervical cancer research, have been de veloped. The ACS recommends that “cervical cancer screening should begin a pproximately three years afte r a woman starts to have vaginal intercourse, but no late r than 21 years of age. Sc reening should be done every year with conventional Papa nicolaou smear tests or ever y two years using the liquid-

PAGE 46

35 based tests. At or after 30 years of age, women who have had thr ee consecutive, normal results may elect to be screened every two to three years. Physicians may suggest that a woman screen more often if she has certain risk factors, such as HIV infection or a compromised immune system. Women who are 70 years of age and older, who have had three consecutive normal Papanicolaou test s in the last 10 years may chose to stop cervical cancer screening. Screening women who have had a total hysterectomy (with removal of the cervix) is not necessary unle ss the surgery was undertaken as a treatment for cervical cancer” (ACS, 2003, p. 48; ACS, 2004, p. 56). In 2005 the ACS recommendations for cer vical cancer screening suggest that “cervical cancer screening should begin approxim ately three years after a woman starts to have vaginal intercourse, but no later than 21 years of age. Screening should be done every year with conventional Papanicolaou smear tests or every two years using the liquid-based tests. At or after 30 years of age, women who have had three consecutive, normal results may elect to be screened every two to three years. Alternatively, cervical cancer screening with HPV DNA testing and or liquid-based cytology could be performed every three years. Physicians may suggest that a woman screen more often if she has certain risk factors, such as HIV infection or a compromised immune system. Women who are 70 years of ag e and older, who have had three consecutive normal Papanicolaou tests in the last 10 years ma y chose to stop cervical cancer screening. Screening women who have had a total hystere ctomy (with removal of the cervix) is not necessary unless the surgery was undertaken as a treatment for cervical cancer” (ACS, 2005, p. 60).

PAGE 47

36 Studies of Covariates of Cervical Cancer Screening Behavior: 1990 to 2005 A search of the literature database at the National Libr ary of Medicine (MEDLINE) and manual searches from 1990 to 2005 were executed on the topic of covariates of cervical cancer screening behavior. The principal investigator was interested in finding out which variables were significant in determ ining cervical cancer screening behavior in the studi es. In the context of cervica l cancer screening behavior, the Preventive Health Model posits that a br oad set of factors (variables) influence a woman’s decision to be scr eened for cervical cancer (M yers et al., 1994). These variables, extracted from a surv ey of the relevant literature, will then be used to build the theoretical constructs of the Preventive Health Model for this study. The results from these selected studies are summarized in the following section of the literature review. Hayward, Shapiro, Freeman and Corey (1988) evaluated the national trends in the provision of cancer screening preventive care from the 1986 Access to Care Survey, a large telephone survey of the U.S. popul ation. The researchers had two main study objectives: (a) to determine the proportion of American women who were currently receiving recommended cervical and breast cance r screening, and (b) to determine which groups were at greatest risk of not receiving these preventive measures. The study sample consisted of 4,659 wome n aged 20 years or older: 1,867 women aged 20 to 39 years, 721 women aged 40 to 49 years, 1,040 women aged 50 to 64 years, and 1,031 women aged 65 years or older. Th e sample was weighted, using methods similar to those used in th e National Health Interview St udy so that the findings were representative of the U.S. population (National Center fo r Health Statistics [NCHS], 1985). All interviewers were trained by a si ngle field coordinator. Households were

PAGE 48

37 selected by the Waksberg screening proce dure for random digit dialing (a procedure designed to eliminate the bulk of nonworki ng and no household telephone numbers while retaining equal selection probabilities for households) (Waksberg, 1978). The study participants were asked whether they had received the following screening procedures during the previous year: Papa nicolaou smear (for women aged 20 years or older), breast examination performed by a physician (for women aged 20 years or older), and mammogram (for women aged 40 years or older) (Hayward et al. 1988). The researchers reported that the ov erall response rate for households and individuals selected for interviews was 76%. The response rate of respondents who answered questions about pr eventive care ranged from 95.9% for Papanicolaou smears to 96.3% for mammograms. The Chi-square test of homogeneity was used to analyze the bivariate associations between each of the sc reening modalities and the nine independent variables: age, race/ethnicity, income, employ ment status, health in surance status, selfreported health status, rural ve rsus urban residence, educati on of respondent, and site of usual medical care. Logistic regression analys is showed age, health insurance status, income, education, employment status and race/e thnicity to be relate d significantly to the receipt of Papanicolaou cervical smears (Hayward et al. 1988). The researchers found that: (a) older women were at a higher risk of not receiving Papanicolaou smear tests than younger wome n, (b) uninsured women were less likely to have had Papanicolaou smears than women with insurance, (c) women with less education were less likely to have been sc reened regularly for cervical cancer than women who were better educat ed, (d) poorer women (income, 150% of the federal poverty level) were less likely to have been screened regularly for cervical cancer than

PAGE 49

38 had women with income >150% of the federal poverty level, (e) Women who were not in the work force (e.g. housewives) were less lik ely to have received Papanicolaou smears than those who were employed and unemploye d, and (f) Black women were more likely to have had Papanicolaou smears than were White and Hispanic women (Hayward et al. 1988). Calle, Flanders, Thun and Martin, (1993) used data from the 1987 National Health Interview Survey Cancer Control S upplement to identify demographic variables associated with underuse of mammogra phy or Papanicolaou smear screening. The characteristics were then examined together to produce profiles of women who underuse the cancer screening services available to th em and who would be most likely to benefit from intervention programs. Calle, et al. (1993) analyzed the resp onses of 12,252 women to questions about Papanicolaou smear screening. The women were 18 years and older and had not reported any history of cancer, except non-melanoma sk in cancer. The analysis of mammography use was restricted to 6,353 women aged 40 a nd older without a history of cancer. Two measures of underuse of both Papanicola ou smear screening and mammography were employed, these were, never having been screen ed and not having been screened in the past year. The researchers examined eight de mographic variables as potential covariates of underuse of screening. These variable s were: age (18 to 39, 40 to 64, 65+ for Papanicolaou smear; 40 to 49, 50 to 64, 65+ for mammography); race/ethnicity (White, Black, Hispanic, other); formal education (few er than 12 years, 12 years, more than 12 years); marital status (married, widowed, divor ced, never married); type of urban area (central city, other metropolitan statistical area, non-metropolitan statistical area); region

PAGE 50

39 (Northeast, Midwest, South, We st); employment (in the labo r force, not in the labor force) and income (below poverty level, pove rty level to 200% of poverty level, 200% to 300% of poverty level, >300% of poverty level). The income variable was constructed based on the reported household income a nd family size, and using the 1987 poverty income guidelines (DHHS as cited in Calle, Flanders, Thun & Martin, 1993) to identify those people living below the poverty level, which was defined as a household income of about $11,000 for a family of four. The researchers found that when all the demographic variables were included in the multivariate model, five variables remained significantly associated with underuse of Papanicolaou smear screening. These variab les were: (a) other race, and (b) nevermarried marital status were the strongest in dependent covariates of never having had a Papanicolaou smear; (c) Hispanic ethnicity, (d ) age 65 years or older, and (e) education of fewer than 12 years were also strong independent covariates of underuse of Papanicolaou smear screening (Calle, Flanders, Thun & Martin, 1993). Katz and Hofer (1994) compared the asso ciation of income and education with breast and cervical cancer scr eening in Ontario, Canada, an d the United States. The study sample was representative of all civilian, noninstitutionalized women aged 18 years and older living in Ontario and the United Stat es. The researchers used the 1990 Ontario Health Survey (OHS) and the 1990 U.S. Na tional Health Interview Survey (NHIS) Health Prevention Supplement Sample Pe rson File, population-based surveys that collected detailed information on health care use, health status, health behavior, and demographics from a sample of the ci vilian, noninstitutionalized population. The dependent variables were: (a) a Papanicola ou test within 2 years of having being

PAGE 51

40 surveyed; (b) a clinical breas t examination within 1 year of having being surveyed, and (c) a screening mammography w ithin 1 year of having being surveyed. Katz and Hofer (1994) selected a two-year interval reco mmendation for the frequency of screening Papanicolaou tests because it was the most comparable interval across the surveys. The independent variables were family in come and education. Family income was divided into six categories, the lowest being less than $15,200 and the highest being greater than $45,600. Education was categor ized as some high school, high school graduate, some college, and college gra duate. The researchers found that in both countries, women with higher incomes and highe r educational levels were more likely to have received Papanicolaou scre ening tests (Katz & Hofer, 1994). Hubbell, Chavez, Mishra and Valdez (1996) studied whether beliefs about cervical cancer influenced the use of Papanicolaou smears among Latinas and Anglo women in Orange County, California. The re searchers designed a survey instrument using questions from the National Health Inte rview Survey, the Behavioral Risk Factor Surveillance Survey and an earlier ethnographic survey of th e Orange County Latinas and Anglo women. Trained bilingual female interviewers c onducted the survey from September 1992 to March 1993, using the comp uter-assisted telephone interview (CATI) system. Approximately 94% of Latino and 99% of Anglo households in Orange County have telephones (U.S. Bureau of Census as cited in Hubbell, Chavez, Mishra & Valdez, 1996). The survey used a cross-sectional samp le of random-digit telephone numbers that included both listed and unlisted numbers, thereby minimizing potential bias by excluding households with unlisted numbers.

PAGE 52

41 Hubbell, et al. (1996) used the Chi-square test to analyze the categorical data and logistic regression analysis to evaluate th e covariates of Papani colaou smear use. The predictor variables for both Latinas a nd Anglo women included: age (<40 years; 40 years), marital status (married; not married), household income (<$25,000; $25,000), insurance status (no health status; any type of hea lth status), education ( to high school; >high school), and employment status (e mployed; not employed and not in the workforce). Acculturation, and country of birt h (born outside the United States; born in the United States) were added to the list of predictors for the Latina sample. The researchers reported that the variab les associated with Papanicolaou smear use among Latinas within three years of having being surveyed were (a) health insurance status, (b) marital status, and (c) acculturation level. Women with health insurance were more likely than those without insurance to have had cervical can cer screening. Latinas who were married were more likely to have had Papanicolaou smear screening and those who were more highly acculturated to living in the United States were more likely to have had a Papanicolaou smear scre ening test (Hubbell, et al., 1996). Fontaine, Faith, Allison a nd Cheskin (1998) examined the relationship between body mass index [BMI] (calculate d as weight in kilograms divided by the square of height in meters), and the use of preven tive health care services in a nationally representative sample of women. The data sources used were the 1992 National Health Interview Survey (NHIS), Cancer Control and Health Insurance supplements, conducted by the National Center for Health Statistic s. The study participan ts consisted of 6,981 women aged 18 years old and older residing in the United States who reported sociodemographic information and the use of preventive health care services.

PAGE 53

42 The independent variables in the study were: age, race (nonwhite or white), family income, education (years complete d), smoking status (nonsmoker-former smoker or current smoker), and health insurance st atus (not covered or unknown or covered by private insurance, Medicare, or both). Th e dependent variables were the number of physician visits in the 12 months before comp leting the survey, and the interval since the most recent mammography, clinical breast ex amination, gynecological examination, or Papanicolaou smear test. The intervals were defi ned as : (a) within the past year, (b) 1 to 3 years ago, (c) more than 3 years ago, (d) unknown specific interval ( 3 years versus >3 years), (e) not ascertained or do not know, or (f) unknown or refused. Multiple linear regression analysis was used to examine the relationship between the BMI and the number of physician visits in the 12 months before completing the survey. Logistic regression analyses were us ed to examine whether BMI was related to delaying each of the four preventive health care procedures. The researchers found that after controlling for age, r ace, income, education smoking status and health insurance status, the BMI increased in direct associ ation with increased frequency of physician visits. However, compared with women of average relative body weight (BMI of 25), obese and severely obese women were significantly more likel y to delay clinical breast examinations, gynecological examinations a nd Papanicolaou smear testing, suggesting that body weight may play a role in delayi ng these forms of preventive health care. Rakowski, Clark and Ehrich (1999) investig ated the association of smoking status with breast and cervical cancer screening across the 1990-1994 Nationa l Health Interview Surveys (NHIS). The data were from th e Health Promotion and Cancer Control Supplements to the 1990-1994 NHIS. The population of women studied were 42 to 75

PAGE 54

43 years of age. Associations were examined between smoking status (never, former, <1 pack/day, 1 pack/day) and three screening indi cators: (a) ever had a screening mammography, (b) had mammogram in the past 2 years, and (c) had the Papanicolaou test in the past 3 years. Data analyses we re conducted by bivariate and multiple logistic regression. The researchers found that women who smoked 1 pack of cigarettes per day were significantly less likely to have ha d a Papanicolaou smear, or a screening mammography compared to women who had never smoked. Simoes, Newschaffer, Hagdrup, Ali-Ab arghoui, Tao, Mack & Brownson (1999) combined data from two probability sa mples, 967 women from the 1994 Missouri Behavioral Risk Factor Su rveillance System (BRFSS) and 816 women from the 1994 Missouri Enhanced Survey (ES) to investigate the predicto rs of: (a) routine cervical cancer screening, and (b) compliance with a recommended cervical cancer screening schedule. The researchers analyzed data on two groups of women: women aged 50 and older, and those younger than 50 years of age. The independent variables were: (a) race/ethnicity (White or African American ), (b) educational atta inment (less than high school, or high school graduate and higher), (c) health insura nce (covered or not covered), (d) financial barriers to medical care in the previous year (had financial barriers, did not have financial barriers), (e) clinical breast examination (never had screening, not had screening within past five years, had screen ing within past five years), (f) screening mammography (never had screenin g, not had screening within past five years, had screening within past five y ears), (g) weight (ove rweight: BMI >27.3 kg/m2, non-overweight: BMI 27.3 kg/m2, (h) smoking status (ever smoked or never smoked), and (i) physical activity during previous month (active or inactive).

PAGE 55

44 Simoes, et al. (1999) used logistic regression modeling to generate prevalence odds ratios to identify predictors of non-compliance to cervical cancer screening guidelines. The researchers s howed that women who were younger than 50 years of age, and had had a mammogram or a clinical br east examination during the previous five years were more likely to comply with th e compliance with the 1999 cervical cancer screening guidelines of the ACS schedule. The odds of non-compliance with compliance with the 1999 cervical cancer sc reening guidelines of the ACS: (a) increased with age, and (b) was higher among White women, smoke rs, women without health coverage, women without a high school education, and among women who had experienced financial barriers to seeking medical care. Jennings-Dozier and Lawrence (2000) i nvestigated whether specific sociodemographic variables, such as age, income, marital status and number of persons living at home, were associated with annual Pap testing adherence in a sample of minority women form the Philadelphia metropolitan ar ea. A convenience sample of 204 Black and Hispanic American women was recruited fr om nonprofit agencies in the Philadelphia metropolitan area. The inclusion criteria fo r the study included women who: (a) could read and comprehend English and/or Spanis h, (b) did not have a medical history of cervical cancer, (c) had not had a hysterecto my, and (d) had lived on the mainland of the United States for at least one year. Women were considered to be adherent to Pap testing if they reported having had a Papanicola ou test within the 14 months preceding enrollment into the study. Women who had never had a Papanicolaou smear, or who had not obtained a cervical cancer screening test within the last 14 months preceding their enrollment into the study were classified as being non-adherent.

PAGE 56

45 The researchers used the Demographic A ssessment Survey (DAS) to collect data on the following variables :age ( 40 years and 40 years), ethnicity (H ispanic or Black) educational level (never attended school to completed graduate/professional school), family income ($5,000 to $50,000), and their hi story of cervical can cer screening (ever had a Papanicolaou test or last time obtained a Papanicolaou screeni ng test) and level of acculturation (for Hispanics). To measure th e level of acculturation in the Hispanic population of women, the study pa rticipants were asked whether they were born on the mainland United States and whether they s poke English at home (Jennings-Dozier & Lawrence 2000). The researchers reported that the Black American women who were adherent to annual Papanicolaou testing were younger, earned a higher in come, had insurance, were unmarried and were more likely to be high school graduates than Black American women who were not adherent to annual Papanicola ou testing. A similar result was reported for the Hispanic American women in the st udy. Those who were adherent to annual Papanicolaou testing were younge r, earned a higher income, had insurance, were married and were at least high school graduates when compared to Hispanic women who were not adherent to compliance with the 1999 cervi cal cancer screening guidelines of the ACS (Jennings-Dozier & Lawrence 2000). Amonkar and Madhavan (2000) determined compliance rates for breast and cervical cancer screening behavior reco mmendations for women residing in the Appalachian states and identified covariates of these compliance rates by using BRFSS from data from 1995 to 1997. Compliance w ith other preventiv e services, having insurance coverage, residing in urban areas better self-reported health, and higher

PAGE 57

46 education were associated with increase d odds of compliance with annual screening recommendations. Obesity and smoking were associated with decreased odds of compliance. Coughlin, Thompson, Seeff, Richards and Stallings (2002) compared the cancer screening patterns of adult ( 18 years of age) Black American and White American women and men living in nonmetropolitan coun ties of the southern Black Belt region of the United States, to those of individuals li ving in nonBlack Belt southern counties and other regions of the United States. The Black Belt region is an area which includes practically contiguous counties in the states of Virginia, North and South Carolina, Georgia, Florida, Alabama, Mississippi, Louisi ana, Texas, Arkansas and Tennessee, with high concentrations of Black Americans. Th ese are predominantly rural areas in which the population is economically dependent on agriculture (Lyson & Falk, 2000). The authors analyzed data from the state-based Be havioral Risk Factor Surveillance System (BRFSS) to find out whether there were dispar ities in the cancer scr eening rates of Black American and White American women and men in nonmetropolitan counties of the southern Black Belt region of the United Stat es, to those of individuals living in nonBlack Belt southern counties and other regions of the United States. The screening tests were use of the Papanicolaou test, mammograp hy, test for fecal occult blood test (FOBT), and flexible sigmoidoscopy or colonoscopy. The researchers found that Black American and White American men in the Black Belt counties were significantly le ss likely to have: had a colonoscopy, sigmoidoscopy or fecal occult blood screening te st than those living in non-Black Belt southern counties or elsewhere in the United States. Black American and White

PAGE 58

47 American women in the Black Belt counties were significantly less like ly to have: (a) had a recent (within the past 2 years) mammogr am than were Black American and White American women living elsewhere in the United States, and (b) had a colonoscopy, sigmoidoscopy or fecal occult blood screening te st than those living in non-Black Belt southern counties or elsewhere in the United St ates. This pattern of regional differences in screening behavior was not found among th e women who had had a recent (within the past 3 years) Papanicolaou test (Coughlin, T hompson, Seeff, Richards & Stallings, 2002). Hiatt, Klabunde, Breen, Swan and Balla rd-Barbash (2002) reviewed 65 papers, published between 1980 and 2001, that had used NHIS data to determine whether the investigators had assessed one or more covariates of screenin g use to identif y factors that could help to explain screeni ng practices or to reveal diffe rences in screening behavior among different subpopulations. Forty-eight st udies examined covariates of screening use, but only 10 of these studies used a theo ry or theoretical model. The researchers believe that it is important to use a theory or theoretical model because they guide the design of the research questi on, and the ensuing data analys is and interpretation of the results. Women with: (a) highe r levels of education and in come, (b) a usual source of health care and (c) insurance coverage were more likely to be screened for cervical cancer. Younger women were more likely to be screened for cervical cancer than older women. These variables were strongly and consistently associated with screening behavior across the studies reviewed. These findings are consistent with the results in the studies reviewed by the principal investigator in which: (a) educationa l level, (b) income level, (c) regular source of care, (d) health insurance coverage, (e) age, (f) employme nt status, (g) ethnic origin,

PAGE 59

48 (h) marital status, (i) BMI (j) smoking status and (k) past screeni ng behavior, (l) region of residence, and (m) primary care provider ad vice about screening, were all found to be significantly associated with cancer screening behavior. Physician Recommendation and Adherenc e to Cancer Screening Guidelines One of the objectives of this study is to investigate the association between primary care provider advice about cervical cancer screening and compliance with the 1999 cervical cancer screening guidelines of th e ACS in an ethnically diverse population of women residing in the United States. Therefor e a search of the lite rature database at the National Library of Me dicine (MEDLINE) and manual searches from 1996 to 2005 were performed relevant to the topic of physician recommendation and adherence to cancer screening guidelines. The results from these selected studies are summarized in the following section of the literature review. Roetzheim, Fox, Leake and Houn (1996) conducted a study to determine whether selected risk factors influenced the breast cancer screening rate s of women Medicareinsured women, aged 65 years and older. The researchers used data from the NCI Medicare Supplement that had been collect ed for the NCI in 1991 and 1993 to evaluate the impact of the 1991 benefit on mammogra phy utilization. Self-re ported rates of screening mammography and clinical breast examinations were compared for women with benign breast disease, women with a fa mily history of breast cancer and women without these risk factors. The sample c onsisted of 5,376 non-Hispanic White, Medicareinsured women surveyed at five National Canc er Institute Breast Cancer Consortium sites in 1991 and 1993. These sites were in Long Island, New York; Los Angeles, California; eastern North Carolina; eastern Massachusetts and Philadelphia, Pennsylvania. It was

PAGE 60

49 reported that the women with a family histor y of breast cancer and benign breast disease were more likely to have had screening mammography than those women lacking these risk factors. Women with a positive family hi story, or personal history were more likely than those without these risk factors, to have reported having had a clinical breast examination. The researchers concluded that th ese results were due to the more frequent provision of clinical breast examinat ions by the physicians and physician recommendations for screening mammography examinations. Champion and Menon (1997) studied the vari ables associated with breast cancer screening behaviors of mammography utiliz ation and breast self-examination in a convenience sample of low income Afri can American women living in a large Midwestern, metropolitan area. Data were collected as pa rt of an ongoing intervention trial to increase breast cancer screening in low income African American women. Four hundred and thirty women were enrolled into the longitudinal study from three multisite service centers in the metropolitan area. Eligib ility for the study incl uded being: (a) 45 to 64 years of age, (b) African American, and (c) having a combined income of 150% of poverty ($21,666 annually for a family of fou r). Variables studied included health insurance, source of health care, cost of having a mammogr aphy, perceived risk of having a mammography, perceived benefits, perceive d barriers and physician recommendation. Variables that were significantly associated with breast self examination included: (a) perceived risk, (b) perceived benefits, (c) perceived barriers and (d) having a regular physician. Variables that signi ficantly predicted mammography utilization included: (a) perceived barriers, (b) having a regular phys ician, (c) and reco mmendation by a health care professional to have a mammography.

PAGE 61

50 Ruchlin (1997) identified and assessed th e differences in breast and cervical cancer screening patterns among women who were 5564, 65-74, 75-84, and over 84 years of age. The data was obtained from a nationally representative sample of 28,584,574 women who were interviewed for th e 1990 Health Promotion and Disease Prevention Supplement to the National Health Interview Survey. The dependent variables were screening behavior : (a) having knowle dge of breast self-examination, (b) ever having had a mammography, and (c) having had a Papanicolaou smear within the last three years. The independent variables were age, educational level, area of residence, health status and health belief measures. The researcher found that women 64 years and older were less likely to have had any of the screening te sts than women 55-64 years of age. White women were more likely to know how to examine their own breasts and were more likely to have ever had a screeni ng mammography than we re non-White women. Having attended college and living in a central city of a metropolitan area were also significantly associated with all three screening measures Finally, Ruchlin (1997) found that over a third of the women in the sa mple had not had a mammography because the procedure had not been recommended to them by their physician. Stoddard et al. (1998) studi ed the characteristics of 11, 292 women 50 to 80 of age, who had not been adherent to the Na tional Cancer Institute (NCI) recommendation that they obtain a screeni ng mammography every one or tw o years. Data were from baseline surveys of women collected at th e five study sites of the NCI Breast Cancer Screening Consortium. The Consortium includes the Fred Hutchinson Cancer Research Center, the State University of New York at Stony Brook, the RAND Corporation, the University of Massachusetts, and the Duke Co mprehensive Cancer Center. The five sites

PAGE 62

51 share a primary goal of identifying effective me ans of increasing uti lization of screening mammography by women aged 50 years and abov e who have not been adherent to the NCI screening recommendations. Non adherent women were define d as those who had not had a screening mammography during the pa st 24 months, or those who had not been screened for 24 months prior to the most recent mammography. The women were enrolled from different subpopulations of the United States. The subpopulations were defined by different cr iteria including ge ographic region, HMO membership, church membership, or non-urba n residence. The study results indicated that the primary predicto rs of regular mammography adoption included physician recommendation for a mammography, and a recent clinical breast examination (Stoddard et al., 1998). O’Malley, Earp, Hawley, Schell, Matthews and Mitchell (2001) investigated the association between physician recommenda tion for mammography and race/ethnicity, and socio-economic status in a population of women who lived in 10 counties across rural, eastern North Carolina. Data for th is study was collected from the 1993 to 1994 base survey of 2000 women 50 years and olde r, which included 2 cohorts of Black women and 2 similar cohorts of White wome n. The primary outcome was self-report of physician recommendation for mammography in the past 2 years. The researchers concluded that physician reco mmendation was a strongly as sociated with mammography use. Taylor et al. (2002) conduc ted a study to examine Pa panicolaou testing barriers and facilitators among Chinese American wo men in Seattle, Washington. Four hundred and thirty-two women aged 20 to 79 years of age completed a questionnaire. The main

PAGE 63

52 outcome measures of the community-based st udy were that the women had had: (a) at least one previous Papanicolaou smear test a nd, (b) a Papanicolaou smear test within two years of the interview for the survey. The researchers reported that women with a history of at least one Papanicolaou smear test: (a) were married, (b) believed that a Papanicolaou smear te st is necessary for sexually inactive women, (c) lacked concer n about cancer being discovered, (d) were embarrassed by having the test, (e) had rece ived a physician or family recommendation for cervical cancer screening, (f ) had obtained family planning services in North America and (g) had a regular primary care provider. Women who had a Papanicolaou smear test within two years of the interview for the su rvey: (a) believed that a Papanicolaou smear test is necessary for sexually inactive women, (c) were not em barrassed by having the test, (d) had received a physician recomme ndation for cervical cance r screening, (e) had received obstetric services in North Ameri ca and (f) had a regular primary care provider (Taylor et al., 2002). It is well documented in breast cance r screening literature that physician recommendation about obtaining a mammogram is a strong covariate of adherence to breast cancer screening guidelines (ACS, 2004; Lippert, Eaker, Vi erkant & Remington, 1999; O’Malley, Earp, Hawley, Schell, Math ews & Mitchell, 2001; Roetzheim, Fox, Leake & Houn, 1996). Mandelblatt and Yabroff (2000) reviewed breast and cervical cancer screening guidelines for women 65 years and older. The researchers reported that older women, especially older minority wo men, remain underrepresented in screening programs and are the most likely to die fr om breast or cervical cancer. Physician recommendation has been found to be one of th e strongest predictors of screening across

PAGE 64

53 all age, socio-economic and ethnic origins, a nd one of the most important reasons women report for not being screened is that their physicians did not recommended the procedure to them. Evidence from the literature suggests that the number of deaths from cervical cancer in the United States can be redu ced by preventive screening (DHHS. Race and health: Cancer management, 1999; CD C, 1998-1999; Franco, Duarte-Franco, & Ferenczy, 2001; Holmquist, 2000; Kl aes, et al. 2001; Koop, 1997; MMWR, 1997; Runowicz & Fields, 1999; Sasieni & Adam s, 1999; Schiffman, Brinton, Devessa & Fraumeni, 1996). Although evidence from th e literature indicates that physician recommendation is strongly associated with adherence to breast cancer screening guidelines, little attention has been pa id to physician recomm endation as a major covariate of cervical cancer screening (Hiatt, Klabunde, Br een, Swan & Ballard-Barbash, 2002; Champion & Menon, 1997). The principal investigator, therefore, proposes to contribute further to the body of knowledge by investigating the association between adherence to cervical cancer screening guidelines and primar y care provider advice about obtaining a Papanicolaou smear test. Ethnic Group and Cancer Screening Behavior Last (1995) defines an ethnic group as a “social group characterized by a distinctive social and cultural tradition, maintained within the group from generation to generation, a common history and origin, and a sense of identification with the group. Members of the group have distin ct features in their way of life, shared experiences and often a common genetic heritage These features may be refl ected in their health and disease experience.” Ethnic origin is used in this study in place of the terms race and/or

PAGE 65

54 ethnicity used by other researchers. This te rm [ethnic origin] as used by the principal investigator also encompasses the distinctive socio-cultural traditions maintained within each group. Health professionals have learned that in terventions to promote health must be culturally relevant to the population in que stion (DHHS, 1998). For example, minority populations may have different perspectives on what constitutes cancer or the relative importance of screening for cervical cancer. Increasing demographic diversity makes it imperative for researchers to develop a mo re diverse approach to the design of intervention program s (Fielding, 1999). It has been hypothesized that the method of grouping data into broad ethnic group categories could mask unde rlying differences among the different cultural groups (Potosky, Breen, Graubard & Parsons, 1998). Fo r example, a study of cancer screening in a multi-ethnic community of Blacks and Hispanics found that rates of mammography and Papanicolaou smear tests differed among Colombian, Dominican, Ecuadorian, Puerto Rican, Caribbean, Haitian and United States -born Black women (O’Malley, Mandelblatt, Gold, Cagney & Kerner, 1997). Hiatt and Pasick (1996) hypothesized th at variations in screening rates among members of the same ethnically-defined group could be partly explained by differences in language and level of acculturation within the group. The researchers reported that the hypothesis was supported by the observation that significantly lower rate s of screening mammography are a nd clinical breast examination are reported, in the United States, among non-English speakers compared to English speakers. When language was evaluated w ithin ethnically defi ned groups, rates of screening for English-speaking Hispanic and Chinese women were closer to those of

PAGE 66

55 White American women. Within Hispanic popul ations, lower levels of acculturation and speaking Spanish were associated with lo wer rates of cancer screening, and older Hispanic women where less likely to speak English, more likely to speak Spanish and were less acculturated than younger Hispanic women (Hiatt & Pasick, 1996; O’Malley, Kerner, Johnson & Mandelblatt, 1999; Stein, Fox & Murata, 1991). Kagawa-Singer (1997) reported that among Chinese American women, especially among unmarried women, restrictions on gyneco logical examinations by male physicians and breast self-examinations (because touc hing one’s own body intimately may be taboo) could be individual barriers that prevent women from participating in cancer screening programs. Barriers such as insufficient of in formation in the various Chinese languages, few female physicians and the absence of edu cational campaigns tailored to the needs of the Chinese American women also contribute to the underscreening of this subpopulation (Mo, 1992). Perez-Stable, Sabogal, Otero-Sabogal, Hiatt and McPhee (1992) collected information regarding the knowledge about and attitudes toward cancer in a sample of adult health plan members, self-identified as either Latino or Anglo. The researchers found that compared to Anglos, Latinos were more likely to believe that: (a) having cancer is like receiving a death sentence, (b) cancer is God’s punishment, (c) it is uncomfortable to touch someone with cancer, an d (d) there is very little one can do to prevent cancer. The researchers concluded that the data collected woul d be invaluable in developing tailored, culturallyappropriate cancer control interventions for Latinos. Lee (2000) enrolled 102 Korean American from Queens, New York into a study to investigate their general knowledge about cervi cal cancer screening. The Health Belief

PAGE 67

56 Model was used as theoretical model to guide the study. The concept of perceived barriers to preventive health actions is a ke y component of the Health Belief Model. Lee (2000) enrolled the women into focus groups ranging in size from 9 to 17 women, and asked them about: (a) their knowledge about th e cause, early detection and prevention of cervical cancer, and (b) about the barriers that prevented th e women from screening for cervical cancer using the Papa nicolaou test. The women in the focus groups expressed a number of perceived barriers that they fe lt prevented them from having the Papanicolaou test. These barriers were groupe d into structural and psychos ociocultural barriers by the researcher. The structural barriers were factors th at affected the study population’s access to health services. The most frequently mentioned structural barriers we re cost and lack of insurance, lack of time and language difficu lties. Women stated that because of these barriers, both they and some of their frie nds would return to Korea for the tests and treatment of cancer or other medical c onditions. The psychosoc io-cultural barriers included perceptions, social customs, and cultu rally induced beliefs and attitudes related to the women’s backgrounds and socializati on. Fear and embarrassment were found to be important psychosociocultural barriers for the Korean Amer ican women, many of whom attributed these barriers to their upbringing (Lee, 2000). Kelaher, Gillespie, Allotey, Manderson, Potts, Sheldrake and Young (1999) used the Transtheoretical Model of Behavioral Change (TTM ) to study participation in cervical cancer scr eening programs by different la nguage and cultural groups in Queensland, Australia. The model used by th e researchers consisted of six stages: (a) precontemplation (no past history of screening, no intention to be screened),

PAGE 68

57 (b) contemplation (no past history of screening, intention to be screened), (c) action (intention to continue screeni ng after initial screening), (d) maintenance (intention to continue regular screening), (e) relapse (no intention to be screened again after initial screening), and (f) relapse risk (no intention to be screened again after regular screening). Focus groups and structured inte rviews were used to classify the sample in terms of the TTM. The study sample consisted of Australian South Sea Islanders, Chinese, German, Greek and Muslim women. Kelaher, et al. (199 9), stated that cervical cancer screening promotion for women of diverse cultures and et hnicities in Australia tended to focus on women in the precontemplation and contemplat ion stages. However, since most of the women in the study sample were in the acti on and maintenance stag es, the researchers concluded that representing cervical cancer screening promotion in terms of the Transtheoretical Model could c onsiderably improve the effectiveness of interventions for women of diverse cultures and ethnicities because different stages of change need different points of intervention (K elaher, et al. 1999; Rimer, 1996). Evidence from the literature suggests that the number of deaths from cervical cancer in the United States can be redu ced by preventive screening (CDC, 1998-1999; DHHS. Race and health: Cancer management 1999; Franco, Duarte-Franco, & Ferenczy, 2001; Holmquist, 2000; Klaes, et al. 2001; Koop, 1997; MMWR, 1997; Runowicz & Fields, 1999; Sasieni & Adams, 1999; Schi ffman, Brinton, Devessa & Fraumeni, 1996), little attention has been pa id to physician recommendati on as a major predictor of cervical cancer scr eening (Hiatt, Klabunde, Breen, Swan & Ballard-Barbash, 2002; Champion & Menon, 1997). The socio-cultural trad itions or customs are reflected in the preventive health behavior of the ethnical ly diverse group of women highlighted in the

PAGE 69

58 review of the literature. Th ese traditions may put the women at a high risk of being underscreened (Potosky, Breen, Graubard & Parsons, 1998). One of the most powerful measures for reducing the mortality rates of cervical cancer, especially in minority populations, is by increasing cervical cancer screening (CDC, 1998-1999; DHHS. Race and health: Cancer management, 1999; Franco, DuarteFranco, & Ferenczy, 2001; Holmquist, 2000; Klaes, et al. 2001; Koop, 1997; MMWR, 1997; Runowicz & Fields, 1999; Sasieni & Ad ams, 1999; Schiffman, Brinton, Devessa & Fraumeni, 1996). Although the literature id entifies physician recommendation as the most powerful predictor for breast cancer sc reening (Mandelblatt & Yabroff, 2000), little is known about the effect of physician r ecommendation on cervical cancer screening, especially in minority populat ions. In order to develop effective programs to increase cervical cancer screen ing, it is important to examine whether population-specific factors such as ethnic origin should be a focus for future intervention programs. The effect of ethnic origin as a moderating variable in the association between primary care provider advice about cervical cancer screening and compliance with the 1999 cervical cancer screening guidelines of the ACS will also be investigated. It is critical, given the public health orientation of this study, to inves tigate whether membership of an ethnic group with its distinctive social and cultural trad itions could have a moderating effect on the association between primary care provider ad vice about cervical cancer screening and compliance with the 1999 cervical cancer screen ing guidelines of the ACS The insights gained from this study will be added to the body of knowledge and used to design more effective, tailored, interventions to enable the women to adhere to the cancer screening

PAGE 70

59 guidelines of the American Cancer Societ y, thereby reducing the mortality rates of cervical cancer, especially in the minority populations. The Preventive Health Model The Preventive Health Model was deve loped specifically to understand and determine the covariates of preventive h ealth behavior (Myers, Ross, Jepson, Wolf, Balshem, Millner, & Leventhal, 1994). It is a theoretical model which has been derived from constructs of the Health Belief Mode l, the Theory of Reasoned Action and Social Cognitive Theory that are thought to be impor tant in predicting preventive health behavior, such as cancer screening (Carver & Scheier as cited in Myers, et al., 1994). This model has been used in studies to predic t: (a) colorectal cancer screening behavior in men and women (Myers et al., 1994, Watts Vernon, Myers & Tilley, 2003), and (b) prostate cancer screen ing behavior in African Amer ican men (Gwede, 2001; Myers, Wolf, Mckee, McGory, Burgh, Nelson, & Nelson 1996). The Preventive Health Model posits that a broad set of f actors influence an individual’s decision to be sc reened for cervical cancer (Mye rs et al., 1994). The first set of factors described as the background factors which may include age, gender, ethnic origin, education, and past preventive scre ening behavior. These background factors may be described as the socio-demographic c ontext within which an individual makes decisions about his or her health care. The background factors constr uct is derived from the Health Belief Model (Str echer & Rosenstock, 1997). The second set of factors are described as representation factors. The representation factors construc t is also derived from the H ealth Belief Model and includes an individual’s perceptions about: (a) his or her own susceptibili ty to a particular disease,

PAGE 71

60 (b) the severity of th e consequences of leaving the dis ease untreated, (c) the benefits of adhering to the recommended guidelines for enga ging in behavior to prevent the disease, and (d) the barriers of adhe ring to the recommended prev entive behavior, including the practical convenience of adhering to said be havior (Janz & Becker 1984; Strecher & Rosenstock, 1997). Social influence factors form the third set of factor s. These factors include an individual’s relationship with his or her health care profe ssional, and the social support for engaging in preventive health behavior provided by significant individuals in his or her life. This social influence factors construct is derived from Social Cognitive Theory which posits that behavior, personal factor s (including cognition), socio-cultural and socio-demographic factors all in teract in a reciprocal manner to determine an individual’s pattern of health behavior. (Bar anowski, Perry & Parcel, 1997). Program factors comprise the fourth set of fact ors in the model. An individual may be exposed to educational interventions such as primary care provider advice or recommendations, the purpose of which is to motivate and reinforce a particular preventive health behavior (Myers, et al., 1994 ). The program factors construct is derived from both Social Cognitive Theory and the Hea lth Belief Model and refers to the contacts made by the health services system for th e purpose of motivati ng and reinforcing the given preventive health behavior (Bara nowski, Perry & Parcel, 1997; Strecher & Rosenstock, 1997). According to the Theory of Reasoned Ac tion (Ajzen & Fishbein, 1980; Fishbein & Ajzen, 1975), the combination of backgr ound, representation, social influence and program factors could influence an individual’ s intention towards engaging in preventive

PAGE 72

61 health behavior. The Preventive Health Model posits that background, psychological representation, social influence and program factors influence an individual’s intention to engage in preventive health behavior, and al so influence an indi vidual to engage in preventive health behavior (Myers, et al., 1994). Myers, et al. (1994) used the Preventi ve Health Model to identify factors associated with prospective adherence to colorectal cancer sc reening. The sampling frame for the study consisted of 12,800 adult men and women (50 to 74 years of age) who were members of HMO PA or HMO NJ, two prepaid health care plans of U.S. Healthcare, an independent practice association health maintenance organization. As HMO members, subjects in the sampling frame were eligible for free colorectal cancer screening through a central screening office. The HMO members were mailed fecal occult blood screening kits on an annual basi s. The kit included thre e tests and a postagepaid return envelope. Individuals who did not return completed fecal occult blood tests to a central laboratory by mail within 15 days received a reminder letter to do the test. Individuals who adhered to th e testing received their resu lts by mail. Their primary care physicians also were notified of the test resu lts by mail. None of the individuals included in the study had received prio r fecal occult blood test mailin gs in the screening program. The outcome behavior, adherence to scr eening, was characteri zed by completion and return of the fecal occult blood tests within 90 days of the date on which the tests were mailed. The researchers found that variables from each of the four domains of the Preventive Health Model were found to be significantly associated with adherence to fecal occult blood testing.

PAGE 73

62 1. The background variables were: (a) gender (female), (b) age (being older) and (c) past fecal occult blood testing; 2. the psychological representation fact ors were: (a) worry about having an abnormal screening test result; (b) se lf-efficacy related to screening, (c) perceptions about the severity and curabi lity of colorectal cancer, and (d) the salience and coherence of fecal occult blood screening; 3. the social influence factors were: (a) rapport between physician and patient, and (b) powerful others locus of control; and 4. the program factors included exposure to health education interventions (Myers, et al. 1994). Myers, et al. (1996) presented the resu lts of a telephone survey conducted among African American men who had been se lected from the patient population of a community-based primary care phys ician practice in Philadelphi a. Researchers identified a total of 2,355 African American men from the population who were 40 to 70 years of age and had complete addresses and telephone numbers in the central computer database of the practice. A random sample of two hundr ed and fifty-two men was selected from this population. Two hundred and eighteen men were eligible to take part in the study, and 154 completed the survey. Factors measur ed in the telephone survey were elicited from the Preventive Health Model. The depe ndent variable was intention to undergo an annual prostate cancer screening examination. A number of the Preventive Health Mode l factors were found to be significantly associated with sc reening intention.

PAGE 74

63 1. The background factors were: (a) age, a nd (b) having had a screening examination in the past year before the interview; 2. the psychological representati on factors were: (a) belief in the salience of prostate cancer screening, (b) belief in the efficacy of screening, and (c) belief in the residual value of screening; 3. the social influence factors were: (a) perceived support from the health care professional, (b) receptivity to health care professional advice about screening, and (c) the influence of family members and friends on screening intention. The researchers did not collect any data on pr ogram factors, so this component of the Preventive Health Model was not included in the theoretical fr amework of the study (Myers, et al. 1996). Gwede (2001) studied factors that influence regular pr ostate cancer screening behavior among African American men. The rese archer sought to determine the barriers and motivating factors for screening among Af rican American men 40 years and older in Hillsborough County, Florida. The key constructs of the Preventive Health Model used in this study were background factors, psychol ogical representation factors and social influence factors. Gwede (2001) found the following statistica lly significant associations with one or other or both of the screening modalities (prostate-specific antigen blood test, or the digital rectal examination). Age was significan tly associated with both the ACS screening guidelines (use of both tests) and the relaxed ACS screening guidelines (use of at least one of the tests). Younger men (40-49 years of age) were less likely to be screened regularly for prostate cancer than older men (65 years of age and over). Men who

PAGE 75

64 received regular medical car e (saw a doctor at least once a year for any medical condition, or annual physical examin ation) were more likely to been screened for prostate cancer compared to those who had not rece ived regular medical care. Men who were more knowledgeable about prosta te cancer screening guidelines were more likely to have been screened regularly than those with less knowledge on the topic. Finally, men who received recommendations from their physicia ns to undergo prostate cancer screening examinations were more likely to have been screened regularly compared to those who had not received recommendations from their physicians. Gwede (2001) found that the results of his study were consistent with the results reported by Myers, et al. ( 1994) and Myers, et al. (1996) Gwede (2001) found that: (a) background factors (such as age and regul ar medical care); (b) psychological representation factors ( importance and benefits of screen ing); and (c) social influence factors (receptivity to the recommendation of the physician for prostate cancer screening) were significantly and positively associated with prostate cancer screening behavior. Watts, Vernon, Myers and Tilley (2003) reporte d that there is little information in the literature about wh ether the predictors of intention to be screened for colorectal cancer in cross-sectiona l studies also predict intention ove r time, or change in intention over time. The researchers studied the predicto rs of intention to screen for colorectal cancer in a population of white, male auto motive workers. They hypothesized that if intentions to screen for colorectal cancer ch ange over time, then it is imperative to know which predictors contribute to the behavior. Th e authors argued further, that since there is a low participation rate in colo rectal cancer screening, behavior al interventions need to be developed to encourage participation in th ese screening programs Information gained

PAGE 76

65 from such cross-sectional studies could then be used in intervention programs to: (a) reinforce an individual’s strong intention in following colo rectal screening advice, or (b) strengthen an individual’s weak intent ion in getting screened for colon cancer. Watts, Vernon, Myers and Tilley (2003) conducted a cross-sectional study among white, male, automotive workers taking part in The Next Step Trial, a worksite health promotion trial to encourage colorectal cance r screening and modifica tion in diet (Tilley et al., 1999a; Tilley et al ., 1999b). The study sample consisted of 2,556 men, enrolled from 28 different worksites, who had (a) responded to baseline (1993) and follow-up surveys (1994 and 1995) from the Next Step Tr ial, (b) did not have colorectal cancer at baseline, and (c) did not develop the dis ease during the study period. The dependent variables were: (a) intention to be screened for colorectal cancer, and (b) change in intention to be screened for colorectal cancer. The independe nt variables were represented by the constructs of the Preventive Health Model which included the background, representation, soci al influence and program fa ctors. The researchers found that a number of the Preventive Health Model factors were significantly associated with colorectal cancer screening intention. 1. The background factors were: (a) having a fa mily history of polyps or colorectal cancer, and (b) having had a screening ex amination during the previous 2 years before the interview; 2. the psychological representation factor s were: (a) belief in the salience of colorectal cancer screen ing, (b) lack of concern about screening-related discomfort, (c) perceived susceptibility to colorectal polyps and cancer, and (d) lack of fear and worry bout bei ng diagnosed with colorectal cancer;

PAGE 77

66 3. the social influence factors were: (a) receptivity to family member support for colorectal cancer screening, and (b) s upport for colorectal cancer screening among family members. Watts, Vernon, Myers and Tilley (2003) found that the results of their study were consistent with the results reported by My ers, et al. (1994) fo r colorectal cancer screening; Myers, et al. ( 1996) and Gwede (2001) for prosta te cancer screening; Orbell (1996) for cervical cancer; and Savage and Clark (1996) for breast cancer screening. The researchers did not include physician recommen dation as a predictor variable because the Next Step Trial was delivered through th e medical departments at each of the 28 worksites, not directly through the plant phys icians. However, they advised that since physician recommendation is such an important predictor variable in intention to undergo other cancer screening beha vior (Gwede, 2001; Myers et al., 1994; Myers et al., 1996; Orbell, 1996; Savage & Clark, 1996), it should be included as a predictor variable in future studies of intention to unde rgo colorectal cancer screening Even though many theories a nd theoretical models of h ealth behavior appear to have similar explanatory or pr edictive powers, no single theo ry, or theoretical model can be used to explain all of th e processes involved in comple x, health behavioral change, such as cervical cancer screening (Jennings 1997; Prochaska, R eading & Evers, 1997; Weinstein, 1993). The Health Belief Mode l (Rosenstock, Strecher & Becker, 1988), Social Cognitive Theory (Bandura 1986), the Theory of Reasoned Action (Ajzen & Fishbein, 1980; Fishbein & Ajzen, 1975), a nd the Transtheoretical Model (Prochaska, 1997) have all been used to try to explain why women do or do not obtain screening tests for cervical cancer (Rimer, 1996).

PAGE 78

67 The Preventive Health Model has been selected as the most appropriate theoretical model for investigating self-initia ted compliance with th e 1999 cervical cancer screening guidelines of the ACS in an et hnically diverse population of American women. The constructs of the Preventive Hea lth Model will be operationalized by the theoretically and conceptually appropriate va riables chosen from the pertinent literature, and matched with relevant va riables from the 1999 Behavioral Risk Factor Surveillance System data set. This theoretical model wi ll be used to guide the formation of the research questions, methodology, analyses, and s ubsequent interpretati on of the results of the study. Overview of the Behavioral Ri sk Factor Surveillance System During the early 1980s, scientific research showed that personal health behavior played a major role in chronic disease morbidity and mortality. During this time telephone surveys emerged as a reliable, valid and acceptable method of collecting information on health behavior. The Behavi oral Risk Factor Surveillance System (BRFSS) a collaborative proj ect of the Centers for Dis ease Control and Prevention (CDC), and the U.S. states and territories wa s initiated in 1984, with 15 states collecting surveillance data on risk behavior thr ough monthly telephone interviews. By 1990, 50 states, the District of Columbia, Puerto Rico, Guam and the Virgin Islands were participating in the BRFSS. The surveys were developed and administered to monitor the state-level prevalence of the major behavior al risk factors in the non-institutionalized adult population 18 years of age and over in the United States (CDC/NCCDPHP. Overview: BRFSS 1999; 2002). The basic philosophy of the BRFSS was then, and continues to be, to collect data on self-repor ted health behavior, rather than on self-

PAGE 79

68 reported attitudes and knowledge because this type of data can be used for planning, implementing, managing and evaluating hea lth promotion and disease prevention programs. Factors assessed by the BRFSS include tobacco use, general health status, health coverage, and the use of cancer sc reening services (CDC/NCCDPHP. About the BRFSS, 2002). The intention of this inves tigation is to study the asso ciation among the covariates of compliance with the 1999 cer vical cancer screening guidelines of the ACS (cervical cancer screening behavior) and to explore th e relationship between primary care provider advice about cervical cancer screening and compliance with the 1999 cervical cancer screening guidelines of the ACS in an ethni cally diverse population of American women. The BRFSS data set has been selected for th is purpose because it contains questions pertaining to (self-reported) health behavior rather than to (self-repo rted) knowledge, attitudes, and perceptions of, or intentions to perform health behavior. Moreover this data set contains survey questions specific to preventive screeni ng behavior, including cervical cancer screenin g behavior (CDC/NCCDPHP. About the BRFSS, 2002; Coughlin, Uhler, Hall & Briss, 2004). Reliability and Validity of the Se lf-Reported Data from the 1999 BRFSS Stein, Lederman and Shea (1994) studied th e reliability of the Behavioral Risk Factor Surveillance System questionnaire usi ng a random sample of White adults (n = 122) and a separate sample of Black and Hisp anic adults (n = 200) in Massachusetts. The questionnaire was administered twice, 21 to 44 days apart, by te lephone there was a 65% response rate for second administration of th e questionnaire. Individu al level reliability (Kappa for categorical variab les, correlation for continuous variables) for demographic

PAGE 80

69 characteristics was > 0.8 for White res pondents and >0.6 for Black and Hispanic respondents. Reliability coefficients for behavioral risk factors were >0.7. The researchers concluded that th e data supported the use of th e Behavioral Risk Factor Surveillance System questionnaire for surveillance and research. Stein, Lederman and Shea (1996) also assessed the reprodu cibility of responses to the women’s health module of the 1992 Behavi oral Risk Factor Surveillance System. The module included questions about breast and cervical cancer screen ing, clinical breast examinations, hysterectomy, and pregnancy status. A random sample of women in Massachusetts (n = 91, respons e rate for the repeat inte rview = 70.0%) and a separate random sample of minority women in the state (n = 179; response rate for the repeat interview = 69.4%) were interviewed twice, by telephone, 21 to 94 days apart. Based on Kappa statistics, concordance exceeded 85% for almost all variables examined, but tended to be lower for nonwhite respondent s. The researchers concluded that the women’s health module of the BRFSS ques tionnaire yielded hi ghly consistent group mean estimates of prevalence when administer ed repeatedly to the same individuals. Limitations and Strengths of the 1999 BRFSS Data Set Although 95% of households in the United States have telephones, coverage ranges from 87%-98% across states and varies for subgroups as well. For example, in the South, minorities and individua ls in lower socio-economic gr oups characteristically have lower telephone coverage. Post-stratification weights were used which partially corrected for any bias caused by non-telephone coverage. The weights also adjusted for differences in probability of select ion, nonresponse and non-telephon e coverage (CDC/NCCDPHP. Overview: BRFSS 1999, 2001). The absence of a telephone at home for a significant

PAGE 81

70 proportion of potential respondents could make the study population less representative than the general population, thereby limiting th e generalizability of the survey results (Marin, Vanoss & Perez-Stable, 1990). The BRFSS has a number of unique strength s that will be discussed in this section. Nationwide, survey items from the su rveillance system have remained relatively constant from year to year. A nother strength of the BRFSS is the relative ease with which data can be compared across the states. Alt hough the surveillance syst em is flexible and allows for additional questions from the states, the standard core questions enable health professionals not only to make comparisons betw een states, but also to track health trends over time and derive national-level conclusi ons from the observations. For example, by tracking trends over time, state-based data from the BRFSS have revealed a national epidemic of obesity (CDC/NCCDPHP. BRFSS: Tracking major health risks in America, 2001; CDC/NCCDPHP. BRFSS: Tracki ng public health trends, 2000). Public Health Significance of the Study Healthy People 2000: National Health Promotion and Disease Prevention Objectives was published in 1990 and identified hea lth improvement goals to be reached by the year 2000. These documents establishe d national health objectives and have served as a basis for the development of state and community health plans (Healthy People 2000, 1990). The Healthy People 2000 initiative had three goals, to: (a) increase the span of healthy life, (b) reduce health disparities, and (c) achieve access to preventive services. One of the priority areas in Healthy People 2000 was to increase the proportion of women aged 18 and older who had ever r eceived a Papanicolaou smear test, and those

PAGE 82

71 who had received the test with in the preceeding 1 to 3 years. Healthy People 2000 was the precursor to Healthy People 2010. (Healthy People 2000, 1990) Healthy People 2010, is a document containing a set of health objectives designed for the nation to reach during the first decade of the 21st century (Healthy People 2010, 2000). The two major goals of this health initiative are: (a) to help individuals of all ages to improve the quality of th eir lives, and (b) to eliminat e disparities in health among different sectors of the population. The vision for Healthy People 2010 is Healthy People in Healthy Communities (U.S. Department of Health and Human Services [USDHHS], 2000). Therefore, in the contex t of the current national em phasis on health disparities research, this proposed study is a particularly appropriate opportunity to contribute new research on screening-related health behavior to the body of literature, and to current national dialogue on the subject. Researchers at the National Inst itutes of Health are also initiating investigations into health disparities to unde rstand why some groups have disproportionately higher rate s of disease than others (NCMHD, What we do, n.d.). ). The Preventive Health Model has not been used to predict preventive cervical cancer screening behavior in women. The Pr eventive Health Model in conjunction with data from the 1999 Behavioral Risk Factor Su rveillance System will be used to identify the covariates of cervical can cer screening behavior in a diverse population of American women. The effect of ethnic origin as a mode rating variable in the association between primary care provider advice about cervical cancer screening and compliance with the 1999 cervical cancer screening guidelines of the ACS will also be investigated. The results of this study will enable pub lic health researchers to develop and design intervention programs that could be tailored to the specified population segments

PAGE 83

72 in the study (Glanz & Rimer, 1997; DHSS. Race and health: cancer management, 1999). The objective of the interventions is to increase cervical screening to more optimal levels among the diverse groups of women in th e nation, thereby, not only reducing the differences in mortality rate, but also reducing the overall mortality rate from cervical cancer in the nation.

PAGE 84

73 CHAPTER THREE: RESEARCH METHODS Purpose of the Study The purpose of the study was to use the Preventive Health Model in conjunction with data for the 1999 Behavioral Risk Factor Surveillance System to : (a) investigate the association of selected, signifi cant, covariates of cervical ca ncer screening behavior with compliance with the 1999 cervical cancer sc reening guidelines of the ACS in this ethnically diverse population of American wo men, (b) investigate the magnitude of the association of ethnic origin and complianc e with the 1999 cervical cancer screening guidelines of the ACS (c) investigate th e association between primary care provider advice about cervical cancer screening and compliance with the 1999 cervical cancer screening guidelines of the ACS in this population, and (d) investigate whether the association between primary care provider ad vice about cervical cancer screening and compliance with the 1999 cervical cancer scr eening guidelines of the ACS is moderated by ethnic origin. Research Questions The research questions posed were as follows: 1. Is there an association between each of th e selected covariates of cervical cancer screening behavior and compliance w ith the 1999 cervical cancer screening guidelines of the ACS as recommended in an ethnically diverse population of American women?

PAGE 85

74 2. Is there a difference in the magnitude of the association between ethnic origin and compliance with the 1999 cervical cancer screening guidelines of the ACS? 3. Is there an association between primary care provider advice about cervical cancer screening and compliance with the 1999 cervical ca ncer screening guidelines of the ACS? 4. Is the association between primary care provider advice about cervical cancer screening and compliance with the 1999 cervical cancer screen ing guidelines of the ACS moderated by ethnic origin? Null Hypotheses The study (null) hypotheses were as follows: 1. There is no association between each of the selected pred ictors of cervical cancer screening behavior, and compliance w ith the 1999 cervical cancer screening guidelines of the ACS in this ethnicall y diverse population of American women. 2. There is no difference in the magnitude of the association between ethnic origin and compliance with the 1999 cervical cance r screening guidelines of the ACS. 3. There is no association between primar y care provider advice about cervical cancer screening and compliance with the 1999 cervical ca ncer screening guidelines of the ACS. 4. Ethnic origin does not moderate the a ssociation between primary care provider advice about cervical can cer screening and compli ance with the 1999 cervical cancer screening guidelines of the ACS.

PAGE 86

75 Study Design The study design was a cross-sectional or prevalence study in which the status of the exposure (primary care provider advice about cervical cancer screening) and the attribute of interest (complia nce with the 1999 cervical cancer screening guidelines of the ACS) were simultaneously assessed among i ndividuals in the defined study sample. Selection of the Study Sample The BRFSS is a cross-secti onal surveillance system involving 52 reporting areas, with natural variation over the sample site s. The BRFSS data were collected from a random sample of adults (one per ho usehold) through a telephone survey (CDC/NCCDPHP, Overview: BRFSS 1999, 2001) A total of 159,989 individuals were interviewed for the 1999 Behavioral Risk F actor Surveillance System, 94,679 of these individuals were women from 18 years to 65 years and older (CDC/NCCDPHP. 1999 BRFSS Codebook, 2002). The BRFSS sample consisted of 74,888 (79%) White, 8237 (9%) Black, and 5,687 (6%) Hispanic, 2234 (2 %)Asian or Pacific Islander and 1450 (1.5%) American Indian or Alaska Native women. The Black and White Hispanic women were combined into a single Hispanic Am erican group because there were not enough data on the Black Hispanic women for analysis. In 1999, the ACS guidelines for cervical cancer screening suggested that “all women who are or who have been sexually ac tive, or who are 18 and older should have an annual Pap test and pelvic examination. Af ter three or more c onsecutive satisfactory examinations with normal findings, the Pap te st may be performed less frequently as per physician recommendation (ACS, 1999, p. 31).

PAGE 87

76 The ACS guidelines for cervical can cer screening update d in 2002 (Saslow, Runowicz, Solomon et al., 2002) recommended that “cervical cancer screening should begin approximately three years after a woman starts to have vagi nal intercourse, but no later than 21 years of age. Screening s hould be done every year with conventional Papanicolaou smear tests or every two years using the liquid-based te sts. At or after 30 years of age, women who have had three cons ecutive, normal resu lts may elect to be screened every two to three years. Physicians may suggest that a women screen more often if she has certain risk factors, such as HIV infec tion or a compromised immune system. Women who are 70 years of age and older, who have had three consecutive normal Papanicolaou tests in the last 10 years may chose to stop cervical cancer screening. Screening women who have had a total hysterectomy w ith removal of the cervix is not necessary unless the surgery wa s undertaken as a treatment for cervical cancer” (ACS, 2003, p. 48; ACS, 2004, p. 56). For the purpose of this study, the 1999 AC S cervical cancer screening guidelines were applied to the population of Black (nonHispanic), White (non-Hispanic), Hispanic, Asian or Pacific Islander and American Indian or Alaska Native women aged 18 to 65 years and over who reported that they had r eceived a preventive Papanicolaou smear test some time during their lifetime, selected from the 1999 BRFSS data set. Women who had never had a preventive Papanicolaou smear who had had a diagnostic Papanicolaou smear or were being, or had been treate d for an abnormal Papanicolaou smear were excluded from the study. The study populat ion consisted of 66,360 (82%) White, 7,236 (9%) Black, 4,774 (6%) Hispanic, 1,817 (2%) Asia n or Pacific Islander and 1,117 (1.4%)

PAGE 88

77 American Indian, or Alaska Native women, a total of 81,304 observations, with one record per individual. Selection of Variables A panel of experts was selected by the pr incipal investigator to provide advice about the selection of the vari ables, conceptual, theoretical and practical relevance of the variables in reference to th e Preventive Health Model a nd the 1999 BRFSS data set, and final formulation of the research question. Me mbers of the panel were required to have expertise in at least one of th e areas identified as being crucial to this research process. These areas of expertise included: (a) cancer epidemiology, (b) principles of cancer screening, (c) cancer prevention research, (d) cancer screening practice, (e) measurement and theory, (f) applied theory, (g) statis tical methods and research design, and (h) knowledge of secondary data analysis of public health data sets using SAS. The members of the panel were considered to be experts in their field by their peers, were members of their respective pr ofessional organizations, and ha d published articles within their areas of expertise. A list of the member s of the expert panel, and their areas of expertise may be found in appendix A. Members of the expert panel recommende d that: (a) additional references be added to the original literature review to strengthen the argument for selecting the variables from the 1999 BRFSS database to be studied, (b) because of the importance of primary care provider recommendation in can cer screening, this variable, or proxy measure for this variable, should be used as the major independent variable in the study, (c) a more detailed explanation of how th e Preventive Health Model is derived from components of the Health Belief Model, th e Theory of Reasoned Action and Social

PAGE 89

78 Cognitive Theory be written, and (d) variable s selected from the 1999 BRFSS data set to operationalize the constructs of the Preventive Health Model be of c onceptual, theoretical and practical value to the principal investigator in the context of this study. The covariates of cervical can cer screening were examined from the studies in the literature review of chapter 2. The covariates were consistent across the studies. None of the authors had used all of th ese predictors in a single study. All of these variables were examined to find out how well they explaine d cervical cancer screening behavior when they were all controlled for in the same study. Two hundred and eighty-four variable s in the 1999 BRFSS data set were examined to find those of conceptual, theore tical and practical importance that matched the cervical cancer screening c ovariates from studies in the literature review. After this process, the selected variables were assesse d by members of the pane l of experts to see whether they could be used to operationalize the constructs of the Preventive Health Model, because variables specify how a construc t is to be measured in a specific situation (Glanz, Lewis & Rimer, 1997). Operationalization of the Constructs of the Preventive Health Model Constructs of the Preventive Health M odel and descriptions of the variables selected from the 1999 BRFSS are shown in Ta ble 1. The following paragraphs contain brief discussions about the theoretical, c onceptual and practical relevance of these variables as empirical counterpa rts of the model constructs.

PAGE 90

79 Background Factors Construct This construct was operationalized by three types of variables; sociodemographic, past screening behavior and health profile variables. These diverse variables may influence a woman’s cer vical cancer screening behavior. Socio-demographic variables. The socio-demographic va riables were age, ethnic origin, marital status, educational level, em ployment status, income level, region of Table 1 Operationalization of the Construc ts of the Preventive Health Model Constructs of the PHM Variables from 1999 BRFSS References Background Factors Construct 1. Socio-demographic variables Age Ethnic origin* Marital status Education level Employment status Income level Region of residence Insurance coverage 2. Past screening behavior variables Mammography Breast exam 3. Health profile variables Tobacco use Obesity (BMI) Hysterectomy Representation Factor Construct Practical convenience Social Influence Factor Construct Rapport with primary care provider Program Factor Construct Screening advice Child-bearing age Ethnic origin Marital status Education level Employment status Income level State FIPS codes Health coverage Had mammogram Had breast exam Smoking status Body Mass Index Had hysterectomy Convenience of medical facility location Satisfaction with primary care provider Cervical cancer screening advice Hiatt, et al., 2002 NCI, 2004 Hubbell, et al., 1996 Hiatt, et al., 2002 Hayward, et al., 1988 Hayward, et al., 1988 Coughlin, et al., 2002 Hiatt, et al., 2002 Simoes, et al., 1999 Simoes, et al., 1999 Simoes, et al., 1999 Simoes, et al., 1999 Smith, et al., 2003 Strecher, et al., 1997 Baranowski, et al., 1997 Baranowski, et al., 1997 *Moderating variable

PAGE 91

80 residence, and insurance coverage. Age and so cio-economic status ar e highly associated with screening behavior. For example, older i ndividuals are less likely to be screened for cervical cancer than younger individuals. Wo men with higher levels of education and income are more likely to be screened for cervical cancer. A number of researchers found that there were no significan t regional differences found among Black American and White American women in the use of the Papanicolaou smear test, (Coughlin, Thompson, Seeff, Richards & Stallings, 2002; Hiatt, Klabunde, Breen, Swan & Ballad-Bar bush, 2002). Members of the expert panel advised that region of residence be included in the Preventive Health Model for greater explanat ory power. There are strong and consistent associations between cervi cal cancer screening and havi ng insurance coverage. Women who have insurance are more likely to ha ve received Papanicolaou smears (Hiatt, Klabunde, Breen, Swan & Ballard-Barbash, 2002). Ethnic origin was used in this study in pl ace of the terms race and/or ethnicity used by other researchers. This variable was chosen to remain in model as a component of the background factors construct because, although cervical cancer screening rates are higher among Black American women than among White American women (Makuc, Fried & Kleinman, 1989; Martin, Parker Wingo & Heath, 1996), Black American women have the highest ageadjusted mortality rate fo r cervical cancer among all the ethnic groups with 5.6 cervical cancer d eaths per 100,000. For Hispanic American women the rate is 3.6 cervical cancer deaths per 100,000; Asians/Pac ific Islanders and American Indians/Alaska Natives followed with 2.8 cervical cancer deaths per 100, 000 and White American women with 2.6 cer vical cancer deaths per 100,000. The age-

PAGE 92

81 adjusted mortality rates for cervical cancer for all ethnic groups is 2.9 deaths per 100, 000 (Ries et al., 2004). Marital status and employme nt status are strong covari ates of cervical cancer screening behavior and were kept in the model (Hayward et al,. 1988; Hubbell, Chavez, Mishra & Valdez, 1996). Past screening behavior variables. Simoes, et al. (1999) found that women who had had a mammogram or a clinical breast ex amination were more likely to have had a Papanicolaou smear test than those women who had not. These variables will also be included in the model. Health profile variables. Health profile variables are a component of the background factors construct. These variables are tobacco use, and obesity measured as body mass index [BMI] calculated as weight in kilograms divide d by the square of height in meters. Health profile variables and their associations with cervical cancer screening are not as well understood as thei r associations with breast ca ncer. Simoes, et al. (1999) found that women who smoke cigarettes and are obese are more likely to be noncompliant with the 1999 cer vical cancer sc reening guidelines of the ACS. Although there is a dearth of literature about the relationship between cervical cancer screening behavior and whether or not a woman has had a hysterectomy, members of the expert panelist advise d that this variable be incl uded in the Preventive Health Model for greater explanatory power. Primar y care providers and the American Cancer Society suggest that women who have had a total hysterectomy, with removal of the cervix may not need to continue with cervi cal cancer screening (Smith, Cokkinides, & Eyre, 2003).

PAGE 93

82 Representation, Social Influence and Program Factors Constructs Each one of the following constructs was operationalized by a single variable from the 1999 BRFSS data set. To arrive at a judgment about whether or not to be screened for cervical cancer, individuals cons ider a number of opti ons including practical convenience and personal benefit (Strech er & Rosenstock, 1997). One of the considerations would be the wh ether or not the health care facility is in a convenient location for the individual. This variable was used to measur e the representation factor of practical convenience. Social influence factors form the third set of factors. These factors include the individual’s relationship or rapport with primary care provide rs. This construct was used measured by individual’s satis faction with the care rece ived from her primary care provider (Baranowski, Perry & Parcel, 1997). Program factors comprise the fourth set of factors in the Preventive Health Model. This concept is derived from Social Cognitive Theory and the Health Belief Model and refers to the contacts made by primary car e providers for motivating and reinforcing a given preventive health behavior, such as ce rvical cancer screening (Baranowski, Perry & Parcel, 1997; Strecher & Rosenstock, 1997). In breast cancer screening literature it is well documented that primary care pr ovider recommendation about obtaining a mammogram is strongly associated with adhe rence to breast cancer screening guidelines (Lippert, Eaker, Vierkant & Remington, 1999; O’Malley, Earp, Hawley, Schell, Mathews & Mitchell, 2001; Roetzheim, Fox, Leake & Houn, 1996). The principal investigator was interested in whether primary care provider advice about cervical cancer screening could also be a covariate of adherenc e to cervical cancer

PAGE 94

83 screening guidelines. The 1999 BRFSS does not contain a direct question on receiving advice or a recommendation from a woman’s primary care provider about cervical cancer screening. However, the question: “During th e past year, have you received advice from your doctor or primary care pr ovider about your sexual practic es and sexually transmitted diseases?” is in the data set. After co nsultation the doctoral committee and panel of experts concurred that the question from the 1999 BRFSS data set could be used as a proxy measure for primary care provider advi ce about cervical can cer screening. This variable was used to measure the program factor of screening advice. Summary of the Proposed Study According to the Theory of Reasoned Action (Fishbein & Ajzen, 1975; Ajzen & Fishbein, 1980), the combination of backgr ound, representation, social influence and program factors is likely to influence an indi vidual’s intention to engage in preventive behavior, and intention is a precursor to taking preventive action. The Preventive Health Model posits that background, psychological representation, social influence and program factors are associated with the act of taking preventive action. In this study the principal investigator used the Preventive Health Model to study the covariates of cervical cancer screenin g behavior in the ethnically dive rse population of women selected for this study. The dichotomous dependent or outcome variable was compliance with the 1999 cervical cancer screen ing guidelines of the ACS Bl ack (non-Hispanic), White (nonHispanic), Hispanic, Asian or Pacific Isla nder and American Indi an or Alaska Native women aged 18 to 65 years and over who re ported having had a routine Papanicolaou smear test within a year of being interv iewed for the 1999 BRFSS were classified as

PAGE 95

84 being adherent to the annual cervical sc reening recommendations of the American Cancer Society. Women who had not had a rout ine Papanicolaou smear test within a year of being interviewed for the survey were classified as not being adherent to the compliance with the 1999 cervical cancer screening guidelines of the ACS recommendations of the American Cancer So ciety. The focal independent variable was primary care provider advice about cervical cancer screening. The moderating variable was ethnic origin. The Preventive Health Model was used as a theoretical and explanatory framework for examining the a ssociation between cervi cal cancer screening behavior and primary care provider advice about cervical cancer screening, while controlling for background, representation and social influence factors. Data Management A second review of the lite rature was performed on st udies of covariates of cervical cancer screening beha vior to identify the reporte d magnitude and range of the odds ratios, significant at a conventional alph a level of 0.05, for each of the variables in the proposed study. The odds ratios in the st udies ranged from 1.3 to 13.7. The women in the 1999 BRFSS identified their ethnic origins as (a) White, (b) Black, (c) Hispanic or Spanish, (d) Asian or Pacific Islander, (e) Am erican Indian or Alaska Native, or (f) “other.” A series of power analyses were performed and it was determined that with a conventional alpha level of 0.05 and a beta level of 0.20: (a) odds ratios 1.1 in the entire sample and in the White American women sub sample, (b) odds ratios 1.2 in the Black American women and Hispanic American women sub samples, (d) odds ratios 1.5 in the Asian or Pacific Islander wome n sub sample, (e) odds ratios 1.6 in the American Indian or Alaska Native women sub sample, (f) odds ratios 1.4 in a combined sample of Asian

PAGE 96

85 or Pacific Islander and American Indian or Alaska Native women, would be detected with 95% certainty in the proposed study. Fo llowing the series of power analyses, a decision was made to remove the Asian or Pacific Islander and American Indian or Alaska Native women from the study because of the lack of power to detect odds ratios of at least 1.3, in either sub sample, or both samples combined. The software package used for data mana gement and the statistical analyses was the Statistical Analysis System, SAS version 9.1 (SAS Institute Inc., 2002-2003). The data was analyzed using logistic regression. The justification for this type of analysis is that the dependent variable is dichotomous, and this fact has to be accounted for in the analysis (Hatcher & Stepanski, 1994; M unro, 1997). An identification number was created for each of the observations and used to check for duplicate records in the BRFSS data set. Observations which had data missi ng on any of the variables were deleted. This strategy, complete case an alysis, is known to yield valid in ferences for logistic regression models (Allison, 1999). The frequency procedure in SAS was used to produce frequency distributions for the variables in the BRFSS data set which had been used to operationalize the Preventive Health Model, as explained earlier in Chapter 3. Univariate procedures were used to explore and understa nd the distributional pr operties of the data set. The procedure produced descriptive stat istics including the mean, median, standard deviation, percentiles, kurtosis, skewness and th e extremes table. (Hatcher & Stepanski, 1997). Bivariate analyses were used to investig ate the relationships among the variables. These procedures provided both a test of st atistical significance and a measure of the association. The Pearson product-moment correla tion coefficient was used to test the null

PAGE 97

86 hypothesis that the correlation between two in terval-level variables was zero in the chosen population. The Pearson correlation also assessed the strength of the relationship between two variables, irrespective of the st atistical significance of the relationship. The Pearson correlation ranged from -1.0 to +1.0, wi th the larger absolute values indicating stronger relationships or associ ations between the variables (Hatcher & Stepanski, 1997). Pearson correlation coefficients of 0 .5 (moderate correlation) to 0.9 (strong correlation) were reported. The Chi-square test of homogeneity was used to measure the bivariate associations between the dependent variable (compliance with the 1999 cervical cancer screening guidelines of the ACS) and each of the independent variables in the study sample. The Chi-square tested the null hypothesis that in th is study population the dependent and independent variables were unr elated. The larger the value of the Chisquare, the stronger was the association betw een the two variables in the sample being examined. The cross tabulation of the Chi-s quare test for differences in proportions showed the frequencies, total sample size, mi ssing data, the p-value of the Chi-square for each of the bivariate associations and whethe r or not this value was significant at the conventional alpha level set at 0.05. Logistic Regression Models The data were analyzed using the logist ic regression procedur e. All the models were developed to: (a) determine unadjusted odds ratios, and (b) refined to determine adjusted odds ratios. The variables were adde d to the logistic model according to their order in the Preventive Health Model (Myers et al. 1994; Myers et al 1996). In the first step of the analysis, the focal independent va riable (primary care provider advice about

PAGE 98

87 cervical cancer screening) wa s regressed on the dependent va riable (compliance with the 1999 cervical cancer screening guidelines of the ACS). In the next step the sociodemographic variables operationalizing the background factors construct, were added en bloc to the regression model. The stepwise selec tion process in the l ogistic procedure was used to determine which independent vari ables, statistically significant at the conventional alpha level of 0.05, would be re moved from the model. The difference in the Chi-square values of the -2 Log Likeli hood for the overall mode l compared with the nested model (with fewer independent va riables) was compared. The Chi-square difference was used to make the final decisi on about which variables would be retained or removed from the model. The health pr ofile variables, past screening behavior variables, representation and social influe nce variables were added separately to each model, and the steps repeated until all the variables were entered into a resultant full model. This procedure resulted in the final, best-fitting model, without any interaction terms, containing all the va riables of statistical an d practical significance. An interaction effect is said to exist when the effect of an independent variable on a dependent variable differs de pending on the value of a third variable, referred to as an effect modifier or a mode rator variable (J accard, 2001). In this study the focal independent variable was primary care provi der advice about cervical cancer screening, the dependent variable was compliance with the 1999 cervical cancer screening guidelines of the ACS, and the moderator vari able was ethnic origin. The values of the moderator variable were Black American wo men, White American women and Hispanic American women. The principal investigator was interested in wh ether or not ethnic origin moderated the relationship between pr imary care provider advice about cervical

PAGE 99

88 cancer screening and compliance with the 1999 cervical cancer scr eening guidelines of the ACS In order to evaluate the effect of the moderator variable on the relationship of interest, interaction terms were defined. Annual cervical cancer screeni ng is a dichotomous variable scored 1 if the woman had received a Papanicolaou smear test within one year of being surveyed for the BRFSS, and 0 if the if the woman had not received a Papanicolaou smear test within one year of being surveyed for the BRFSS. The focal pred ictor variable, primar y care provider advice about cervical cancer screening scored 1 if su ch advice was given and 0 if cervical cancer screening advice was not given. Ethnic origin was the modera tor variable and since it had three levels these was represented by two du mmy variables. Primary care provider did not give advice about cervical cancer screen ing is the referenc e for cervical cancer screening advice, and White American women was the reference group for ethnic origin. Product term analysis was used to analyze th e interaction among thes e variables. All the dummy variables for cervical cancer screeni ng were multiplied by all the variables of ethnic origin. This process yielded two produc t terms or interaction terms which were added to the model with all the background, re presentation and social influence factors. This constituted the full and fina l model with interaction terms. The logististic procedure in SAS was used to estimate a binary logit model via maximum likelihood. The Likelihood Ratio Chi-s quare and the Score St atistic tested the hypothesis that all of the expl anatory variables have coe fficients of zero. These two statistics along with the Akaike’s informa tion criterion (AIC) and the Schwarz criterion (SC) were used to compare the relative fit of the data to the various models. In general,

PAGE 100

89 the lower values of these statistics corres ponded to the more desirable models (Allison, 1999). The Wald Chi-square statistic was used to test the null hypothesis that each individual coefficient is equa l to zero. This statistic was examined to find out whether any or all of the two 2-way inte ractions were statistically signi ficant. If at least one of the interactions was found to be significant, the interaction effect w ill be tested by using logistic regression to determine whether the interaction terms significantly improve model fit better than when there were no interaction terms included in the model. A model Chi-square was estimated for the each of the models. The Chi-square for the model without the interaction terms was subt racted from the Chi-square for the model with the interaction terms. The difference in the Chi-square value is distributed as a Chisquare with degrees of freedom equal to the difference in the degrees of freedom between the two models. A statistically significant Chi-square difference would imply that the interaction effect was significant (Jaccard, 2001). The next step was to examine the logistic coefficients for the non-product terms in the interaction model. The main effect for a va riable that is also in a 2-way interaction has to be interpreted as the effect of that va riable when the other va riable in the product term is zero (Allison, 1999). The dummy variable was part of the product term in the equation therefore the logistic coefficient was conditioned on the moderator variable being zero. These points were taken into cons ideration when the effects of variables in the product terms were interpreted (Jaccard, 2001). The next step was to examine the logistic coefficients for the product terms. Primary care provider advice about cervical cancer screening was a dichotomous variable (scored 1 = advice given, 0 = advice not

PAGE 101

90 given). The odds ratio comp aring the women given a dvice about cervical cancer screening by their primary care provider w ith women not given advice about cervical cancer screening by their primary care provide r was calculated for each of the ethnic origins (i. e. each level of the moderating vari able). If the odds rati os were identical in value (except for sampling error) then there wa s no interaction effect. Different values of the odds ratios would indicate that the e ffect of primary care provider advice about cervical cancer screening vari ed depending on the ethnic origin of the women. The results were then evaluated for statistical sign ificance by calculating the p-value and the confidence interval around the point estimate. None of the interaction eff ects was found to be statistica lly significant, therefore the final step in the analysis was to interp ret the meaning of the odds ratios in the main effects model, the model without the interact ion terms. The results were evaluated for statistical significance by calculating the confidence interval ar ound the point estimate and by calculating the p-value (Allison, 1999). Regression Diagnostics These following diagnostic tests were pe rformed on the full and final logistic regression model developed during the mode l building stage of the analysis. The tolerance statistic was used to assess multic ollinearity. High values of the tolerance statistic were correlated with low multicoll inearity. Several sta tistics produced by the logistic procedure was used to measure th e influence of each observation. Influence statistics indicated how much some featur e of the model changes if a particular observation was deleted from the model f it (Allison, 1999). The hat matrix diagonal identified cases, or observations which infl uenced the logistic regression model more

PAGE 102

91 than others. The leverage st atistic values were between 0.0 (no influence on the model) and 1.0 (completely determined the model). Infl uence statistics measured the effect that deleting an observation would have on each of the regression coefficients. The DFBETA statistic indicated cases or observations which were poorly fitted by the model. The statistic measured the change in the logit co efficients if a case was dropped from the model. The cutoff criterion for observations with poor fit was where the DFBETA >1.0. The C and the CBAR statistics are analogs of the Cook’s D in ordi nary least squares (OLS) regression, and are a third measure of influence on an individual observation. These statistics are standardized measures of the approximate change in all regression coefficients which would occur if an individu al observation was deleted from the model. The Pearson (RESCHI) and deviance (RESDEV) residuals are standardized residuals used to identify observations that were not well explained by the model. The cutoff criterion for observations with poor fit wa s where the RESCHI >2.0, and RESDEV >3.0 (Allison, 1999). These procedures were used to evaluate the study null hypotheses that: (a) there is no association among the selected covariates of cervical cancer sc reening behavior, and compliance with the 1999 cervical cancer sc reening guidelines of the ACS in an ethnically diverse population of American women, (b) there is no difference in the magnitude of the associa tion between ethnic origin and compliance with the 1999 cervical cancer screening guide lines of the ACS, (c) there is no association between primary care provider advice about cervical cancer screening behavior and compliance with the 1999 cervical cancer sc reening guidelines of the ACS, and (d) ethnic origin does

PAGE 103

92 not moderate the association between primary care provider advice about cervical cancer screening and compliance with the 1999 cervical cancer screen ing guidelines of the ACS.

PAGE 104

93 CHAPTER FOUR: RESULTS Description of the Study Sample The results of the data analysis to answer the research questions for this study are presented in this chapter. The study population selected from the 1999 BRFSS data set consisted of: (a) Black (non-Hispanic), White (non-Hispanic) and Hispanic (Black and White) American women 18 to 65 years and over (b) who reported that they had received a preventive Papanicolaou smear, and (c) resided in the United States. Women in the data set (a) who had never had a preventive Papa nicolaou smear, (b) who had had a diagnostic Papanicolaou smear or were being treated for an abnormal Papanicolaou smear, (c) had been treated for an abnormal Papanicolaou sm ear, or (d) resided in Puerto Rico were excluded from the study. The women res ponded to questions posed to them by researchers from the Centers for Diseas e Control and Prevention (CDC, 1999). The data were analyzed to see if ther e were any significant differences among the women who answered all of the questions asked by the research ers, and those who refused to answer some of the questions posed. From the frequency analyses, it was found that White American women (11.5%) were the most likely to refuse to answer questions, followed by Black American wo men (3.0%), and then Hispanic women (0.5%). From the logistic regression analys es, it was found that Black American women (OR=1.7, 1.4-1.9) were more likely to have answ ered the questions than White American women. Black American, White America n, and Hispanic American women

PAGE 105

94 (OR=0.7, 0.5-0.9) who had an annual household income of less than $25, 000 were also significantly more likely to ha ve answered than women in the study sample with higher annual incomes. The questions to which the response was eith er “don’t know/not sure” or “refused” were then deleted from the da ta set. The study popula tion consisted of 52,878 (84.7%) White, 5,742 (9.2%) Black, and 3,795 (6.1 %) Hispanic American women, a total of 62,415 observations, with one record per individual. The focal independent variable in this study was cervical cancer screening advice The question “During the past year, have you received advice from your primary care provider about your sexual practices a nd sexually transmitted diseases?” was included in one of the optional modules in the 1999 BRFSS questionnaire. This question is a proxy for the question “H as your primary care provide r ever advised you about cervical cancer screening?” Given that the qu estion was not from the core section of the BRFSS questionnaire, the resear chers from the CDC were not mandated to ask this question of women in all the 1999 BRFSS repor ting regions of the United States. These optional CDC modules are sets of questions on sp ecific topics that stat es can elect to use on their own questionnaires (CDC, 1999). This question was only asked of women living in the states of Louisiana, Missouri, South Carolina, Texas, Virginia a nd Wyoming who were 18 to 64 years of age. The study sample, therefore, consisted of: (a) Black (non-Hispanic), White (nonHispanic) and Hispanic (Black and White) American women, 18 to 64 years of age who (b) reported that they had r eceived a preventive Papanicola ou smear test, and (c) resided in the states of Louisiana, Missouri, Sout h Carolina, Texas, Virginia and Wyoming. The

PAGE 106

95 study sample consisted 4,302 (79.8%) Wh ite, 873 (16.2%) Black, and 217 (4.0%) Hispanic American women, a total of 5,392 obser vations, with one record per individual. To assess for differences between the wome n in the study sample and those in the study population, all the women wh o had been asked about whether or not they had been advised by their primary health provider about cervical cancer scre ening were coded 1. There were 5,392 women in the study sample who were from the study population of 62,415 women. All the women who had not been asked about whether or not they had been advised by their primary health provi der about cervical cancer screening, and who originally had missing data fo r this variable were code d 0. Therefore 57,023 women out of 62,415 remained in the population of interest. To assess for selection bias in the study sample, the association between the focal independent variable, cervical cancer screening advice and the dependent variable, last Pap smear for subjects in the study population and in the study sample were measured usi ng the Chi-square test of homogeneity. Significant differences were found betw een the women in the study population and those in the study sample ( 2 = 26.0468, p=<.0001). The women in the study sample were younger (18 to 64 years of age) than t hose in the study population (18 to 99 years of age). The ethnic origin of the women in the study sample was significantly different from those in the study population. Black Ameri can women comprised 16.2% of the study sample compared to 9.2% of the women in the study population. Hispanic American women comprised 4.0% of the study sample and 6.1% of the study population. White American women comprised 79.8% of th e study sample and 84.7% of the study population. The region of residence differed significantly between the women from the study sample and the women from the study population. There were not any women from

PAGE 107

96 the Northeast included in the study sample, whereas they accounted for 16.0% of the study population. Fifty-seven percent of the wo men in the study sample resided in the South, compared to 36.7% of the women in the study population. A summary of questions selected from the BRFSS ques tionnaire, the corresponding variables hypothesized to be associated with complia nce to the 1999 cervical cancer screening guidelines of the ACS, and the response cat egories are displayed in Table 2. Details pertaining to the response categories are discussed later in the chapter. Constructs of the Preventive Health Model The Preventive Health Model posits that a broad set of f actors influence an individual’s decision to be screened for cervical cancer (Myers et al., 1994). The following elements make up the constructs of the model. Background Factors Construct This construct of the Preventive Health M odel consisted of thr ee different types of variables; socio-demographic, past screening behavior and he alth profile variables. These diverse variables may influence a woman’s cervical cancer screening behavior. Socio-demographic variables. Socio-demographic variab les included age, ethnic origin, marital status, educational level, employment status, income level, region of residence, and insurance coverage. Ethnic origin was the moderator variable a nd because it consiste d of three levels: (a) Black, non-Hispanic, (b) White, non-Hispan ic, and (c) Hispanic (Black or White) it was represented by two dummy variables wi th the White, non-Hispanic American women as the referent group.

PAGE 108

97Table 2 Variables Associated with Compliance with the 1999 Cervical Cancer Screening Guidelines of the ACS in the Study Sample Questions from the 1999 BRFSS Survey Questionnaire Variables from 1999 BRFSS Response Categories What is your age? What is your race/are you of Spanish or Hispanic origin? Are you married? What is the highest grade or year of school you completed? Are you employed? Is your annual household income from all sources more than $25,000? Census Bureau Regions: Region 1, Region 2, Region 3, Region 4 Do you have any kind of health care coverage? Have you ever had a mammogram? Have you ever had a clinical breast exam? Do you smoke cigarettes, every day, some days, or not at all? BMI more than 27.3 kg/m2 Have you had a hysterectomy? Child-bearing age Ethnic origin Black Hispanic White Marital status Education level Employment status Income level State FIPS codes Have health coverage Had mammogram Had breast exam Smoking status Body Mass Index Had hysterectomy 18-44 years/45-64 years Black, non-Hispanic Hispanic, (Black or White) White, non-Hispanic* Yes/No* Some college education Less than college education* Yes/No* Yes/No* Northeast, Midwest, South*, West Yes/No* Yes/No* Yes/No* Yes*/No Yes*/No Yes*/No *Referent group

PAGE 109

98Table 2 continued Variables Associated with Compliance with the 1999 Cervical Cancer Screening Guidelines of the ACS In the Study Sample Questions from the 1999 BRFSS Survey Questionnaire Variables from 1999 BRFSS Response Categories How would you rate the convenience of your medical facility location?** How would you rate satisfaction with your primary care provider?** Has your primary care provider ever advised you about cervical cancer screening? How long since your last Pap smear? Convenience of medical facility location Satisfaction with primary care provider Cervical cancer screening advice Last Pap smear Good/Poor* Good/Poor* Yes/No* Within 1 year/Over 1 year* Referent group ** Questions asked in Virginia Dummy variables also were created for marital status (married versus not married), education level (have some co llege education versus have no college education) and income level (annual income of $25,000 or mo re versus annual income of less than $25,000). The age variable was defined as bein g of childbearing age (18 to 44 years of age) or over childbearing age (45 to 64 year s of age).The insurance coverage variable was described to the respondents as having (or not having) “any kind of health care coverage, including health insu rance, prepaid plans such as HMOs or government plans such as Medicare” (CDC, 1999, p.5). The region of residence variable was constructed from the United States census bureau regi ons with federal information processing standards (FIPS) codes (Table 3). The “FIPS codes are a standardized set of numeric or alphabetic codes issued by the National Instit ute of Standards and Technology (NIST) to ensure uniform identification of geographi c entities through all federal government agencies” (U.S. Census Bureau, 2003).

PAGE 110

99Table 3 Census Bureau Regions of the United States with State FIPS Codes Region 1: Northeast Connecticut (09) New Jersey (34) Maine (23) New York (36) Massachusetts (25) Pennsylvania (42) New Hampshire (33) Rhode Island (44) Vermont (50) Region 2: Midwest Indiana (18) Iowa (19) Illinois (17) Kansas (20) Michigan (26) Minnesota (27) Ohio (39) Missouri (29) Wisconsin (55) Nebraska (31) North Dakota (38) South Dakota (46) Region 3: South Delaware (10) Alabama (01) Arkansas (05) District of Columbia (11) Kentucky (21) Louisiana (22) Florida (12) Mississippi (28) Oklahoma (40) Georgia (13) Tennessee (47) Texas (48) Maryland (24) North Carolina (37) South Carolina (45) Virginia (51) West Virginia (54) Region 4: West Arizona (04) Alaska (02) Colorado (08) California (06) Idaho (16) Hawaii (15) New Mexico (35) Oregon (41) Montana (30) Washington (53) Utah (49) Nevada (32) Wyoming (56) Modified from U.S. Census Bureau, Geography Division, 2003

PAGE 111

100 Past screening behavior variables. Simoes, et al. (1999) found that women who had had a mammogram (had a mammogram vers us not had a mammogram) or a clinical (i.e. physical) breast examination (had a phys ical breast exam vers us not had a physical breast exam) at some time during their lives, were more likely to have had a Papanicolaou smear test than those wome n who had not. These variables were also included in the model. Health profile variables. Health profile variables were a component of the background factors construct. These variables we re tobacco use status, and obesity measured as body mass index [BMI] calculated as weight in kilograms divided by the square of height in meters, and whether or not a woman has had a hysterectomy. Health profile variables and their associations with cervical cancer screening are not as well understood as their associations with breas t cancer. Simoes, et al. (1999) found that women who are obese, and smoke tobacco, are more likely to be non-compliant with cervical cancer screening guidelines. Obesity was measured as not obese (BMI < 27.3 kg/m2) and obese (BMI 27.3 kg/m2). Women in the 1999 BRFSS were asked “Do you smoke cigarettes every day, some days or not at all?” Tobacco use status was measured as smoke and do not smoke. Alt hough there is a dearth of literature about the relationship between cervical cancer scr eening behavior and wh ether or not a woman has had a hysterectomy (not had hysterectomy versus had hysterectomy) the variable was included in the Preventive Health Model to see if greater explanatory power could be gained.

PAGE 112

101 Representation, Social Influence and Program Factors Constructs Each one of these constructs of the PH M was operationalized by a single variable from the 1999 BRFSS data set. To arrive at a judgment about whether or not to be screened for cervical cancer, individuals cons ider a number of opti ons including practical convenience and personal benefit (Strech er & Rosenstock, 1997). One of the considerations was whether or not a health care facility is conveniently located. This variable was used to measure the representa tion factor construct of practical convenience of medical facility (good versus poor) in terms of adequacy of location. Social influence factors formed another se t of factors. In this study there is a single social influence factor which is th e woman’s relationship or rapport with her primary care provider. The social influen ce factor construct was measured by the woman’s satisfaction (good vers us poor) with the care rece ived from her primary care provider (Baranowski, Perry & Parcel, 1997). Program factors comprised an additional se t of factors in the Preventive Health Model. This concept, derived from Social C ognitive Theory and the Health Belief Model, referred to the contacts made by primary car e providers for motiva ting and reinforcing a given preventive health behavior, such as ce rvical cancer screening (Baranowski, Perry & Parcel, 1997; Strecher & Rosenstock, 1997). In the breast cancer screen ing literature it is well documented that primary care pr ovider recommendation about obtaining a mammogram is strongly associated with comp liance to breast cancer screening guidelines (Lippert, Eaker, Vierkant & Remington, 1999; O’Malley, Earp, Hawley, Schell, Mathews & Mitchell, 2001; Roetzheim, Fox, Leak e & Houn, 1996). A single variable, advice

PAGE 113

102 about cervical cancer screeni ng received from a primary care provider (yes versus no) was used to measure the program factor construct of the PHM in this study. Data Analyses Frequency Analyses Frequency procedures were used to obtain descriptive statistics on the characteristics of women in the study sample. The characteristics of this sample, presented in terms of the constructs of the Preventive Health Model, are summarized in Table 4. The questions “Thinking of the distance or time you travel to get to the place you usually go to [ medical facility ] how would you rate the c onvenience of that place?” and “How would you rate your satisfaction wi th the overall health care provided by your primary care provider?” were asked of women in Virginia (CDC, 2002). A total of 1,081 women in Virginia answered the question a bout the practical convenience of the location of their medical facility. The women (97.3%) rated the convince of the location as being “very good” and 2.7% of the women rated the convenience as being “poor.” A total of 1,198 Virginian women answered the ques tion on rapport with their primary care provider. The women (96.9%) rated rapport with their primary care providers as being “very good” and 3.1% of the women rated ra pport as being “poor.” Given that the questions were only asked of women in one of the six reporting sites in the study sample, these variables were not included in fu rther analysis of the data set. There were statistically significant differe nces in the sources of health care coverage among the women in the study sample. The major sources of health coverage for the Black American women were thr ough their employers (76.5%), through someone else’s employer (9.5%) and through Medi caid or Medical Assistance (7.6%).

PAGE 114

103Table 4 Descriptive Statistics of the Study Sample Constructs of the PHM # Black Women (16.2%) Number (%) # Hispanic Women (4.0%) Number (%) # White Women (79.8%) Number (%) # Study Sample (100.0%) Number (%) Age 18 44 years 45 – 64 years Ethnic origin Black Hispanic White Marital status Married Not married Educational level Some college education No college education Employment status Employed Unemployed Income level (annual) At least $25,000 Less than $25,000 Insurance Coverage Yes No Region of residence Midwest South West Mammography Had mammogram Not had mammogram Breast exam Had breast exam Not had breast exam 595 (11.0%) 278 (5.2%) 873 (16.2%) 268 (5.0%) 605 (11.2%) 412 (7.6%) 461 (8.6%) 649 (12.0%) 224 (4.2%) 375 (7.0%) 498 (9.2%) 667 (12.4%) 206 (3.9%) 86 (1.6%) 785 (14.6%) 2 (0.04%) 512 (9.5%) 361 (6.7%) 744 (13.8%) 129 (2.4%) 156 (2.9%) 61 (1.1%) 217 (4.0%) 124 (2.3%) 93 (1.7%) 119 (2.2%) 98 (1.8%) 148 (2.7%) 69 (1.3%) 128 (2.4%) 89 (1.7%) 166 (3.1%) 51 (1.0%) 25 (0.7%) 146 (2.7%) 46 (0.9%) 106 (2.0%) 111 (2.1%) 196 (3.6%) 21 (0.4%) 2616 (48.5%) 1686 (31.3%) 4302 (79.8%) 2698 (50.0%) 1604 (30.0%) 2598 (48.2%) 1704 (31.6%) 3140 (58.2%) 1162 (21.6%) 3042 (56.4%) 1260 (23.4%) 3694 (68.5%) 608 (11.3%) 1347 (25.0%) 2144 (39.8%) 811 (15.0%) 2443 (45.3%) 1859 (34.5%) 4020 (74.6%) 282 (5.2%) 3367 (62.4%) 2025 (37.6%) 873 (16.2%) 217 (4.0%) 4302 (79.8%) 3090 (57.3%) 2302 (42.7%) 3129 (58.0%) 2263 (41.9%) 3937 (73.0%) 1455 (27.0%) 3545 (65.8%) 1847 (34.3%) 4527 (84.0%) 865 (16.0%) 1458 (27.0%) 3075 (57.0%) 859 (15.9%) 3061 (56.8%) 2331 (43.2%) 4960 (92.0%) 432 (8.0%) Percentages may not equal 100% because of rounding.

PAGE 115

104Table 4 continued Descriptive Statistics of the Study Sample Constructs of the PHM # Black Women (9.2%) Number (%) # Hispanic Women (6.1%) Number (%) # White Women (84.7%) Number (%) # Study Sample (100.0%) Number (%) Tobacco use Smoke Do not smoke Obesity (BMI) Obese Not obese Hysterectomy Had hysterectomy Not had hysterectomy Practical convenience* Very good Poor Rapport with primary care provider* Very good Poor Screening advice Yes No Compliance with the 1999 cervical cancer screening guidelines of the ACS guidelines Yes No 173 (3.2%) 700 (13.0%) 450 (8.4%) 423 (7.8%) 174 (3.2%) 699 (13.0%) 222 (20.5%) 3 (0.3%) 236 (19.7%) 13 (1.1%) 403 (7.5%) 470 (8.7%) 698 (13.0%) 175 (3.3%) 40 (0.7%) 177 (3.9%) 75 (1.4%) 142 (2.6%) 31 (0.6%) 186 (3.5%) 34 (3.2%) 2 (0.2%) 40 (3.3%) 0 (0.0%) 82 (1.5%) 135 (2.5%) 165 (3.1%) 52 (1.0%) 1161 (21.5%) 3141 (58.3%) 1355 (25.1%) 2947 (54.7%) 900 (16.7%) 3402 (63.1%) 796 (74.0%) 24 (2.2%) 886 (74.0%) 23 (1.9%) 1426 (26.5%) 2876 (53.3%) 3602 (56.8%) 1240 (23.00%) 1374 (25.5%) 4018 (74.5%) 1880 (34.9%) 3512 (65.1%) 1105 (20.5%) 4287 (79.5%) 1052 (97.3%) 29 (2.7%) 1162 (96.9%) 36 (3.0%) 1911 (35.4%) 3481 (64.6%) 3925 (72.8%) 1467 (27.2%) Percentages may not equal 100% because of rounding. The major sources of health coverage for Hispanic and White American women were virtually identical. Hisp anic women were covered thr ough their employers (52.8%), provided through someone else’s employer (20.9 %) or through indemnity plans (10.4%). The major sources of health coverage fo r White American women were through their

PAGE 116

105 employers (55.7%), through someone else’s employer (27.4%) or through indemnity plans (9.6%). The different sources of health care coverage are su mmarized in Table 5. Table 5 Sources of Health Care Coverage for the Study Sample Health Care Coverage Black Women Number (%) Hispanic Women Number (%) White Women Number (%) Health care coverage provided by employer 443 (76.5%) 86 (52.8%) 1983 (55.7%) Health care coverage provided by someone else's employer 55 (9.5%) 34 (20.9%) 978 (27.4%) Indemnity plan 17 (2.9%) 17 (10.4%) 342 (9.6%) Medicare 0 (0.0%) 0 (0.0%) 5 (0.1%) Medicaid or Medical Assistance 44 (7.6%) 15 (9.2%) 108 (3.0%) CHAMPUS, VA 15 (2.6%) 9 (5.5%) 102 (2.9%) Indian Health Service 0 (0 .0%) 1 (0.6%) 1 (0.03%) Some other source 5 (0.9%) 1 (0.6%) 39 (1.1%) Total 581 (100%) 163 (100%) 3558 (100%) Percentages may not equal 100% because of rounding. Univariate Analyses The data were screened using th e univariate procedure in SAS to review the quality of the data before conducting more sophisticated analyses. The univariate procedure produced a number of descriptive statistics includi ng skewness, kurtosis, and a table of extreme observations for identifying outli ers in the data set. This feature of SAS also produced a significance test for the nu ll hypothesis that the data came from a normally distributed population. It is important to test for normality of the data because extremely non-normal data may lead to errone ous conclusions in inferential statistical analyses. If the assumption of normality is vi olated, the likelihood of making a Type I or alpha error (i.e. rejecting the null hypothesis when it is true) increases. A Type II or beta

PAGE 117

106 error results when one fails to reject the null hypothesis when it is false. The second problem with extremely non-normal data is that they may cause bias in the correlation coefficients. Logistic regre ssion does not need the depende nt variable to be normally distributed, neither does it assume a linear relationship between the dependent and independent variables (Tabachnick & Fidell, 1996). Therefore, thes e restrictions do not apply to the analysis of this data set. The univariate procedure produ ced a number of descrip tive statistics of use in screening the quantitative variables. Sel ected values for the skewness and kurtosis statistics are displayed in Table 6. Table 6 Summary of Selected Descriptive Univariate Statistics Skewness and Kurtosis for Variables in the Study Sample Variable Name Skewness Kurtosis Had breast exam -3.1 7.6 Hispanic ethnic origin 4.7 19.9 The skewness statistic characterized the degree of asymmetry of the variable distribution around its mean. The variable had breast exam exhibited negative skewness indicating that the distributi on had an asymmetrical tail that extended towards the lower values in the distri bution. The variable Hispanic ethnic origin exhibited positive skewness that indicated that the distributions had asymmetrical tails extending towards the higher values in the distri bution. Normal distributions pr oduce a skewness statistic of about zero, with values of 2.0 standard errors of skewness.

PAGE 118

107 Kurtosis characterized the relative peak edness or flatness of the distributions compared to the normal distribution. Mes okurtic (normal) distributions produce a kurtosis statistic of about zero. Positive kurto sis is indicated by a relatively peaked, or leptokurtic distribution, and negative kurtosis is indicated by a relatively flat, or platykurtic distribution. The va riables in Table 6 displayed leptokurtic distributions. The other variables in the data set had values of the standard errors of the skewness statistic ranging from -1.5 to + 1.8, and values of the standard errors of the kurtosis statistic ranging from -1.9 to + 1.7. The table for extrem e observations was examined to identify any observations that had contri buted outliers to the data set. No such observations were found. Bivariate Analyses The correlation procedure further facili tated decisions on which variables were redundant in the model building process, a nd could, therefore, be removed from the analyses. This procedure was used to assess the strength of association between each of the variables in the study sample. The Pear son correlation coefficient ranges from no correlation (0.0) to perfect correlation ( 1.0). The associations of variables with a Pearson correlation of 0.3 and higher were reported. The results of the correlation analyses are summarized in Table 7. The strength of the correlations varied from the absolute value of 0.3 (weak) to the absolute value of 0.6 (moderate). The negative value of the coefficient between had hysterectomy had mammogram meant that if a women had had a mammogram, she was less likely to have had a hysterectomy. The negative value of the coefficient between child-bearing age had mammogram meant that the younger the woman in the study

PAGE 119

108 sample, the less likely she was to have had a mammogram. There were no strongly correlated variables (absolute values of 0.7 to 0.9). Therefore, variables were not dropped from the model at this point in the analyses. Table 7 Summary of the Correlation Analyses: Assessment of the Strength of Association between each of the Variables in the Study Sample Variable Names Pearson Coefficient Strength of Correlations P-value Employment status Education level 0.3 weak correlation <.0001 Had hysterectomy Had mammogram -0.3 weak correlation <.0001 Income level Have health coverage 0.3 weak correlation <.0001 Education level Had mammogram 0.4 moderate correlation <.0001 Marital status Income level 0.4 moderate correlation <.0001 Child-bearing age Had mammogram -0 .5 moderate correlation <.0001 Child-bearing age Had hysterectomy 0.4 moderate correlation <.0001 Child-bearing age Cerv ical cancer screening advice 0.3 weak correlation <.0001 The Chi-square test of homogeneity was used to measure the bivariate associations between the dependent variable (compliance with the 1999 cervical cancer screening guidelines of the ACS) and each of the independent variables. The Chi-square tested the null hypothesis that the dependent and independent variab les were unrelated. The larger the value of the Chi-square, the stronger was the association between the two variables under examination in the sample. The cross tabulation of the Chi-square test for differences in proportions showed the frequenc ies, total sample size, missing data, the pvalue of the Chi-square for each of the biva riate associations and whether or not this value was statistically significant at the conventional alpha leve l of 0.05 (Hatcher & Stepanski, 1994). The results for the Chi-square analyses are reported in Table 8. The

PAGE 120

109 bivariate associations between marital status and last Pap smear (p = 0.2844) and Hispanic ethnic origin (p = 0.2731) were not stat istically significant. Table 8 Chi-Square Analyses: Measurement of the Bivariate Associations between each Independent Variable and the Dependent Variable in the Study Sample Cross Tabulations Chi-Square Value P-value Education level Last Pap smear 42.6 <.0001 Employment status Last Pap smear 5.2 0.0223 Had mammogram Last Pap smear 4.1 0.0427 Had hysterectomy Last Pap smear 65.0 <.0001 Have health coverage Last Pap smear 113.3 <.0001 Income level Last Pap smear 30.0 <.0001 Marital status Last Pap smear 1.1 0.2844 Cervical cancer screening advice Last Pap smear 118.2 <.0001 Had breast exam Last Pap smear 90.7 <.0001 Smoking status Last Pap smear 36.6 <.0001 Child-bearing age Last Pap smear 88.7 <.0001 Black ethnic origin Last Pap smear 30.0 <.0001 Hispanic ethnic origin* Last Pap smear 1.2 0.2731 White ethnic origin Last Pap smear 28.1 <.0001 Body Mass Index Last Pap smear 33.8 <.0001 State FIPS codes Last Pap smear 71.4 <.0001 Logistic Regression Analyses After using the PHM as a guide for selec ting the preliminary variables from the 1999 BRFSS, the data were analyzed using stepwise, binomial logistic regression featuring nested models. The justification for us ing this type of analys is is the fact that the dependent variable (compliance with th e 1999 cervical cancer sc reening guidelines of the ACS) is a categorical variab le and this feature had to be accounted for in the analysis. Logistic modeling was used to find the best model to explain the association between the focal independent variable, cervical cancer screening advice and compliance with the

PAGE 121

110 1999 cervical cancer screening guidelines of the ACS, while simultaneously controlling for confounding of the association by the other independent variables. The first step in this procedure was to construct models of the unadjusted odds ratio estimates for each of the independent variables to see which of the variables was an independent determinant of compliance with the 1999 cervica l cancer screening guidelines of the ACS. Each independent variable was regressed on the dependent variable. The results of this procedure are summarized in Table 9. Table 9 Models of Unadjusted Odds Ratio Estimates for Covariates of Compliance with the 1999 Cervical Cancer Screening Guidelines of the ACS Variable estimate Unadjusted OR 95% CI P-value Child-bearing age 0.58 1.79 1.59 – 2.02 <.0001 Black ethnic origin 0.47 1.60 1.34 – 1.91 <.0001 Hispanic ethnic origin 0.18 1.19 0.87 – 1.64 0.2737 White ethnic origin -0.43 0.65 0.55 – 0.76 <.0001 Marital status -0.07 0.94 0.83 – 1.06 0.2844 Education level 0.40 1.49 1.32 – 1.69 <.0001 Employment status 0.16 1.17 1.02 – 1.33 0.0224 Income level 0.32 1.40 1.24 – 1.59 <.0001 State FIPS codes -0.00 1.00 0.99 1.00 0.3065 Have health coverage 0.81 2.24 1.93 – 2.61 <.0001 Had mammogram 0.12 1.13 1.00 – 1.28 0.0428 Had breast exam 0.94 2.57 2.10 – 3.13 <.0001 Smoking status 0.41 1.50 1.32 – 1.72 <.0001 Body Mass Index 0.37 1.44 1.27 – 1.63 <.0001 Had hysterectomy 0.58 1.78 1.54 – 2.04 <.0001 Cervical cancer screening advice 0.75 2.11 1.84 2.41 <.0001 The variables Hispanic ethnic origin, marital status, employment status and state FIPS codes were not statistically signifi cant determinants of compliance. White ethnic origin had a statistically significant inverse asso ciation with screening compliance. That is, White American women were, independent of any other covari ates, less likely to

PAGE 122

111 comply with the 1999 cervical cancer screen ing guidelines of the ACS. All the other variables were independent determinants of cervical cancer screening compliance. All of the variables were retained for modeling th e adjusted odds ratio for cervical cancer screening compliance with logistic regressi on procedures. The PHM was used to guide the construction of the regression models. The next step in the analyses was to build the best fitting model for the variables. The focal independent variable, cervical cancer screening advice, was regressed on the dependent variable last Pap smear The results are reported in Table 9. Last Pap smear was retained as a variable in the model. In the next step, all of the soci o-demographic variables were added en bloc to the regression model in the order outlined by the PHM. The iterati ve stepwise procedure in SAS was used to guide the selection of the be st variables to incl ude in the regression model for the background factor s construct of the PHM. The residual Chi-square test was used to assess the overall logistic model by comparing the difference in the Likelihood Ratio Chi-square between the ove rall model and the nested model. In general, if the pvalue of the Chi-square at the alpha leve l of 0.05 was statistically significant, the variables were retained in the model. If th e p-value was not statistically significant, the redundant variables were dropped from the mode l. This process wa s repeated until the full regression model containing all the releva nt variables was obtained (Table 10). The iterative, stepwise, logistic pr ocedures used in building the full, and final logistic model are presented in Appendix B. Two interaction terms ( interaction_Black ethnic origin and interaction_Hispanic ethnic origin ) were formed to evaluate the effect of the moderator variable ethnic origin

PAGE 123

112 on the relationship between primary care provi der recommendation and cervical cancer screening compliance. Ethnic origin consisted of three levels Black ethnic origin, Hispanic ethnic origin and the referent level, White ethnic origin Table 10 Model H: Final Model of Adjusted Odds Ratio Estimates for Covariates of Compliance with the 1999 Cervical Cancer Screening Guidelines of the ACS Variable Name estimate Adjusted OR 95% CI P-value Child-bearing age 0.57 1.78 1.50 – 2.08 <.0001 Black ethnic origin 0.60 1.83 1.51 – 2.22 <.0001 Hispanic ethnic origin 0.30 1.34 1.00 – 1.89 0.0847 Have health coverage 0.80 2.23 1.90 – 2.63 <.0001 Had mammogram 0.60 1.82 1.55 – 2.14 <.0001 Had breast exam 0.84 2.32 1.87 – 2.88 <.0001 Smoking status 0.30 1.35 1.17 – 1.55 <.0001 Body mass index 0.38 1.46 1.28 – 1.67 <.0001 Had hysterectomy 0.47 1.60 1.36 – 1.88 <.0001 Cervical cancer screening advice 0.63 1.90 1.61 – 2.17 <.0001 The interaction terms were added to the model (Table 11). The iterative stepwise procedure was used to select th e variables that best fit the logistic regression model. The model Chi-square was used to assess th e overall logistic model by comparing the difference in the Likelihood Ratio Chi-square between the full model and the full model with interaction terms. The residual Chi-squa re test was used to determine which model was a better fit for the data. The p-value of the Chi-square at the alpha level of 0.05 was not statistically sign ificant, (p = 0.6998). Therefore th e interaction terms were dropped from the model. Although Hispanic ethnic origin was not statistically significant it was forced to remain in the model because the Hispanic American women needed to be controlled for in the analysis. The final re gression model is displayed in Table 10.

PAGE 124

113Table 11 Model I: Adjusted Odds Ratio Estimates for Covariates of Compliance with the 1999 Cervical Cancer Screening Guidelines of the ACS with Interaction Terms Variable estimate Adjusted OR 95% CI P-value Child-bearing age 0.57 1.78 1.50 – 2.08 <.0001 Black ethnic origin 0.62 1.83 1.51 – 2.22 <.0001 Hispanic ethnic origin 0.21 1.34 1.00 – 1.89 0.3007 Have health coverage 0.82 2.23 1.90 – 2.63 <.0001 Had mammogram 0.60 1.82 1.55 – 2.14 <.0001 Had breast exam 0.84 2.32 1.87 – 2.88 <.0001 Smoking status 0.30 1.35 1.17 – 1.55 <.0001 Body mass index 0.38 1.46 1.28 – 1.67 <.0001 Had hysterectomy 0.47 1.60 1.36 – 1.88 <.0001 Interaction_Black ethnic origin -0.03 1.00 0.66 – 1.44 0.8768 Interaction_White ethnic origin 0.32 1.38 0.64 – 3.00 0.4162 Cervical cancer screening advice 0.62 1.86 1.56 – 2.19 <.0001 Regression Diagnostics Analyses The following diagnostic tests were perfor med on the final logistic regression model developed during the model building stag e of the analysis. The tolerance statistic was used to assess multicollinearity. High values of the tolerance statistic were correlated with low multicollinearity. Th e tolerance statistic ranged from 0.64 to 0.98 indicating low multicollinearity among the independent variab les. Several statistics produced by the logistic procedure were used to measure the influence of each observation. Influence statistics indicated how much a given feature of the model changes if a particular observation is deleted from the model fit (Allis on, 1999). The influence statistics are the: (a) hat matrix diagonal sta tistic, (b) DFBETA statistic, (c) C statistic and (d) CBAR statistic. The hat matrix diagonal, also descri bed as a leverage statis tic, identified cases, or observations that influenced the logist ic regression model more than others. The leverage statistic values were between 0.0 (no influence on the model) and 1.0

PAGE 125

114 (completely determined the model). Influence st atistics measured the effect that deleting an observation would have on each of the regression coefficients. The DFBETA statistic indicated cases or observations that were poorly addressed by the model. The DFBETA statistic measured the change in the logit co efficients if a case was dropped from the model. The cutoff criterion for observations with poor fit was where the DFBETA was >1.0. The C and the CBAR statistics are a th ird measure of influence on an individual observation. These statistics are standardized measures of the approximate change in all regression coefficients that would occur if an individual observation was deleted from the model. The Pearson (RESCHI) and deviance (RESDEV) residuals are standardized residuals used to identify observations that are not well explained by the model. The cutoff criterion for observations with poor fit was where the RESCHI was >2.0, and RESDEV was >3.0 (Allison, 1999) These procedures were used to evaluate each of study null hypotheses asso ciated with the research questions. Summary of the Data Analyses Research Question 1 The first research question posed whethe r there was an association between each of the selected covariates of cervical cancer screening behavior a nd compliance with the 1999 annual cervical cancer scr eening guidelines of the ACS as recommended in this ethnically diverse population of American wo men. The null hypothesis posited that there was no association between each of the se lected determinants of cervical cancer screening behavior, and compliance with the 1999 cervical cancer scre ening guidelines of the ACS in this ethnically dive rse population of American women.

PAGE 126

115 Nine of the fourteen selected covariat es (64.3%) of cervica l cancer screening behavior were found to be significantly a ssociated with the dependent variable compliance with the 1999 annual cervical screening guidelines of the ACS. The independent covariate variables were black ethnic origin, child-b earing age, have health coverage, had mammogram, had breast exam, smoking status, body mass index, had hysterectomy and cervical cancer screening advice (Table 10). Black ethnic origin. Black American women we re 1.83 (CI=1.51-2.22) times more likely than the referent group, White American women, to have had an annual Papanicolaou smear test as recomme nded by the 1999 guidelines of the ACS. Child-bearing age. Black, Hispanic and White Am erican women who were of childbearing age (18 to 44 years of age) were 1.78 (CI=1.50-2.08) times more likely to have had an annual Papanicolaou smear te st than those women who were not of childbearing age (45 to 64 years of age). Have health coverage. Women who had health coverage were 2.23 (CI=1.90-2.63) times more likely to have ha d an annual Papanicolaou smear test than women who did not have health coverage. Had mammogram. Women who had had a mammogram were 1.82 (CI=1.55-2.14) times more likely to have ha d an annual Papanicolaou smear test than women who had not ever had a mammogram. Had breast exam. Women who had had a clinical (physical) breast exam were 2.32 (CI=1.87-2.88) times more likely to have had an annual Papanicolaou smear test than women who in the study sample who had not had a breast exam.

PAGE 127

116 Smoking status. Women who were non-smokers were 1.35 (CI=1.17-1.55) times more likely to have had an annual Papani colaou smear test than women who were smokers. Body mass index. Women who were not obese (BMI <27.3 kg/m2) were 1.46 (CI=1.28-1.67) times more likely to have had an annual Papanicolaou smear test than women who were obese (BMI 27.3 kg/m2). Had hysterectomy. Women who had not had a hysterectomy were 1.60 (CI=1.36-1.88) times more likely to have had an annual Papanicolaou smear test than women who in the study sample who had had a hysterectomy. Cervical cancer screening advice. Women who had been advised by their primary care provider about cervical cancer screeni ng were 1.90 (CI=1.61-2.17) times more likely to have had an annual Papanicolaou smear te st than women who ha d not been advised by about cervical cancer screening by their primary care provider. Research Question 2 The second research question inquired whether there was a difference in the magnitude of the associa tion between ethnic origin and compliance with the 1999 cervical cancer screening guide lines of the ACS. The null hyp othesis posited that there was no difference in the magnitude of th e association betwee n ethnic origin and compliance with the 1999 cervical cancer screening guidelines of the ACS. The results of the analyses showed that Black American women were 1.83 times more likely than the referent group, White American women, to have had an annual Papanicolaou smear test as recomme nded by the 1999 guidelines of the ACS.

PAGE 128

117 Research Question 3 The third research question asked whet her there was an association between primary care provider advice about cervical cancer screening and compliance with the 1999 cervical cancer screening guidelines of the ACS. The null hypothesis posited that there was no association between primary care provider advice a bout cervical cancer screening and compliance with the 1999 cervical cancer screen ing guidelines of the ACS. The results of the analyses indicated th at in the study sample, the Black, White, and Hispanic American women who had been advised by their primary care provider about cervical cancer screening were 1.90 (C I=1.61-2.17) times more likely to have had an annual Papanicolaou smear test than women who had not b een advised by about cervical cancer screening by their primary care providers. Research Question 4 The fourth research question asked whet her the association between primary care provider advice about cervical cancer screening and complia nce with the 1999 cervical cancer screening guidelines of the ACS was moderated by ethnic origin. The null hypothesis posited that ethnic or igin did not moderate the association between primary care provider advice about cervical cancer screening and compliance with the 1999 cervical cancer screening guideli nes of the ACS. The results i ndicated that the association was not moderated by ethnic origin (p=0.8768, p=0.4162). In other words, there was not any difference in primary care provider advice and subsequent cervical cancer screening compliance among Black, White and Hispanic American women.

PAGE 129

118 CHAPTER FIVE: SUMMARY, DISCUSSION, CONCLUSIONS, RECOMMENDATIONS Summary Black American women have the highest age-adjusted mortality rate for cervical cancer among all ethnic groups with 5.6 cervical cancer deat hs per 100,000. For Hispanic American women the rate is 3.6 cervical cancer deaths per 100,000, and White American women 2.6 cervical cancer de aths per 100,000. The age-adjusted mortality rates for cervical cancer for all ethnic groups is 2.9 deaths per 100, 000 (Ries et al., 2004). The literature suggests that the num ber of deaths from cervical cancer in the United States could be reduced by preventive screening, a public health intervention strategy (CDC, 1998-1999; DHHS. Race and health: Cancer ma nagement, 1999; Franco, Duarte-Franco, & Ferenczy, 2001; Holmquist, 2000; Klaes et al. 2001; Koop, 1997; MMWR, 1997; Runowicz & Fields, 1999; Sasieni & Adam s, 1999; Schiffman, Brinton, Devessa & Fraumeni, 1996).This investigation was proposed on the premise that if cervical cancer screening rates could be increased in this population of women, particularly among Black American women, then the mortality ra tes would decrease (CDC, 1998-1999; DHHS. Race and health: Cancer management, 1999; Franco, Duarte-Franco, & Ferenczy, 2001; Holmquist, 2000; Klaes, et al. 2001; Koop, 1997; MMWR, 1997; Runowicz & Fields, 1999; Sasieni & Adams, 1999; Schiffman, Brinton, Devessa & Fraumeni, 1996).

PAGE 130

119 Therefore, the primary objective of this investigation was to study the association among the covariates of compliance with the 1999 cervical cancer scr eening guidelines of the ACS (cervical cancer screening behavior ) and to explore the relationship between primary care provider advice about cervical cancer screening and compliance with these screening guidelines in an ethnically di verse population of American women. The secondary purpose of this study was to exam ine the utility of th e Preventive Health Model as a theoretical model in guiding research in cervical cancer screening behavior in this ethnically diverse population of women, using survey data from the 1999 Behavioral Risk Factor Surveillance System. The findings of this study will be further discussed in terms of the research questions posed. Research Question 1 Is there an association between each of th e selected covariates of cervical cancer screening behavior and complia nce with the 1999 cervical can cer screening guidelines of the ACS in this ethnically divers e population of American women? Nine covariates of cervical cancer sc reening behavior were found to be significantly associated with compliance with the 1999 annual ce rvical screening guidelines of the ACS. The results indicated that Black, Hispanic and White American women of child-bearing age (18 to 44 years of age) who (a) had health care coverage, (b) had had a screening mammogram, (c) had had a clinical (physical) breast examination, (d) were non-smokers, (e) were not obese, (f) had not had a hysterectomy and (g) had been advised by their primary car e providers about cervical cancer screening were approximately twice as likely as to have been compliant with the 1999 cervical

PAGE 131

120 cancer screening guidelines of the ACS, as those women without these characteristics. Conversely, the women in the study sample who were older and not of child-bearing age (45 to 64 years of age) who (a) did not have health care coverage, (b) had not had a screening mammogram, (c) were smokers, (d) were obese (had a body mass index 27.3 kg/m2), (e) had had a hysterectomy, were less lik ely to have been compliant with the 1999 annual cervical cancer screening of the ACS. Research Question 2 Is there a difference in the magnitude of the association between ethnic origin and compliance with the 1999 cervical cancer screening guidelines of the ACS? Black American women were nearly twice as likely as White American women to have had an annual Papanicolaou smear test as recommended by the 1999 cervical screening guidelines of the ACS. Convers ely, White American women were 0.55 times less likely than the Black Am erican women to have had an annual Papanicolaou smear test. Although the effect of the Hispanic Am erican group was not found to be statistically significant, the information indicated that Hispanic American women had screening characteristics that were similar to thos e of the White American women. Therefore, Hispanic American women were less likely th an Black American women to have had an annual Papanicolaou smear test. In practical terms the results indicated that White American women of childbearing age (18 to 44 years of age) who had health care coverage, had had a screening mammogram, had had a clinical (physical) br east examination and had been advised by their primary care providers about cervical can cer screening were approximately half as likely as Black American women of the same status, to have had an annual Papanicolaou

PAGE 132

121 smear test. Hispanic American women were also more likely to have had similar screening characteristics as those of the Wh ite American women, and thus, they were less likely to have had an annual Papanicolaou smear test. The cervical cancer screening rate was defi ned as the percentage of women 18 to 64 years of age who were enrolled in a heal th plan and had had a Papanicolaou screening smear within a year of being interviewe d for the 1999 BRFSS (National Committee for Quality Assurance, 2002). The goal of Healthy People 2000 was for women in the nation to attain an average cervical cance r screening rate of 85% and for Healthy People 2010 the goal was 90% (Healthy People 2010, 2000; Healthy People 2000 Final Review, 2001) The average cervical cancer screening rates in 1999 for the women in the study sample were: (a) 80% for Black American women, (b) 71.2% for White American women, and (c) 76.0% for Hispanic American wo men. It can, therefore, be observed that even though Black American women were twi ce as likely as White American women to have been compliant with the 1999 annual ce rvical cancer screening guidelines of the ACS, both groups fell short of meeting the minimum Healthy People 2000 cervical cancer screening rate goal. Hispanic Ameri can women are also included in this suboptimal group of screeners because their cervical cancer screening rate is higher than that of White American women, but lower than the rate of Black American women. Research Question 3 Is there an association between primary care provider advice about cervical cancer screening and compliance with the 1999 cervical cancer scr eening guidelines of the ACS?

PAGE 133

122 One of the objectives of this study was to investigate the association between primary care provider advice about cervical cancer screening and compliance with the 1999 annual cervical cancer screeni ng guidelines of the ACS. Th e results showed that there was a statistically significant association betw een these two variables, namely that Black, Hispanic and White American women, who had been advised by their primary care provider about cervical cancer screening, were about twice as likely to have had an annual Papanicolaou smear test as women who had not been advised about cervical cancer screening by their primary care providers. Research Question 4 Is the association between primary care provider advice about cervical cancer screening and compliance with the 1999 cervical cancer scre ening guidelines of the ACS moderated by ethnic origin? It also was found that ethnic origin di d not moderate the association between primary care provider advice about cervical cancer screening and compliance with the 1999 cervical cancer screening guidelines of the ACS. In other words, primary care provider advice about cervical cancer screeni ng did not differentially impact the Black, White or Hispanic American women in any statistically significant manner. The responses to research questions 3 and 4 were extremely important because they emphasized that primary care provider ad vice about cervical cancer screening was equally important for comp liance with the 1999 annual ce rvical cancer screening guidelines of the ACS among Black American, Hispanic American and White American. The implication of the responses to these quest ions will be discussed in the next section of the chapter.

PAGE 134

123 Discussion Implications for Public Health Intervention Strategies Intervention strategies need to be developed that are spec ifically tailored to older women who are no longer of child -bearing age. To develop thes e strategies it is important to understand the reasons fo r underutilization of cervical cancer screening in this population of women. Calle et al (1993) suggest that the de crease in regular cervical cancer screening among older women could be du e to a decrease in regular gynecological examinations among these women. Wells and Horm (1997) report that when the reproductive needs and care of women about 45 to 59 years of age change, they are less likely to obtain Papanicolaou screening tests. The reason for this dec line in frequency or regularity is that the women seek less of, or a different kind of care than they did during their reproductive years. The authors suggest that it is imperative to address this population of women with tailored messages. Amonkar and Madhavan (2002) imply that the decrease in cervical cancer screening among older women is due to the less stringent recommendations made by primary health providers once three or more annual Papanicolaou smears have been normal. Gulitz, Bustillo-Hernandez and Kent (1998) report that lack of primary care provider recommendation is a major predictor of underutilization of cervical can cer screening in this populat ion of women. Mandelblatt and Yabroff (2000) also report that physic ians do not consistently recommend cervical cancer screening to olde r women (not of child-bearing age) and physician recommendation is one of the strongest pred ictors of screening. Because older women participate less regularly in cervical cancer screening than younger women, the cervical lesions, when discovered, may be at more advanced stage and less responsive to

PAGE 135

124 treatment than those of younger women (Wh ite, Begg, Fishman, Guthrie, & Fagan, 1993; Masood, 1997). It can be seen from the lite rature that the Black American, Hispanic American and White American women who are 45 to 64 years of age, ar e at a higher risk of developing cervical cancer than their younge r counterparts, in part, because of their poorer compliance with cervical cancer screening guidelines. Primary care provider recommendation has an important impact on compliance with cervical cancer scre ening guidelines. It is, therefor e, imperative for primary care providers to encourage compliance with scre ening, guidelines, particularly women who are beyond child-bearing age. Primary care pr oviders could discuss the rationale of cervical cancer screening in older women especia lly if the women still have intact uteri. It is also important for women to know about their personal screeni ng options, for example whether she should continue to be screened on an annual basis, less frequently or not at all. It would also be helpfu l to provide the patients with information about scheduling a Papanicolaou test. These women could then be given literature to take home with them, with the assurance that they could call the pr imary care provider’s office if they had any further questions about cervical cancer screen ing. These women coul d be sent a reminder close to the time of time of their next examination. If the women do not return for their next visit, they should be contacted. Vogt, Glass, Glasgow La Chance and Lichtenstein (2003) concluded the best way to follow-up on these patients is to send them a second reminder card or letter. If there was still no response, then a personal telephone call would elicit the best resp onse from the patient. Another intervention strategy to encourage older women to increase their cervical cancer screening rates would be to provide out reach and integrated preventive services at

PAGE 136

125 community-based sites, as opposed to the o ffices of primary care providers. Educational programs about the importance of cervical cancer screening in older women and Papanicolaou smear tests could be offere d on-site in housing complexes (White, Begg, Fishman, Guthrie & Fagan, 1993). Mobile clin ical units could be used to provide identical services and encourage older women to participate in community-based cervical cancer screening programs (Masood, 1997). Utilizing White Americans as the referent group in public health research is standard procedure, and there are a number of reasons for this practice. White Americans usually comprise the largest ethni c group in a multi-ethnic study, SAS automatically defaults to the largest group as the referent group, researchers are taught to utilize the largest group as the referent group, and Wh ite Americans often have better health outcomes than the other ethnic groups, therefor e, it is reasonable to make this group the base to which all the others are compare d. In this study White American women were selected as the referent group because thei r screening outcomes (as opposed to screening rates) are better than the Black American and Hispanic American women, i.e. their mortality rates are lower than in either of the other two groups. However, the screening rates of White American women (71.2%) are far from optimal, so the implications of the higher cervical cancer screen ing rates of the Black Amer ican women (80.0%) in this sample need to be interpreted with cauti on because their screeni ng rates are also not optimal. Therefore, in terms of reality and practicality, it does not matter that the Black American women have higher cervical cance r screening rates; both groups of women need active public health interv ention strategies to increase their screening rates. The

PAGE 137

126 same argument can be applie d to the Hispanic American women who also have a suboptimal cervical cancer screening rate (76.0%). However, a way to improve the rates would be to develop community-based participatory networks to raise the awarene ss of the benefits and importance of regular cervical cancer screen ing throughout a woman’s lifetime. Community-based participatory networks are crucial for the implementation and success of any public health intervention to improve annual cancer screening rates. In terventions work best when the community itself is vested in the objectiv es and goals of the public heal th intervention. The networks often consist of equal partnerships among community members, community-based organizations, academic institutions and health agencies working together to address the health issues of the community. For an in tervention to be implemented effectively, community members themselves have to participate in designing, developing, implementing and evaluating the programs. It is also essential for th e community partners to ensure that the messages are crafted and delivered in such a manner as to encourage cervical cancer awareness among the women so that they understand: (a) the need for women of all ages to be screened regularl y for cervical cancer throughout their lifetime, (b) women’s reproductive health, especially the location of th e cervix in relationship to the uterus, (c) the significance of being as ymptomatic for cervical cancer, and the reasoning behind cervical cance r screening, and screening in general, (d) how a Papanicolaou test is performed, and the mean ing of the test results (e) the stages of cervical cancer and, (f) the importance of compliance with cervical cancer screening guidelines.

PAGE 138

127 In this population of women the message that needs to be emphasized is that cervical cancer is can be prevented, treated and cured if detected early. Cervical cancer prevention and early detection of cervical can cer can be enhanced by annual screening as recommended by the ACS (Masood, 1997). There still remains the disturbing fact that although Black American women do have higher screening rates th an White American women, their mortality rates remain higher (Makuc, Fried & Kl einman, 1989; Martin, Parker, Wingo & Heath, 1996; CDC/NCCDPHP. BRFSS: Prevalence data, 2002 ). Given the cross-sectional design of the study, it is not possible to make inferen ces about which women, if any, died from cervical cancer. However, it is known that Bl ack American women present with a more advanced stage of cervical cancer at di agnosis, an occurrence thought to be a consequence of underutilization of cancer sc reening services in this group (Shavers & Brown, 2002). Factors which have been reported to be associated with disproportionate cervical cancer mortality among Black American women, resulting in underutilization of cervical cancer screening services include: (a) lack of knowledge about cer vical cancer and the implications of abnormal Papanicolaou smear results, (b) lack of adequate follow-up of abnormal Papanicolaou smears, (c) lack of co mmunity participation in helping women to navigate the health system, and (d) lack of community involvement in discussions about the serious repercussions of not followi ng-up on abnormal Papanicolaou smear test results, especially for Black American women (Allen-Barash, Foster, Mitchell & Fletcher, 1997; Bigby, Ko, Johnson, David, & Fe rrer, 2003). It is imperative that the gravity of underutilization is emphasized in discussions among the community-based

PAGE 139

128 partners, during the process of designing a nd developing culturall y, and linguistically appropriate cervical cancer educational messages for women in the community. Such discussions could also help to hi ghlight previously unknown reasons for the differences in cervical cancer mortality among Black American, White American and Hispanic American women in the study sample that in turn, could be attributed to the processes involved in obtaining an annual Papanicolaou smear test. Evaluation of the Preventive Health Model A secondary objective of this study was to assess the utility of the Preventive Health Model. This theoretical model was de veloped by researchers in the field of cancer prevention and control specifically to understand preventive health behavior (Myers et al. 1994; Myers et al. 1996). The Preventive H ealth Model was evaluated for its utility according to the criteria of theory evaluation suggested by Tzeng and Jackson (1991). These researchers generated six criteria from the literature: (a) formalization, (b) flexibility, (c) parsimony, (d) falsifiability, (e) fruitfulne ss, and (f) scie ntific selfregulation. “Formalization is the extent to which a theo ry, or theoretical m odel is clear and its statements and variables are explicitly defi ned and consistently used” (Tzeng & Jackson, 1991, p.72). The Preventive Health Model incor porates all of these criteria. It is a theoretical model that has been derived from constructs of the Health Belief Model, the Theory of Reasoned Action and Social Cognitive Theory that are thought to be important in predicting preventive hea lth behavior, such as cancer sc reening (Carver & Scheier as cited in Myers, et al., 1994).

PAGE 140

129 “A good theory, or theoretical model needs to remain flexible enough to accommodate new evidence, and thus, can neve r become completely closed ” (Tzeng & Jackson, 1991, p. 62). The Preventive Health Mode l is flexible as seen by the fact that investigators have used different combinati on of constructs from the Preventive Health Model in their various research projects. Myers et al. (1994) added the construct of preventive intention to the model to see wh ether the degree of inte ntion to perform or engage in preventive health behavior increa sed the predictive power of the model. Myers et al. (1996) and Gwede (2001) did not collect any data on pr ogram factors, therefore the construct of program factors was removed from their models. Finally, Watts, Vernon, Myers and Tilley (2003) used al l the constructs of the Preven tive Health Model in their study of colorectal cancer screeni ng among male automotive workers. “Parsimony is the extent to which a theo ry, or theoretical model attempts to account for complex phenomena in terms of parsimonious constructs” (Tzeng & Jackson, 1991, p. 72). The most useful theory is one that can generate accurate predictions with the fewest prior assumptions and the simplest pr opositions. Different constructs of the model have been used by various researchers to determine covariates of preventive health behavioral intent, or the be havior itself, using a few pr ior assumptions and simple propositions. The Preventive He alth Model was developed sp ecifically to understand and determine the covariates of preventive h ealth behavior (Myers, Ross, Jepson, Wolf, Balshem, Millner, & Leventhal, 1994). “A good theory or theoretical model must be amenable to operational definitions” (Tzeng & Jackson, 1991, p. 62,). Operational de finitions are the processes by which constructs or variables are defined in term s of the methods, proce dures and techniques

PAGE 141

130 used to explain them. These definitions ar e important in evaluation of a theory or theoretical model because th ey provide a uniform standard by which they can be objectively tested and proved wrong (falsifi ed). The Preventive Health Model is amenable to operational definitions and its constructs were operationalized by variables from the 1999 BRFSS questionnaire. It also has been tested objectively by different researchers in the field of preventive health behavior. “Fruitfulness is the extent to which a theory or theoretical model stimulates further research” (Tzeng & Jackson, 1991, p. 72). The Preventive Health model has been used in studies to explain: (a) colorectal cancer screening behavior in men and women (Myers et al., 1994, Watts, Vernon, Myers & Tilley, 2003), and (b) prostate cancer screening behavior in African American men (Myers, Wolf, Mckee, McGory, Burgh, Nelson, & Nelson 1996; Gwede, 2001). The theoretical model is being used in this study to explain the relationship between primar y care provider advice about cervical cancer and compliance with cervical cancer screening guidelines. “A good theory should keep itself ‘in-ch eck’ by utilizing scie ntific techniques designed to optimize empirical objectivity and theoretical objectivity (Tzeng & Jackson, 1991, p. 64). The best method of scientific self -regulation is to have research articles published in scholarly, peer-rev iewed journals. In this way the research methods and results using the Preventive Health Mode l have been added to the existing body of knowledge by the investigators. Variables from the 1999 BRFSS were use d, in this study, to measure the constructs of the Preventive Health Model. A number of the socio-demographic variables that were used to measure the background fact ors construct were dropped from the final

PAGE 142

131 logistic regression model because they di d not gain statistical significance. These variables were marital status, education level, employment status, income level and state FIPS codes. The representation factor construct and the social influence factor construct were each measured by a single vari able. These two elements were convenience of medical facility location and satisfaction with primary care provider. There was not enough data collected on these va riables for them to be included in the final logistic regression model. The Preventive Health Model was modified accordingly, and the results of the study are displayed in Table 12. Table 12 The Modified Preventive Health Model Constructs of the PHM Variables from 1999 BRFSS References Background Factors Construct 1. Socio-demographic variables Age Ethnic origin Insurance coverage 2. Past screening behavior variables Mammography Breast exam 3. Health profile variables Tobacco use Obesity (BMI) Hysterectomy Program Factor Construct Screening advice Child-bearing age Ethnic origin Health coverage Had mammogram Had breast exam Smoking status Body Mass Index Had hysterectomy Cervical cancer screening advice Hiatt, et al., 2002 NCI, 2004 Hiatt, et al., 2002 Simoes, et al., 1999 Simoes, et al., 1999 Simoes, et al., 1999 Simoes, et al., 1999 Smith, et al., 2003 Baranowski, et al., 1997 It was found that socio-demographic variab les, past screening behavior variables and health profile variables that measured the background factors construct were highly associated with compliance to the cervical can cer screening guidelines. The variables that

PAGE 143

132 measured the representation factor construct and the social influence factor structure could not be assessed in this study because of lack of data. Therefore, these constructs were not included in the modified Pr eventive Health Model. The variable cervical cancer screening advice was used to measure the program fact or construct and also was found to be associated strongly with compliance to cervical cancer screening guidelines. The Preventive Health Model has been utilized effectively to guide the research on factors associated with compliance to these cervical cancer screening guidelines as evaluated by the criteria suggested by T zeng and Jackson (1991). Conclusions Limitations of the Study The major limitation of this study is that the study design was a cross-sectional or prevalence study in which the status of the exposure (primary care provider advice about cervical cancer screenin g) and the attribute of interest (compliance with the 1999 cervical cancer screening guidelines of the ACS) were assessed simultaneously among individuals in the study sample. Therefore, it was not possible to determine whether primary care provider advice about cervical cancer screeni ng resulted from, or preceded, compliance with screening guidelines. This limitation mu st be taken into acc ount when interpreting the data because temporal and causal inferenc es cannot be determine d. It was possible to conclude that there was an associatio n between compliance with cervical cancer screening guidelines and primary care provi der advice about cervical cancer screening when all the other variables were accounted for (adjusted) in the statistical logistic regression model and theoretical Preventive Hea lth Model. The data produced may be of

PAGE 144

133 great value to public health professionals in assessing the health care needs of an ethnically diverse population of wo men (Hennekens & Buring, 1987). These results are only generalizable to the population of women who identified themselves as being: (a) Black (non-Hisp anic), White (non-Hispanic) and Hispanic (Black and White) American women, 18 to 64 y ears of age who (b) re ported that they had received a preventive Papanicolaou smear test (c) resided in the states of Louisiana, Missouri, South Carolina, Texas, Virginia and Wyoming in 1999, and (d) responded to the question “During the past year, have you received advice from your primary care provider about your sexual practices and sexua lly transmitted diseases?” [This question is a proxy for the question “Has your primary care provider ever advised you about cervical cancer screening?”]. However, the methods us ed to obtain these results can be adapted for use in other studies. Contribution of the Study to the Body of Knowledge There are a number of contributions this study has made to the body of knowledge. The Preventive Health Model has not been used to determine preventive cervical cancer screening behavi or in women. It has been used in studies to explain: (a) colorectal cancer screening behavior in men and women (Myers et al., 1994, Watts, Vernon, Myers & Tilley, 2003), and (b) prostate cancer screening be havior in African American men (Myers, Wolf, Mckee, Mc Gory, Burgh, Nelson, & Nelson 1996; Gwede, 2001). This study has shown the “fruitfulness “ of the Preventive Health Model by stimulating further research in a different ar ea of preventive health behavior than had been studied previously. The background f actors construct was extended to encompass the health profile variables. These variables had not been us ed by any other researcher to

PAGE 145

134 determine preventive health behavior. This ach ievement indicates the flexibility of the model as, an important factor in preventive health behavior research, because there are many different combinations of variables th at can affect health behavior. Another contribution to the body of knowledge is that this is the only study in which the Preventive Health Model has been used with secondary data from a national data set to answer a series of research questions on preventive health behavior. It showed how well this theoretical model could be used to guide research. Finally, this is the only study in which “intention to perform a preventive hea lth behavior” was not an intermediate or final outcome. The Preventive Health Model was used with the 1999 BRFSS data set to answer the specific research questions, because the data set contained questions pertaining to (self-reported) health behavior rather than to (self-repo rted) knowledge, attitudes, and perceptions of, or intentions to perform health behavior. Moreover, this data set contains survey que stions specific to preventive screening behavior, including cervical cancer screenin g behavior (CDC/NCCDPHP. About the BRFSS, 2002; Coughlin, Uhler, Hall & Briss, 2004). This study also adds dimensions to th e body of knowledge related to disparities research. It results revealed that Blac k American, White American and Hispanic American women were more similar in th eir screening behavior than dissimilar. Therefore, when thinking about cost-effective, public health interv ention strategies, it would be better to concentrate on the sim ilarities across the groups first and design intervention strategies to addr ess as many of the women as possible. The next step would be to look at the dissimilari ties across ethnic origins, a nd where appropriate, tailor extremely specific interventions to the popul ation segments in question. The study also

PAGE 146

135 showed that caution should be used in interpreting results when using the White American women as the base or referent group. It is imperative not to assume that their outcomes are better than those of other wo men. The data must be checked carefully before drawing that conclusion. Finally, this study contributes to the literature by showing that the disparity among the women is in age, women of child-bearing age (1844 years) versus women not of child-bearing ag e (45 to 64 years), rath er than in ethnic origin. There are few data in the literature demonstrating an associ ation between primary care provider recommendation a nd compliance with cervical sc reening guidelines. These findings contribute to the body of knowledge, namely that compliance with cervical cancer screening guidelines is associated with specific primar y care provider advice about this screening. The results of this study would enable public health researchers to develop and design intervention programs that could be tailored to specifie d population segments (DHSS. Race and health: cancer management, 1999; Glanz & Rimer, 1997). Such interventions may increase cervical screening to more optimal levels among the diverse groups of women in the nation, thereby, not only reducing the differences in mortality rate, but also reducing the overall mortality rate from cervical cancer. HPV DNA Screening, HPV Vaccine Trials and Public Health Intervention Strategies In 2005 the ACS included HPV DNA r ecommendations for cervical cancer screening. “Screening should be done every year with conventiona l Papanicolaou smear tests or every two years using the liquid-based tests. At or after 30 years of age, women who have had three consecutive, normal results may elect to be screened every two to three years. Alternatively, cervical cancer screening with HPV DNA testing and or

PAGE 147

136 liquid-based cytology could be performed every three years. Physicians may suggest that a woman screen more often if she has certain risk factors, such as HIV infection or a compromised immune system” (ACS, 2005, p. 60). As previously stated in Chapter 2, the most likely causative agent of cervical cancer and its precursors is the human papi llomavirus (HPV) (Cuzick, 2000; Hoffman & Cavanagh, 1996; Palefsky, 2003; Rohan, Burk, & Fr anco, 2003). Persistent infection with one or more types of HPV is an important etiological factor in the development of precancerous lesions with progression to invasive cervical cancer. Worldwide, HPV DNA is detected in about 99.7% of all inva sive cancers (Sellors et al., 2003). The incidence of carcinogenic HPV is highest in younger women aged 15 to 20 years of age, however, the infection clears spontaneously in 60% of this population (Cohen, 2005; Gray & Walzer, 2004; Sellors et al., 2003). In the United States, the prevalence of HPV infection declines to very low levels by th e time a woman reaches 50 years of age, and identification of high-risk HPV genotypes, by the hybrid capture tec hnique, is extremely rare in post-menopausal women. Therefore a ny HPV infection found in older women is most likely to be due to persistent inf ection with high-risk HPV genotypes (Ferenczy, Gelfand, Franco & Mansour, 1997; Schiffman & Ca stle, 2003). It is al so known that: (a) HIV-positive women have a higher inciden ce of cervical HPV infection than HIVnegative women, (b) cervical HPV infection is more persistent in the HIV-positive population, (c) the incidence of invasive cer vical cancer is increased in HIV-positive women, and (d) HIV-positive women with lesions have higher HPV viral loads than HIV-negative women (Palefsky, 2003; Schiffman, et al., 2000).

PAGE 148

137 Several researchers have examined the role of using HPV DNA tests as the primary (preventive) tool in screening fo r invasive cervical ca ncer (Cuzick, 1995). However, there are a number of issues that need to be resolved before this can be realized. One of the major issues in using HPV DNA testing as the primary screening test for cervical cancer is that its sp ecificity is presently worse th an that of cytology. That is, the probability of the HPV DNA screening test correctly identifying a woman who does not have cervical cancer (the true negative rate) is worse th an that of the conventional Papanicolaou smear screening test (Rohan, Bur k, & Franco, 2003). Other issues that need to be determined include: (a) the most appr opriate ages at which to begin and end HPV DNA testing, (b) the frequency of screening, (c) the utility of incor porating measures of high HPV viral load in lesion management, and (d) the cost-effectiveness of this approach to screening. Large-scale, randomize d, controlled clinical tr ials are needed to evaluate: (a) whether widespread HPV DNA test ing is either feasible or affordable, (b) whether such testing will eventually lead to fewer cases of invasive cervical cancer, or reduce the mortality of the disease (Cuzick, 2000; Schiffman, et al., 2000; Rohan, Burk, & Franco, 2003). Cytology will continue to be the major screening method for cervical cancer prevention until these issues with HPV DNA screening have been resolved (Schiffman, et al., 2000). It has been recommended that HP V DNA testing be used as an adjunct to routine cyto logical screening, especially as it has a higher sensitivity than cytology. That is, the probability of the HPV DNA screening test correctly identifying a woman who has invasive cervical ca ncer (the true positive rate) is better than that of the conventional Papanicolaou smear screening te st (Cuzick, 1995; Rohan, Burk, & Franco, 2003; Tjalma, Arbyn, Paavonen, Van Waes & Bogers, 2004). Even though addition of

PAGE 149

138 the HPV DNA test to the Papanicolaou smear test is thought to increase the sensitivity for detecting invasive cervical cancer, and allows for larger screening intervals (Franco, 2003; ACS, 2005), clinical trials have not shown any reducti on of cancer incidence in populations where HPV DNA testi ng was added to cytological screening (Hinkula et al., 2004). Ongoing trials need to be extended in terms of size and duration, in order to examine ways in which HPV DNA testing could be used as a primary (preventive) tool in screening for invasive cervical cancer (T jalma, Arbyn, Paavonen, Van Waes & Bogers, 2004). The cytological screening programs that have helped to reduce the incidence rates of invasive cervical cancer in developed countri es in the world, are either not available or are too expensive for women in developing coun tries. Researchers believe that women in the developing world could benefit the most from vaccination programs which target invasive cervical cancer (Cohen, 2005; Sc hreckenberger, & Kaufmann, 2004; Tjalma, Arbyn, Paavonen, Van Waes & Bogers, 2004). Ho wever, women in the United States who are not screened, or who are underscreened could also benefit from such vaccination programs. Two types of HPV vaccines can be differe ntiated. These are: (a) therapeutic vaccines which induce cell mediated immune re sponses against epithelial cells infected with HPV, and (b) prophylactic vaccines which prevent primary HPV infection by inducing virus-neutralizing antibodies which protect against new, but not established infections (Schreckenberger, 2004; Tjal ma, Arbyn, Paavonen, Van Waes & Bogers, 2004). Preclinical and clinical studies have shown that therapeutic vaccines need more appraisal because the problem of lack of clin ical responses due to tumor immune evasion

PAGE 150

139 (Schreckenberger, 2004). Stanley, 2003 suggests that therapeutic HPV vaccines are likely to only clear the cervical cancer completely in a tiny proportion of the cases whether the vaccine is used alone or with immunomodulators. The re searcher suggests that the vaccines may be used as adjunct therapy to pr event the reoccurrence of invasive cervical cancer, or HPV infection, after the cancerous lesions have been surgically removed. Two different prophylactic vacci nes have been developed an d tested in more than 3000 human participants in phase II clinical trials. These vaccines were made by Merck & Company of Rahway, New Jersey, and GlaxoSmithKline (GSK) Biologicals of Rixenart, Belgium. Both vaccines prevented persistent infection in 100% of the women vaccinated against the disease, and redu ced cervical abnormalities by 90% (Cohen, 2005; Harper, et al., 2004). However, these vaccines still have to be proven to be safe and effective in larger phase III clinical trials wi th long term follow-up, and also confirm that vaccination does prevent cervical cancer. The vaccines are currently under study in over 50,000 human participants the United States, and countries of South America, Europe, Asia and Africa, and the trials are proj ected to end between 2007 and 2010 (Cohen, 2005; Tjalma, Arbyn, Paavonen, Van Waes & Bogers, 2004). There are a number of issues that need to be resolved before HPV vaccination could possibly be implemented as an effec tive public health measure for reducing the global burden of cervical can cer (Harper, et al., 2004; Tjalma, Arbyn, Paavonen, Van Waes & Bogers, 2004). The first issue is the number of HPV genotype s that should be included in the vaccines to target cervical cancer. Merck and GSK used HPV 16 and 18 as the foundation of their vaccines, but these HPV genotypes together only give protection against 70% of the cervical cancer cases (C ohen, 2005). It is not known

PAGE 151

140 whether antibodies created for one genotype of HPV will protect against other genotypes, which indicates the need for multivalent vaccines. HPV genotypes 16, 18, 45, 31 and 33 are responsible for 85% of all cervical cancer. Therefore, in order to be effective for at least 80% of the population, the v accines, in theory, should contai n at least four to five of the most common oncogenic HPV genotypes of that region or country, because oncogenic HPV genotypes are region and country specific. However, the combination of several HPV genotypes in one vaccine could (a) cause other unforeseen problems, (b) create more manufacturing difficulties and (c) increase the cost of production of the vaccine thereby increasing its price (Cohen, 2005; Schiffma n & Castle, 2003; Tjalma, Arbyn, Paavonen, Van Waes & Bogers, 2004). Other important issues that need to be resolved are questions regarding the duration of the vaccine-induced immunity, a nd the minimum protectiv e level of the HPV antibodies. These have not been evaluated a nd follow-up studies are needed to establish the facts. It is known that antibodies acqui red after a natural infection with HPV will persist for decades, but decrease over time in women who do not have HPV-associated lesions (Carter, et al., 2001). For the vaccine to be efficacious it would need to confer immunity for several decades, conceivably, with a booster vaccinati on after 5 or 10 years (Tjalma, Arbyn, Paavonen, Van Waes & Bogers, 2004). Issues involving the optimal timing for HPV vaccination and which groups to vaccinate need to be addressed. Until the duration of the vaccine-induced immunity is known, it will be difficult to determine the appropriate age groups that would most rapidly prevent the spread of HPV infec tion in the population. Evidence from the preclinical and clinical studies suggests that the prophylactic HPV vaccine should be

PAGE 152

141 administered to young men and women before they become sexually active. However, it is not known whether men will be protected against HPV infection or HPV-induced clinical disease resulting from infection. Individuals who are immunocompromised are at a higher risk of persistent HPV infection a nd the development of HPV-related diseases, including cervical cancer. It is not known whether a prophylactic HPV vaccine would be of benefit to these i ndividuals. It is imperative that st udies be initiated to determine exactly which individuals in the population would benefit fr om being vaccinated against HPV (Cohen, 2005; Harper, et al., 2004; Tj alma, Arbyn, Paavonen, Van Waes & Bogers, 2004). Until these issues with HPV DNA screening and HPV vaccination are resolved, women should abide by the cervical cancer screening guidelines recommended to them by their primary care providers. With the confus ion that could arise from reports in the media about HPV DNA testing and HPV vaccination, a clear and concise message that all women should be screened regularly througho ut their life, is even more relevant and crucial than ever before. “When and even if the cervical cancer screening can be stopped is unclear, it seems that cervi cal cancer screening should be continued for at least a whole generation of women who are already infected” (Tjalma, Arbyn, Paavonen, Van Waes & Bogers, 2004, p. 758). Recommendations for Future Research The cross-sectional design of this study limited the inferences that could be drawn from the results. Neither the temporal rela tionship, nor the causal relationship between the dependent variable (compliance with the annual 1999 cervi cal cancer screening guidelines of the ACS) and the focal independe nt variable (primary care provider advice

PAGE 153

142 about cervical cancer screen ing) could be assessed. Th e cross-sectional design only allowed for associations to be drawn between each one of the independent variables and the dependent variable in the study. Pattern s of screening behavior could not be determined because the women were asked ab out their screening practices at a single point in time. Thus, it was not possible to tell whether a woman had a pattern of being screened annually, or whether she had just been screened at the point in time when she was asked the question “How long has it b een since you had your last Pap smear?” A similar study should be performed usi ng a longitudinal, prospective cohort study design. This would improve the capacity to establish a temporal sequence and causality. In this study the women would be follo wed for at least ten to fifteen years, and questions from the BRFSS would be util ized. The BRFSS questionnaires contain questions which have been designed to coll ect data on general se lf-reported health behavior. The data are then used for pl anning, implementing, managing and evaluating health promotion and disease prevention progra ms in the United States and its territories. Factors assessed by the BRFSS include tobacco use, general health status, health coverage, and the use of cancer screening services (CDC/NCCDPHP. About the BRFSS, 2002). In order to adapt the questions for us e with the more specific research methods dictated by the study questions, the questions fr om the BRFSS need to be modified in the following manner. A question concerning prim ary care provider advice about cervical cancer screening should be de veloped. For example, women should be asked specifically about how often their primary care provider advi sed them to screen for cervical cancer. This question should be put in to the Women’s Health core section and be asked of women in all of the reporting sites.

PAGE 154

143 Two other questions should be added to th e Women’s Health core section, one of them should be a measure of health literacy and the other should be a measure of acculturation. Health literacy is “the abil ity to read and unders tand written materials commonly encountered in health care setti ngs” (Scott, Gazmararian, Williams & Baker, 2002, p.395). Acculturation refers to “the extent to (and the process through) which … minorities participate in the cultural traditi ons, values, beliefs, assumptions, and practices of the dominant … society (acculturated), remain immersed in their own cultures (traditional), or participate in the traditions of their own culture and of the dominant … culture as well (bicultural) ” (Landrine & Klonoff, 1994, p. 104) Low acculturation and low health literacy are thought to contribute to lower rates of cervical cancer screening, so it is important to control for these va riables in the study (Selden, Zorn, Ratzan & Parker, 2000). There is not a question asked on the specifi c type of primary care provider visited by the women (e.g. family practitioner, inte rnist etc.) in the BR FSS questionnaire. To better assess the association between comp liance to annual cervical cancer screening guidelines and primary care provider advice about cervical cancer screening, a question needs to be included on the type of primar y care provider visited by the women in the study. Fos and Zungia (1999) defined primary care providers as being: (a) family practice physicians, (b) general practice physicians (c) obstetrics-gynecology physicians, (d) internal medicine physicia ns, (e) pediatric physicians, and (f) nurse practitioners. A final modification to the BRFSS questionnair e would be to include a response for question refusal, which would contribute to the design of the questionnaire for further surveys.

PAGE 155

144 The BRFSS is a computer-assisted, tele phone interview survey. Therefore, by definition, women without telephones were ex cluded from participating in the study. Women without telephones and without health plans coul d be enrolled into the prospective study by using supplementary que stionnaires and enrollment techniques designed and approved by members of the co mmunity-based participatory networks. Doing so would allow results from the study to be more generalizable than those of the current cross-sectional study. A longitudinal, prospectiv e cohort study design would allow the temporal relationship, and th e causal relationship between the focal independent variable (primary care provider advice about cervical cancer screening) and the dependent variable (compliance with the annual 1999 cervi cal cancer screening guidelines of the ACS) to be assessed.

PAGE 156

145 REFERENCES Adami, H. O., Ponten, J., Sparen, P., Bergstrom, R., Gustafsson, L. & Friberg, L. G. (1994). Survival trend after invasive cervical cancer di agnosis in Sweden before and after cytologi cal screening. 1960-1984. Cancer, 73 (1), 140-7. Adams, M., Borysiewicz, L., Fiander, A., Man, S., Jasani, B., Navabi, H., Lipetz, C., Evans, A. S., & Mason, M. (2001). Can cer studies of human papillomavirus vaccines in pre-invasive and invasive cancer. Vaccine 19, 2549-2556. Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior and Human Decision Processes, 50, 179-211. Ajzen, I. & Fishbein, M. (1980). Understanding attitudes and pr edicting social behavior. Englewood Cliffs, NJ: Prentice Hall. Ajzen, I. & Madden, T. (1986). Predictors of goa l-directed behavior: Attitudes, intention, and perceived behavioral control. Journal of Experiential Social Psychology, 22, 453-474. Alle-Barash, N., Foster, V., Mitchell, W. H., & Fletcher, R. (1997). Breast and cervical cancer screening. Washington Public Health 15, 1-6. Allison, P. D. (1999). Logistic regression using the SAS system: Theory and application Cary, NC: SAS Institute Inc. Allison, P. D. (2001). Missing data. Sage University Papers on Quantitative Applications in the Social Sciences, 07-136. Thousand Oaks, CA: Sage.

PAGE 157

146 American Cancer Society, Inc. (1999). Cancer facts & figures 1999. American Cancer Society, Inc. Atlanta, GA. American Cancer Society, Inc. (2000). Cancer facts & figures 2000. American Cancer Society, Inc. Atlanta, GA. American Cancer Society, Inc. (2001). Cancer facts & figures 2001. American Cancer Society, Inc. Atlanta, GA. American Cancer Society, Inc. (2002). Cancer facts & figures 2002. American Cancer Society, Inc. Atlanta, GA. American Cancer Society, Inc. (2003). Cancer facts & figures 2003. American Cancer Society, Inc. Atlanta, GA. American Cancer Society, Inc. (2004). Cancer facts & figures 2004. American Cancer Society, Inc. Atlanta, GA. American Cancer Society, Inc. (2005). Cancer facts & figures 2005. American Cancer Society, Inc. Atlanta, GA. Amonkar, M. M. & Madhaven S. (2002). Comp liance rates and predictors of cancer screening recommendations among Appalachian women. Journal for Health Care for the Poor & Underserved, 13 (4), 443-60. Ayre, J. E. (1947). Selective cytolo gy smear for diagnosis of cancer. American Journal of Obstetrics and Gynecology, 53, 609-617. Bandura, A. (1986). Social foundations of thought and acti on: A social cognitive theory. Englewood Cliffs, NJ: Prentice Hall. Baranowski, T., Perry, C. L. & Parcel G. S. (1997). How individuals, environments, and health behavior interact: Social Cognitive Theory. In Glanz, K., Lewis, F. M., &

PAGE 158

147 B. K. Rimer, (Eds.). Health behavior and health education: Theory, research and practice (2nd ed.) (pp. 153-178). San Francisco: Jossey-Bass. Bigby, J., Ko, L. A., Johnson, N., David, M. M. A., & Ferrer, B. (2003). A community approach to addressing ex cess breast and cervical can cer mortality among women of African descent in Boston. Public Health Reports, 118 (4), 338-47. Bjorge, T., Trope, C. G., & Engelan d, A. (1999). Screening for cancer. Tidsskrift for Den Norske Laegeforening, 119 (8), 1129-36. Blair, J. & Czaja, R. (1982). Locating a special population using random digit dialing. Public Opinion Quarterly 46, 585-590. Bocciolone, L., La Vecchia, C., Levi, F., Lucchin i, F. & Franceschi, S. (1993). Trends in uterine cancer mortality in the Americas, 1955-1988. Gynecologic Oncology, 51 (3), 335-44. Brenda, J. A. (1994). Pathology of cervica l carcinoma and its prognostic implications. Seminars in Oncology, 21 (1), 3-11. Brinton, L. A., & Fraumeni, J. F., Jr. (1986) Epidemiology of uterine cervical cancer. Journal of Chronic Diseases, 39 (12), 1051-1065. Brinton, L. A., Herrero, R., Reeves, W. C., de Britton, R. C., Gaitan, E. & Tenorio, F. (1993). Risk factors for ce rvical cancer by histology. Gynecologic Oncology, 51 (3), 301-6. Broders, A. C. (1932). Carcinoma in situ cont rasted with benign pe netrating epithelium. Journal of the American Medical Association, 99, 1670-4.

PAGE 159

148 Bosch, F. X. & de Sanjose, S. (2003). Human papillomavirus and cervical cancer – burden and assessment of causality. Journal of the National Cancer Institute. Monographs, 31, 3-13. Bronson, R. C., Jackson-Thompson, J., Wilkerso n, J. C. & Kiang, F. (1994). Reliability of information on chronic disease risk fact ors collected in the Missouri Behavioral Risk Factor Surveillance System. Epidemiology, 5 (5), 545-9. Buckley, C. H., Butler, E. B., & Fox, H. ( 1982). Cervical intraep ithelial neoclassic. Journal of Clinical Pathology, 35, 1-13. Called, E. E., Flanders, W. D., Thune, M. J. & Martin, L. M. (1993). Demographic predictors of mammography and Pap smear screening in U.S. women. American Journal of Public Health, 83, 53-60. Carpenter, A. B. & Dave, D. D. (1999). Th in Prep Pap test: Performance and biopsy follow-up in a university hospital. Cancer, 87 (3), 105-12. Carter, J. J., Madeleine, M. M., Shera, K ., Schwartz, S. M., Cushing-Haugen, K. L., Wipf, G. C., Porter, P., Daling, J. R ., McDougall, J. K., & Galloway, D. A. (2001). Human papillomavirus 16 and 18 L1 serology compared across anogenital cancer sites. Cancer Research, 61, 1934-1940. Centers for Disease Contro l and Prevention. (2002). 1999 BRFSS summary quality control report. Retrieved September 29, 2002 from http://www.cdc.gov/nccdphp/brfss/pdf/99quality.pdf

PAGE 160

149 Centers for Disease Control and Preventi on, National Center for Chronic Disease Prevention and Health Pr omotion. (2002, March 22). 1999 BRFSS Codebook. Retrieved September 29, 2002, from http://www.cdc.gov/brfss/ti-surveydata1999.htm Centers for Disease Control and Preventi on, National Center for Chronic Disease Prevention and Health Pr omotion. (2002, March 22). About the BRFSS. Retrieved September 29, 2002, from http://www.cdc.gov/nccdphp/brfss/about.htm Centers for Disease Control and Preventi on. National Center for Chronic Disease Prevention and Health Promotion. (1999). Behavioral risk factor surveillance system survey questionnaire. Atlanta, GA: U.S. Department of Health & Human Services. Centers for Disease Control and Preventi on. National Center for Chronic Disease Prevention and Health Pr omotion. (2002, March 30) Behavioral risk factor surveillance system training guide : Data collection and management. Retrieved September 29, 2002, from http://www2.cdc.gov/nccdphp/brfss2/training_gu/edit.asp Centers for Disease Control and Preventi on. National Center for Chronic Disease Prevention and Health Promotion. (2002, April 1) Behavioral risk factor surveillance system training guide: Prevalence data. Retrieved January 25, 2003, from http://apps.nccd.cdc.gov/brfss/race.asp Centers for Disease Control and Preventi on. National Center for Chronic Disease Prevention and Health Pr omotion. (2002, March 30) Behavioral risk factor

PAGE 161

150 surveillance system training guide: Quality assurance. Retrieved September 29, 2002, from http://www2.cdc.gov/nccdphp/br fss2/training_gu/quality.asp Centers for Disease Control and Preventi on, National Center for Chronic Disease Prevention and Health Pr omotion. (2002, March 30). Behavioral risk factor surveillance system training guide: Questionnaire development Retrieved September 29, 2002 from http://www2.cdc.gov/nccdphp/brfss2 /training_gu/Q_development.asp Centers for Disease Control and Preventi on, National Center for Chronic Disease Prevention and Health Pr omotion. (2002, March 30). Behavioral risk factor surveillance system training guide: Survey methodology Retrieved September 8, 2002, from http://www2.cdc.gov/nccdphp/br fss2/training_gu/method.asp Centers for Disease Control and Preventi on. National Center for Chronic Disease Prevention and Health Pr omotion. (2002, March 30) BRFSS in action. Retrieved September 29, 2002, from http://www.cdc.gov/nccdphp/brfss/dataused.htm Centers for Disease Control and Preventi on. National Center for Chronic Disease Prevention and Health Promotion. (2002, August 7) BRFSS survey techniques. Retrieved September 29, 2002, from http://www2.cdc.gov/nccdphp/br fss2/training_ov/bst1.htm Centers for Disease Control and Preventi on. National Center for Chronic Disease Prevention and Health Promotion. (2002, August 30) BRFSS: Tracking major health risks in America (BRFSS At a glance). Retrieved September 29, 2002, from http://www.cdc.gov/nccdphp/aag/aag_brfss.htm/

PAGE 162

151 Centers for Disease Control and Preventi on. National Center for Chronic Disease Prevention and Health Promotion. (2002, August 7) BRFSS: Tracking public health trends (BRFSS strengths: comparability). Retrieved September 29, 2002, from http://www2.cdc.gov/nccdphp/br fss2/training_ov/tpht9.htm Centers for Disease Control and Preventi on, National Center for Chronic Disease Prevention and Health Promotion. (2002, August 7). BRFSS weighting formula Retrieved September 29, 2002, from http://www.cdc.gov/nccdphp/brfss/ti-docs.htm Centers for Disease Control and Preventi on, National Center for Chronic Disease Prevention and Health Promotion. (1999, Dec 12). Notes for data users Retrieved September 29, 2002, from http://www.cdc.gov/nccdphp/brf ss/prevdata/usernote.htm Centers for Disease Control and Preventi on, National Center for Chronic Disease Prevention and Health Pr omotion. (2002, March 22). Overview: BRFSS 1999. Retrieved September 29, 2002, from http://www.cdc.gov/nccdphp/brfss/ti-surveydata1999.htm Centers for Disease Control and Preventi on, National Center for Chronic Disease Prevention and Health Promotion, Divisi on of Adult and Community Health. (2000, Jan. 7). Behavioral risk factor survei llance system training guide. Retrieved June 11, 2000, from http://ww2.cdc.gov/nccdphp/brfss2/training Centers for Disease Control and Prevention, National Center for Health Statistics, Healthy People 2000 review, (1998-1999).

PAGE 163

152 Champion, V., & Menon, U. (1997). Predic ting mammography and breast selfexamination in African American women. Cancer Nursing, 20 (5), 315-22. Chao, A., Becker, T. M., Jordan, S. W., Darling, R., Gilliland, F. D., & Key, C. R. (1996). Decreasing rates of cervical cancer among American Indians and Hispanics in New Mexico (United States). Cancer Causes and Control, 7 (2), 205-13. Clark, R. (1995). Principles of cancer screening. Cancer Control Journal JMCC, 2, 485-92. Clark, R. A., & Reintgen, D. S. (1996). Princi ples of cancer screening. In D S. Reintgen & R. A. Clark (Eds.), Cancer Screening (pp. 1-20). St. Louis: Mosby. Cohen, J. (2005). High hopes and dilemma s for a cervical cancer vaccine. Science, 308, 618-621. Coughlin, S. S., Uhler, R. J., Hall, H. I., & Briss, P. A. (2004). Nonadherence to breast and cervical cancer screening: What are the linkages to chronic disease risk? Preventing Chronic Disease. Retrieved June 16, 2004, from http://www.cdc.gov/pcd/issues/2004/jan/03_0015.htm Coughlin, S. S., Thompson, T. D., Seeff, L., Richards, T. & Stallings, F. (2002). Breast, cervical, and colorectal carcinoma in a demographically defined region of the southern U. S. Cancer, 95 (10), 2211-22. Cuzick, J., Szarewski, A., Terry, G., Ho, L., Hanby, A., Maddox, P., Anderson, M., Kocjan, G., Steele, S. T., & Guilleba ud, J.(1995). Human papillomavirus testing in primary cervical screening. The Lancet, 345 (8964), 1533-1536.

PAGE 164

153 Department of Health and Human Services 1987. Poverty income guidelines: Annual revision. Federal Register, 52, 5340-5341. Department of Health and Hu man Services. (1999, Sep. 13). Race and health: cancer management Retrieved November 09, 2002, from http://raceandhealth.hhs. gov/2ndpgblue/2pgCancer.htm Department of Health and Huma n Services. (2000, November 13). Eliminating racial and ethnic disparities in health. Retrieved September 29, 2002, from http://raceandhealth.hhs.gov/ sidebars/sbinitOver.htm Department of Health and Human Services. 1998. Healthy People, 2010 Objectives: Draft for public comment. September. Washington, DC. Department of Health and Human Services. 2000. Healthy People 2010 (Conference Edition in Two Volumes). Washington, DC. Dorland’s illustrated medical dictionary (29th ed.). (2000). Philadelphia, PA: W. B. Saunders Company. Ferenczy, A., Gelfand, M. M., Franco, E., & Mansour, N. (1997). Human papillomavirus infection in postmenopausal women with and without hormone therapy. Obstetrics & Gynecology, 90, 7-11. Ferrante, J. M., Gonzalez, E. C., Roetzhei m, R. G., Pal, N., & Woodard, L. (2000). Clinical and demographic predictors of late-stage cervical cancer. Archives of Family Medicine, 9 (5), 439-45. Fishbein, M., & Ajzen, I. (1975). Beliefs, attitudes, intention and behavior: An introduction to theory and research. Reading, MA: Addison-Wesley.

PAGE 165

154 Fishbein, M., & Guinan, M. ( 1996). Behavioral science and public health: a necessary partnership for HIV prevention. Public Health Reports, 3 (Suppl.) 5-10. Flores, E. T., Castro, F. G., & Fernandez-Es quer, M. E. (1995). Social theory, social action, and intervention research: Imp lications for cancer prevention among Latinos. Journal of the National Cancer Institute. Monographs (18), 101-8. Fontaine, K. R., Faith, M. S., Allison, D. b., & Cheskin, L. J. (1998). Body weight and health care among women in the general population. Archives of Family Medicine, 7 (4), 381-4. Fos, P. J. & Zungia, M. A. (1999). Assessmen t of primary health care access status: An analytic technique for decision making. Health care Management Science 2, 229-238 Fox, J. (1997). Applied regression analysis, linea r models, and related models. Thousand Oaks, CA: Sage Publications, Inc. Franco, E. L. (2003). Primary screening of cervical cancer with human papillomavirus tests. Journal of the National C ancer Institute Monographs, 31, 89-96. Franco, E. L., Duarte-Franco, E. & Frenczy, A. (2001). Cervical cancer: epidemiology, prevention and the role of human papillomavirus infection. Canadian Medical Association Journal, 164 (7), 1017-1025. Fries, J. E., Koop, C. E., Beadle, C. E., Cooper, P. P., England, M. J., Greaves, r. F., Sokolov, J. J., & Wright, D. (1993). Reduc ing Health care costs by reducing the need and demand for medical services The Health Project Consortium. New England Journal of Medicine, 329 (5), 321-5.

PAGE 166

155 Gentry, E. M., Kalsbeek, W. D ., Hogelin, G. C., Jones, J. T., Gaines, K. L., Forman, M. R., Marks, J. S., & Trowbridge, F.L. (1985) The behavioral risk factor surveys: II. Design, methods, and estimates from combined state data. American Journal of Preventive Medicine, 1 (6), 9-14. Gibson, L., Spiegelhalter, D. J., Camilleri-Fe rrante, C., & Day, N. E. (1997). Trends in invasive cervical cancer incidence in East Anglia from 1971 to 1993. Journal of Medical Screening, 4 (1), 44-8. Glanz, K., & Rimer, B. K. (1995). Theory at a glance: A gu ide to health promotion practice. (NIH publication no. 95-3896). Bethesda, MD: National Institutes of Health, National Cancer Institute. Glanz, K., Lewis, F. M., & Rimer, B. K. (Eds.). (1997). Health behavior and health education: Theory, research and practice (2nd ed.). San Francisco: Jossey-Bass. Glanz, K. & Rimer, B. K. (1997). Theory at a glance: A guide to health promotion practice. US Department of Health and Human Services, Public Health Service, National Institutes of Health, National Cancer Institute. Gove, P. B. (Ed.). (1993). Webster’s third new inter national dictionary of the english language. Springfield, MA: Merriam-Webster Inc. Gray, S. H. & Walzer, T. B. (2004). New st rategies for cervical cancer screening in adolescents. Current Opinion in Pediatrics, 16 (4), 344-349. Greenlee, R.T, Murry, T, Bolden, S, & Wingo, P.A. (2000). Cancer statistics 2000. CA: A Cancer Journal for Clinicians, 50 (1), 7-33. Guidozzi, F. (1996). Screening for cervical cancer. Obstetrical and Gynecological Survey, 51 (4), 247-52.

PAGE 167

156 Gulitz, E., Bustillo-Hernandez, M., & Kent E. B. (1998). Missed cancer screening opportunities among older women: A provider survey. Cancer Practice, 6 (6), 325-32 Gwede, C. (2001). Predictors of prostate cancer screening among African American men: A test of the preventive health model. Doctoral dissertation, University of South Florida. Harper, D. M., Franco, E. L., Wheeler, C., Fe rris, D. G., Jenkins, D., Schuind, A., Zahaf, T., Innis, B., Naud, P., De Carvalho, N. S ., Roteli-Martins, C. M., Teixeira, J., Blatter, M. M., Korn, A. P., Quint, W., & Dubin, G. (2004). Efficacy of a bivalent L1 virus-like particle prevention of infection of infection with human papillomavirus types 16 and 18 in young women: A randomized controlled trial. The Lancet, 364 1757-1765. Harewood, G. C., Wiersema, M. J., & Melton, J. L. III (2002). A prospective, controlled assessment of factors influencing acceptance of scr eening colonoscopy. The American Journal of Gastroenterology, 97 (12), 3186-94. Hatcher, L. & Stepanski, E. J. (1994). A step-by-step approach to using the SAS system for univariate and multivariate statistics. Cary, NC: SAS Institute Inc. Hayward, R. A., Shapiro, M. E., Freeman, H. E., & Corey, C. R. (1988). Who gets screened for cervical and breast cancer? Results from a new national survey. Archives of Internal Medicine, 145 (5), 1177-81. Healthy People: The Surgeon General’s Report on He alth Promotion and Disease Prevention. Washington, DC: US Department of Health and Human Services, Public Health Service, 1979.

PAGE 168

157 Healthy People 2000: National Health Promotion and Disease Prevention Objectives. Washington, DC: US Department of Hea lth and Human Services, Public Health Service, 1990. Healthy People 2000 Final Review: National Health Promotion and Disease Prevention Objectives. Washington, DC: US Department of Health and Human Services, Public Health Service, 2001. Healthy People 2010: National Health Promotion and Disease Prevention Objectives. Washington, DC: US Department of Hea lth and Human Services, Public Health Service, 2000. Health Promotion/Disease Prevention: Objectives for the Nation. Washington, DC: US Department of Health and Human Services, 1981. Helms, L. J. & Melnikow, J. (1999). Dete rmining costs of health care for costeffectiveness analyses: The case of cervi cal cancer prevention and treatment. Medical Care, 37 (7), 652-661. Hennekens, C. H. & Buring, J. E. (1987). Epidemiology in medicine (1st ed.). Boston, Massachusetts: Little, Brown and Company. Herrero, R. (1996). Epidemio logy of cervical cancer. Journal of the National Cancer Institutes. Monograph. 21, 1-6. Hiatt, R. A., Klabunde, C., Breen, N., Swa n, J. & Ballard-Barbash, R. (2002). Cancer screening practices from National Health Interview Surveys: Past, present, and future. Journal of the National Cancer Institute, 94 (24), 1837-1846. Hiatt, R. A. & Pasick, R. J. (1996). Unsolved problems in early br east cancer de tention: Focus on the underserved. Breast cancer Research & Treatment, 40 (1), 37-51.

PAGE 169

158 Hinkula, M., Pukkala, E., Kyyronen, P., Laukkanen, P., Koskela, P., Paavonen, J., Lehtinen, m., & Kauppila, A. (2004). A population-based study on the risk of cervical cancer and cervical intraepit helial neoplasia among grand multiparous women in Finland. British Journal of Cancer, 90 (5), 1025-9. Ho, G. Y., Bierman, R., Beardsley, L., Chang, C. J. & Burk, R. D. (1998). Natural history of cervicovaginal papillomavir us infection in young women. New England Journal of Medicine, 338, 423-428. Hochbaum, G. M. (1958). Public participation in me dical screening programs: a sociopsychological study. (PHS Publication No. 572). Washington, DC: US government Printing Office. Hoffman, M. S. & Cavanagh, D. (1996). Cervical cancer. In D. S. Reintgen & R. A. Clark (Eds.), Cancer Screening (pp. 41-54), St. Louis: Mosby. Hoffman-Goetz, L., Breen, N. L., & Meissner, H. (1998). The impact of social class on the use of cancer screening within three racial/ethnics groups in the United States. Ethnicity and Disease, 8 (1), 43-51. Holmquist, N. D. (2000). Revising the effect of the Pap test on cervical cancer. American Journal of Public Health, 90 (4), 620-3. Hristova, L., & Hakama, M.(1997). Effect of sc reening for cancer in the Nordic countries on deaths, cost and quality of life up to the year 2017. Acta Oncologica, 36 (Suppl. 9), 1-60. Hubbell, F. A., Chavez, L. R., Mishra, S. I. & Valdez, R. B. (1996). Beliefs about sexual behavior and other predictors of Pa panicolaou smear screening among Latinas and Anglo women. Archives of Internal Medicine, 156 (20), 2353-8.

PAGE 170

159 Institute of Medicine, Committee for the Study of the Future of Public Health. The Future of Public Health. Washington, DC: National Academy Press, 1988. Institute of Medicine, Nationa l Academy of Sciences (1988). The future of public health. Washington, DC: National Academy Press. Jaccard, J. (2001). Interaction effects in logistic regression Sage University Papers Series on Quantitative Applications in the Social Sciences, 07-135, Thousand Oaks, CA: Sage. Jacobsen, D. L., Womack, S. D, Peralta, L ., Zenilman, J. M., Feroli, K., Maehr, J., Daniel, R. W., & Shah, K. V., 2000. Concor dance of human papillomavirus in the cervix and urine among inner city adolescents. The Pediatric Infectious Disease Journal 19 (8), 722-729. Janz, N. K. & Becker, M. H. (1984). Th e health belief model: a decade later. Health Education Quarterly, 11, 1-47. Jayant, K., Rao, R. S., Nene, B. M., & Dale P. S. (1994). Early diagnosis of cervical cancers. Our experience at the Rural Cancer Registry, Barshi, India. Indian Journal of Cancer, 31 (2), 59-63. Jennings, K. M. (1997).Getting a Pap smear: focus group responses of African-American and Latina women. Oncology Nursing Forum, 24, 827-835. Jennings-Dozier, K. & Lawrence, D. (2000). So cio-demographic predic tors of adherence to compliance with the 1999 cervical cance r screening guidelines of the ACS in minority women. Cancer Nursing, 23 (5), 350-356. Kagawa-Singer, M. (1997). Addressing issues for early detection and screening in ethnic populations. Oncology Nursing Forum, 24 (10), 170-11.

PAGE 171

160 Kasl, S. V., & Cobb, S. (1966). Health behavior illness behavior a nd sick behavior: I. Health and illness behavior. Archives of Environmental Health, 12, 246-266. Katz, S. J., & Hofer, T. P. (1994). Socio-econo mic disparities for preventive care persists despite universal coverage: breast and cervical cancer screening in Ontario and the United States. Journal of the American Medical Association, 272 (7), 530534. Kelaher, M., Gillespie, A. G., Allotey, P ., Manderson, L., Potts, H., Sheldrake, M., & Young (1999). The Transtheoretical Model and cervical cancer screening: Its application among culturally diverse co mmunities in Queensland, Australia. Ethnicity & Health, 4 (4), 259-76. King, M. (1990). Health is a sustainable state. Lancet, 336, 664-607. Kiviat, N. (1996). Natural history of ce rvical neoplasia: Overview and update. American Journal of Obstetrics and Gynecology, 175 (4, Pt. 2), 1099-104. Kiviat, N. B., Koutsky, L. A., Critchlow, C. W., Lorincz, A. T., Cullen, A. P., Brockway, J., & Holmes, K. K. (1992). Prevalence a nd cytologic manifestations of human papillomavirus (HPV) types 6, 11, 16, 18, 31, 33, 35, 42, 43, 44, 45, 51, 52, and 56 among 500. International Journal of gynecological Pathology, 11 (3), 197203. Klaes, R., Friedrich, T., Spitkovsky, D., Ri dder, R., Rudy, W., Petry, U., DallenbachHellweg, G., Schmidt, D., & von Knebel Doebevitz, M. (2001). Overexpression of p16 (INK4A) as a specific marker for dysplastic and neoplastic epithelial cells of the cervix uteri. International Journal of Cancer, 9 (2), 276-84.

PAGE 172

161 Koop, C. E. (1997). Forward. In B. Burgower (Ed.) Principles of Public Health Practice (pp vi-vii). Albany, NY: De lmar Publishers Inc. Kretzer, E. K., & Larson, E. L. (1998). Beha vioral interventions to improve infection control practices. American Journal of Infection Control, 26 (3), 245-253. Ku, N. N. (1999). Automated Papanicolaou smear analysis as a screening tool for female lower genital malignancies. Current Opinions in Obst etrics & Gynaecology, 11 (1), 41-3. Kurman, R. J. (1992). Human papilloma virus infection of the cervi x: relative risk of associations of 15 co mmon anogenital types. Obstetrics and Gynecology, 79, 328-37. Kurman, R. J., Malkasian, G. D., Jr., Se dlis, A., & Solomon, D. (1991). From Papanicolaou to Bethesda: The rationa le for a new cytologic classification. Obstetrics and Gynecology, 77 (5), 779-782. Landis, S. H., Murray, T., Bolden, S., & Wingo, P. A. (1999). Cancer Statistics. CA: A Cancer Journal for Clinicians 49 (1) 8-31. Landrine, H. & Klonoff, E. A. (1994). Th e African American acculturation scale: Development, Reliability, and Validity. Journal of Black Psychology, 20 (2), 104-127. Last, J. M. (Ed.). (1995). A dictionary of epidemiology (3rd ed.). New York: Oxford University Press, Inc. Lee, M. C. (2000). Knowledge, barriers a nd motivators to cervical cancer screening among Korean American women: a focus group approach. Cancer Nursing, 23 (3), 245-253.

PAGE 173

162 Liberatos, P., Link, G., & Kelsey, J. L. ( 1998). The measurement of social class in epidemiology. Epidemiologic Reviews, 10, 87-121. Linder, J. (1998). Recent advances in thin-layer cytology. Diagnostic Cytology, 18 (1), 24-32. Linder, J. & Zahniser, D. ( 1998). ThinPrep Papanicolaou test ing to reduce false-negative cervical cytology. Archives of Pathology and Laboratory Medicine, 122 (2), 139-44. Lippert, M. T., Eaker, E. D ., Vierkant, R. A. & Remmingt on, P. L. (1999). Breast cancer screening and family history among rural women in Wisconsin. Cancer Detection & Prevention, 23 (3), 265-72. Lipskie, T. L. (1998). A summary of cancer screening guidelines. Chronic Diseases in Canada, 19 (3), 12-30. Lorincz, A. T., Reid, R., Jenson, A. B., Gree nberg, M. D., Lancaster, W., & MacLean, C. D. (1992). Principles of cancer screening. Medical Clinics of North America, 80 (1), 1-14. Lyson, T. A. & Falk, A. A. (Eds.) (2000). Forgotten places. Uneven development and the loss of opportunity in rural America. Lawrence, KS: University Press of Kansas. Makuc, D. M., Freid, V. M. & Kleinman, J. C. (1989). National trends in the use of preventive health care by women. American Journal of Public Health 79, 21-26. Mandelblatt, J., Freeman, H., Winczewski, D ., Cagney, K., Williams S., Trowers, R., Tang, J., Gold, K., Lin, T. H., & Kerner, J. (1997). The costs and effects of cervical and breast cancer in a public hospital emergency room. The Cancer Control Center of Harlem. American Journal of Public Health, 87 (7), 1182-9.

PAGE 174

163 Mandelblatt, J. S. & Yabroff, K. R. (2000). Breast and cervical cance r screening for older women: Recommendations and challenges for the 21st century. Journal of the American Medical Women’s Association, 55 (4), 210-5. Martin, L. M., Parker, S. L., Wingo, P. A., & Heath, C. W. (1996). Cervical cancer incidence and screening: Status re port on women in the United States. Cancer Practice, 4 (3), 130-4. Masood, S. (1997). Why women s till die from cervical cancer. Journal of the Florida Medical Association, 84 (6), 379-83. McIndoe, W. A., McLean, M. R., Jones, R. W., & Mullins, P. R. (1984). The invasive potential of carcinoma in situ of the cervix. Obstetrics and Gynecology, 64 (4), 451-8. Miller, A. B. (1995). An epidemiol ogical perspective on cancer screening. Clinical Biochemistry, 28 (1), 41-8. MMWR (1997). Regulatory closure of ce rvical cancer cyto logy laboratories: Recommendations for a public health response. Morbidity & Mortality Weekly Reports, 46 (RR-17), 1-19. Mo, B. (1992). Modesty, sexuality, and breas t health in Chinese American women. Western Journal of Medicine, 157 (3), 260-4. Munoz, N., & Bosch, F. X. (1992). HPV and cer vical neoplasia: Review of case-control and cohort studies. IARC Scientific Publications, 119, 251-61. Lyon. Munoz, N., & Bosch, F. X., de Sanjose, S., & Shah, K. V. (1994). The role of HPV in the etiology of cervical cancer. Mutation Research, 305, (2), 293-301.

PAGE 175

164 Munoz, N., & Bosch, F. X., de Sanjose, S., Tafu r, L., Izarzugaza, I., Gili, M., Viladiu, P., Navarro, C., Markos, C., & Ascunce, N. (1992). The causal link between human papilloma virus and invasive cervical cancer: A population-based case-control study in Colombia and Spain. International Journal of Cancer, 52 (5), 743-9. Munoz, N., & Bosch, F. X., de Sanjose, S., Viladiu, P., Tormo, J., Moreo, P., Ascunce, N., Gonzalez, L. C., Tafur, L., & Gili, M. (1993). Human papilloma virus in the etiology of cervico uterine cancer. Boletin de la Oficina Sanitaria Panamericana, 115 (4), 301-9. Munoz, N., Bosch, F. X., Castellsague, X., Diaz, M., de Sanjose, S., Hammouda, D., Shah, K. V., & Meijer, C. J. L. M. (2004). Against which human papillomavirus types shall we vaccinate and screen ? The international perspective. International Journal of Cancer, 111, 278-285. Munro, B. H. (1997). Statistical methods for health care research (3rd ed.). Philadelphia, PA: Lippincott-Raven Publishers. Myers, R. E., Ross, E., Jepson, C., Wolf, T., Balshem, A., Millner, L., & Leventhal, H. (1994). Modeling adherence to colorectal can cer screening. Preventive Medicine, 23, 142-151. Myers, R. E., Wolf, T. A., McKee, L., McGory, G., Bur gh, D. Y., Nelson, G., & Nelson, G. A. (1996). Factors associated with in tention to undergo annual prostate cancer screening among African Ameri can men in Philadelphia. Cancer, 78 (3), 471-9. Nasiell, K., Nasiell, M., & Vaclavinkova, V. (1983). Behavi or of moderate cervical dysplasia during long-term follow-up. Obstetrics and Gynecology, 61 (5), 609-14.

PAGE 176

165 Nasiell, K., Roger, U., & Nasiell M. (1986). Behavior of mild cervical dysplasia during long-term follow-up. Obstetrics and Gynecology, 67, 665-9. National Cancer Institute. (2001, Aug. 01). Cancer facts. Retrieved November 11, 2002, from http://cis.nci.nih.gov/fact/5_16.htm National Cancer Institute, DCCPS, Surveill ance Research Program, Cancer Statistics Branch (2004). Surveillance, Epidemiology, and End Results (SEER) program SEER*Stat Database: Mortality – All COD, public-use with state, total U. S. for expanded Races/Hispanics (1990-2000). National Cancer Institute, DCCPS, Surveill ance Research Program, Cancer Statistics Branch (2005). Surveillance, Epidemiology, and End Results (SEER) program SEER*Stat Database: Mortality – All COD, public-use with state, total U. S. for expanded Races/Hispanics (1990-2002). National Cancer Institute, Statistical Re search and Applicat ions Branch (2003). DEVCAN: Probability of Developing or Dying from Cancer Software, version 5.1. National Cancer Institute Workshop (1989) The 1988 Bethesda System for reporting cervical/vaginal cytologic diagnoses. Journal of the American Medical Association, 262, 931-4. National Center for Health Statistics (1985). The National Health Interview Survey Design, 1973-1984, and Procedures, 1975-1983. (PHS Publication No. 85-1320). U.S. Department of Health and Human Services.

PAGE 177

166 National Committee for Quality Assurance (2002). The State of Health Care Quality. Retrieved April 10, 2005, from http://www.ncqa.org/sohc2002/SOHC_2002_CCS.html National Institutes of Health (1996). National Institutes of Health consensus development conference statement: Cervical cancer, April 1-3, 1996. National Institutes of Health Consensus Development Panel. Journal of the National Cancer Institute Monographs, 21, vii-xix National Institutes of Health. (1999, Sep. 30). Addressing health disparities: The NIH Program of Action. Retrieved November 11, 2002, from http://healthdisparities.nih.gov/whatare.html National Institutes of Health. (1999, Oct. 20). NIH Guide: Understanding and eliminating minority health disparities. Retrieved November 11, 2002, from http://grants.nih.gov/ grants/guide/rfa-f iles/RFA-HS-00-003.html National Institutes of Health, Na tional Cancer Institute. (n.d.). Cancer.gov Dictionary. Retrieved November 11, 2002, from http://www.cancer.gov/dictionary National Vital Statistics System (N VSS), CDC, NCHS, November 30, 1999. NCMHD Congressional testimony. National In stitutes of Health. (n.d.). Retrieved November 11, 2002, from http://ncmhd.nih.gov/about_ncmhd/senate.asp NCMHD history. National Inst itutes of Health. (n.d.). Retrieved November 11, 2002, from http://ncmhd.nih.gov/about_ncmhd/history.asp NCMHD What we do. National Institutes of Health. (n.d.). Retrieved November 11, 2002, from http://ncmhd.nih.gov/about_ncmhd/what.asp

PAGE 178

167 NIH Consensus Development Conferen ce on Cervical Cancer (1996). NIH urges wider screening for cervical cancer (Press release, April 03). Office of Management and Budget (1997). Colle cting and presenting federal data on race and ethnicity. Federal Register, 62, FR 58781-58790. O’Malley, A. S., Kerner, J., Johnson, A. E. & Mandelblatt, J. ( 1999). Acculturation and breast cancer screening among Hisp anic women in New York City. American Journal of Public Health, 89 (2), 210-227. O’Malley, A. S., Mandelblatt, J., Gold, K., Ca gney, K. A. & Kerner, J. (1997). Continuity of care and the use of breast and cervi cal cancer screeni ng services in a multiethnic community. Archives of Internal Medicine 157 (13), 1462-1470. O’Malley, M. S., Earp, J. A., Hawley, S. T., Sc hell, M. J., Mathews, H. E. & Mitchell, J. (2001). The association of race/ethnicit y, socio-economic status, and physician recommendation for mammography: who gets the message about breast cancer screening? American Journal of Public Health, 91 (1), 49-54. On-line Medical Dictionary (19972004). Retrieved May 01, 2004 from http://cancerweb.ncl.ac.uk/cgi-bin /omd?On-line+Medical+Dictionary On-line Medical Dictionary (19972003). Retrieved January 18, 2004 from http://cancerweb.ncl.ac.uk/cgi-bin /omd?On-line+Medical+Dictionary Orbell, S. (1996). Cognition and affect afte r cervical cancer screen ing: the role of previous test outcome and personal oblig ation in future upt ake expectations. Social Science Medicine, 23, 1237-1243. Ostor, A. G. (1993). Natural history of cervica l intraepithelial neoplasia : a critical review. International Journal of Gy necological Pathology, 12, 186-92.

PAGE 179

168 Palefsky, J. M. (2003). Cervical human papillomavirus infection and cervical intraepithelial neoplasia in women positiv e for human immunodeficiency virus in the era of highly active antiretroviral therapy. Current Opinion in Oncology, 15 (5), 382-388. Papanicolaou, G. N. & Traut, H. F. (1941). The diagnostic value of vaginal smears in carcinoma of the uterus. American Journal of Obst etrics and Gynecology, 41 (2), 193-206. Parkin, D. M., Pisani, P. & Ferlay, J. (1993) Estimates of the worl dwide incidence of 18 major cancers in 1985. International Journal of Cancer, 54, 594-606. Parkin, D. M., Pisani, P. & Ferlay, J. (1999) Estimates of the worl dwide incidence of 25 major cancers in 1990. International journal of Cancer, 80, 827-41. Perez-Stabile, E. J., Sabogal, F., Otero-Sabogal, R., Hiatt, R. A., & McPhee, S. J. (1992). Misconceptions about cancer among Latinos and Anglos. Journal of the American Medical Association, 268 (22), 3219-23. Peterson, O. (1956). Spontaneous course of cervical precancerous conditions. American Journal of Obstetrics and Gynecology, 72, 1063. Poljak, M., Marin, I. J., Seme, K. & Vince, A. (2002). Hybrid capture II HPV test detects at least 15 human papilloma virus genotype s not included in its current high-risk probe cocktail. Journal of Clinical Virology, 25 (Suppl 3) 89-97. Potosky, A. L., Breen, N., Graubard, B. I. & Parsons, P. E. (1998). The association between health care coverage and the use of cancer screening tests. Results from the 1992 National Health Interview Survey. Medical Care, 36 (3), 257-70.

PAGE 180

169 Prochaska, J. O., Velicer, W. F., Rossi, J. S., Goldstein, M. G., Marcus, B. H., Rakawski, J. W., Fiore, C., Harlow, L. L., Reddi ng, C. A., Rosenbloom, D. et al. (1994). Stages of change and decisional ba lance for twelve problem behaviors. Health Psychology, 13, 39-46. Prochaska, J. O., Redding, C. A. & Evers, K. E. (1997). Health behavior and health education. In K. Glanz, F. M. Lewis & B. K. Rimer (Eds.), The transtheoretical model and stages of change (pp 60-84). San Francisco: Jossey-Bass. Rakowski, W., Clark, M. A., & Ehrich, B. (1999) smoking and cancer screening for women ages 42-75: Associations in the 1990-1994 National Health Survey Interviews. Preventive Medicine, 29 (6, Pt. 2), 489-95. Reagan, J. W., Seideman, I. L., &. Sar acusa, Y. (1953). The cellular morphology of carcinoma in situ and dysplasia of atypi cal hyperplasia of th e uterine cervix. Cancer, 6, 224-35. Rimer, B. K. (1996). Cancer screening. In D. S. Reintgen & R. A. Clark (Eds.), Adherence to cancer screening (pp 261-276). St. Louis: Mosby. Richart, R. M., & Barron, B. A. (1962). A follow-up study of patients with cervical dysplasia. American Journal of Obstetrics and Gynecology, 105, 386-93. Ries, L. A. G., Eisner, M. P., Kosary, C. L., Hankey, B. F., Miller, B. A., Clegg, L., Mariotto, A., Feuer, E. J., & Edwards, B. K. (Eds.) (2004). SEER Cancer Statistics Review, 1975-2001, National Cancer Institute. Bethesda, MD. Retrieved April 28, 2004 from http://seer.cancer.gov/csr/1975_2001/

PAGE 181

170 Roberts, J. M., Gurley, A. M., Thurloe, J. K., Bowditch, R., & Laverty, C. R. (1997). Evaluation of the ThinPrep Pap test as an adjunct to the c onventional Pap smear. Medical Journal of Australia, 167 (9), 466-9. Roetzheim, R., Fox, S. A., Leake, B. & Houn, F. (1996). The influence of risk factors on breast carcinoma screening of the Me dicare-insured older women. National Cancer institute Breast Cancer Screening Consortium. Cancer, (78), 2526-34. Rohan, T. E., Burk, R. D. & Franco, E. L. (2003). Toward a reduction of the global burden of cervical cancer. American Journal of Obstetrics and Gynecology, 189 (4), 37-39. Rosenstock, I. M., Strecher, V. J., & Becker, M. H. (1988). So cial learning theory and the health belief model. Health Education Quarterly, 15 (2), 175-183. Rosetti, D., Gerli, S., Saab, J. C. & DiRenz o, G. C. (2000). Atypical squamous cells of undetermined significance (ASCUS), lowgrade squamous intraepithelial lesion (LSIL), high-grade squamous intraep ithelial lesion (HSIL), an histology. Lebanese Medical Journal, 48 (3), 127-130. Rosner, B. A. (1995). Fundamentals of biostatistics (4th ed.). Belmont, California: Duxbury Press. Ruchlin, H. S. (1997). Prevalence and correla tes of breast and cervi cal cancer screening among older women. Obstetrics & Gynecology, 90 (1), 16-21. Runowicz, C. D., & Fields, A. L. (1999). Screening for gynecologic malignancies: A continuing responsibility. Surgical Oncology Clinics of North America, 8 (4), 703-23, vii.

PAGE 182

171 SAS v9.1 [Computer software]. (2002-2003). SAS In stitute Inc. Cary, NC: Statistical Software. Sasieni, P., & Adams, J. (1999). Effect of screening on cervical cancer mortality in England and Wales: Analysis of tren ds with an age period cohort model. British Medical Journal, 318 (7193), 1244-5. Saslow, D., Runowicz, C. D., Solomon, D., Mo scicki, A., Smith, R. A., Eyre, H. J. & Cohen, C. (2002). American Cancer Society guidelines for the early detection of cervical neoplasia and cancer. CA A Cancer Journal for Clinicians, 52, 342-362. Sato, S., Matunaga, G., Tsuji, I., Yajima, A ., & Sasaki, H. (1999). Determining the cost effectiveness of mass screening for ce rvical cancer using common analytic models. Acta Cytologia, 43 (6), 1006-14. Savage, S. A., & Clark, V. A. (1996). Fact ors associated with screening mammography and breast self-examination intentions. Health Education Research, 11, 409-421. Schifeling, D. J., Horton, J. & Tafelski, T. J. (199710. Common cancers – genetics, origin, prevention, screen ing: Parts I and II. Disease-A-Month, 43 (10), 681-742. Schiffman, M. H. (1992). Recent progress in defining the epidemiology of human papilloma virus infection and cervical neoplasia. Journal of the National Cancer Institute, 84 (6), 394-8. Schiffman, M. H., Brinton, L. A., Devessa, S. S., & Fraumeni, J. F., Jr. (1996). Cervical cancer. In D. Schottenfeld, & J. F. Fraumeni, Jr. (Eds.), Cancer Epidemiology and Prevention (2nd ed., pp. 1090-1116). New York: Oxford University Press. Schiffman, M. & Castle, P. E (2003). Human papillomavirus: Epidemiology and Public Health. Archives of Pathology and Laboratory Medicine, 127, 930-934.

PAGE 183

172 Schiffman, M., Herrero, R., Hildesheim, A., Sh erman, M. E., Bratti, M., Wacholder, S., Alfaro, M., Hutchinson, M., Morales, J., Greenberg, M. D., & Lorincz, A. T. (2000). HPV DNA testing in cervical cancer screening: Results from women in a high-risk province of Costa Rica. Journal of the Americ an Medical Association, 283 (1), 87-93. Schoenborn, C. A., & Marano, M. (1998). Curre nt estimates from the National Health Interview Survey: United States, 1987. Vital and Health Statistics, Series 10. (DHHS publication PHS 88-1594). Washingt on, DC: U.S. Government Printing Office. Schreckenberger, C. & Kaufmann, A. S. (2004) Vaccination strategies for the treatment and prevention of cervical cancer. Current Opinion in Oncology, 16 (5), 485-491. Scott, T. L., Gazmararian, J. A., Williams, M. V., & Baker, D. W. ( 2002). Health literacy and preventive health care use among medicare enrollees in a managed care organization. Medical Care, 40 (5), 395-404. Sellors, J. W., Karwalajtys, T. L., Ma hony, J. B., James, B., Lytwyn, A., Chong, S., Sparrow, J., & Lorincz, A. (2003). Incide nce, clearance and predictors of human papillomavirus infection in women. Shaver, V. L. & Brown, M. L. (2002). Racial and Ethnic disparities in the Receipt of Cancer Treatment. Journal of the National Cancer Institute, 94 (5), 334-57. Shea, S., Stein, A. D., Lantigua, R., & Basch, C. E. (1991). Reliability of the behavioral risk factor survey in a triethnic population. American Journal of Epidemiology, 133 (5), 489-500.

PAGE 184

173 Shingleton, H. M., Jones, W. B., Russell, A ., Fremgen, A., Chmiel, J. S., Ocwieja, K., Winchester, D. P., & Clive, R. (1996). Hy sterectomy in invasive cervical cancer: A national pattern of care study of th e American College of Surgeons. Journal of the American College of Surgeons, 183 (4), 393-400. Shingleton, H. M., Patrick, R. L., Johnson, W. W., & Smith, R. A. (1995). The current status of the Papanicolaou smear. Ca: A Cancer Journal for Clinicians, 45 (5), 305-20. Simoes, E. J., Newschaffer, C. J., Hagdrup, N., Ali-Abarghoui, F., Tao, X., Mack, N., & Brownson, R., G. (1999). Predictors of compliance with recommended cervical cancer screening schedule : a population-based study. Journal of Community Health, 24 (2), 115-130. Simon, M. S., Gimotty, P. A., Coombs, J., McBride, S., Moncrease, A., & Burack, R. C. (1998). Factors affecting participatio n in a mammography screening program among members of an urban Detroit health maintenance organization. Cancer Detection and Prevention, 22 (1), 30-38. Smith, R. A. (1999). Principles of successful cancer screening. Surgical Oncology Clinics of North America, 8 (4), 587-609. Smith, R. A., Mettlin, C. J., Davis, K. J., & Eyre, H. (2000). American Cancer Society guidelines for the early detection of cancer. Ca: A Cancer Journal for Clinicians, 50 (1), 34-49. Solomon, D., Davey, D., Kurman, R., Moriat y, A., O’Connor, D., Preg, M., Raab, S., Sherman, M., Wilbur, D., Wright, T ., & Young, N. (2002). The 2001 Bethesda

PAGE 185

174 System: Terminology for reporting results of cervical cytology. Journal of the American Medical Association, 287 (16), 2114-2119. Stanley, M. (2003). Genital human papillomav irus infections-current and prospective therapies. Journal of the National Can cer Institute Monographs, 31, 117-124. Stein, A. D., Lederman, R. I., & Shea S. (1994) The behavioral risk factor surveillance system questionnaire: Its reliability in a statewide sample. American Journal of Public Health, 83 (12), 1768-72. Stein, A. D., Lederman, R. I., & Shea S. ( 1996). Reproducibility of the women’s module of the behavioral risk factor surveillance system questionnaire. Annals of Epidemiology, 6 (1), 47-52. Stein, J. A., Fox, S. & Murata, P. J. (1991) The influence of ethnicity, socio-economic status, and psychological barri ers on the use of mammography. Journal of Health and Social Behavior, 32 (2), 101-13. Stoddard, A. M., Rimer, B. K., Lane, D., F ox, S. A., Lipkus, I., Luckmann, R., Avrunin, J. S., Sprachman, S., Constanza, M. & Urban, N. (1998). Underusers of mammogram screening: Stage of adoption in five U. S. subpopulations. The NCI Breast Cancer Screening Consortium. Preventive Medicine, 27 (3), 478-87. Strecher, V. J. & Rosenstock, I. M. (1997). The health belief model. In Glanz, K., Lewis, F. M., & B. K. Rimer, (Eds.). Health behavior and health education: Theory, research and practice (2nd ed.) (pp. 41-59). San Francisco: Jossey-Bass. Suba, E. J., Nguyen, C. H., Nguyen, B. D. & Raab, S. S. (2001). De novo establishment and cost-effectiveness of Papanicolaou cytology screening services in the Socialist Republic of Vietnam. Cancer, 91 (5), 928-39.

PAGE 186

175 Tabachnick, B. G., & Fidell, L. S. (1996). Using multivariate statistics, (3rd ed.). New York: Harper Collins. Taylor, V. M., Jackson, J. C., Tu, S. P ., Yasui, Y., Schwartz, S. M., Kuniyuki, A., Acorda, E., Lin, K., & Hislop, G. (20 02). Cervical cancer screening among Chinese Americans. Cancer Detection & Prevention, 26 (2), 139-45. Tilley, B. C., Glanz, K., Kristal, A. R., Hi rst, K., Li, S., Vernon, S. W., & Myers, R. (1999b). Nutrition intervention for high-risk auto workers: results of the Next Step Trial. Preventive Medicine, 28 284-292. Tilley, B. C., Vernon, S. W., Myers, R., Glanz, K., Lu, M., Hirst, K., & Kristal, A. R. (1999a).The Next Step Trial: impact of a worksite colorectal cancer screening program. Preventive Medicine, 28, 276-283. Tjalma, W. A. A., Arbyn, M., Paavonen, J., Van Waes, T. R. & Bogers, J. J. (2004). Prophylactic human papillomavirus vaccines: The beginning of the end of cervical cancer. International Journal of Gynecological Cancer, 14, 751-761. Turnock, B. J. (1997). Public health: What it is and how it works. Gaithersburg, MD: Aspen Publishers, Inc. Tzeng, O. C. & Jackson, J. W. (1991). Common methodological framework for theory construction and evaluation in the social and behavioral sciences. Genetic, Social, and General Psychology Monographs, 117 (1), 49-76. U. S. Census Bureau, Geography Division, (2003). Federal Information Processing Standards (FIPS) Codes. Retrieved February 21, 2005, from http://www.census.gov/fips/fips.html

PAGE 187

176 Vilos, G. A., (1998). The history of the Pa panicolaou smear and the odyssey of George and Andromache Papanicolaou. Obstetrics & Gynecology, 91 (3), 479-83. Vogt, T. M., Glass, A., Glasgow, R. E., La Chance, P. A., & Lichtenstein, E. (2003). The safety net: A cost-effective approach to improving breast and cervical cancer screening. Journal of Woman’s Health, 12 (8), 789-798. Vonka, V., Kanka, J., Hirsch, I., Zavadova, H., Krcmar, M., Suchankova, A., Rezacova, D., Broucek, J., Press, M., & Domoraz kova, E. (1984). Prospective study on the relationship between cervical neoplasia and herpes simplex type-2 virus. II. Herpes simplex type-2 antibody presence in sera taken at enrollment. International Journal of Cancer, 33 (1), 61-6. Waksberg, J., (1978). Sampling meth ods for random digit dialing. Journal of the American Statistical Association, 73, 40-49. Watts, B. G., Vernon, S. W., Myers, R. E., & Tilley, B. C. (2003). Intention to be screened over time for colorectal cancer in male automotive workers. Cancer Epidemiology, Biomarkers & Prevention, 12, 339-349. Webster’s New American Dictionary (1995). New York, NY: Smithmark Publications. Weinstein, N. D. (1993). Testing four competi ng theories of health-protective behavior. Health Psychology, 12 (4), 324-333. Wells, B. & Horm, J. (1998). Targeting the underserved for breast and cervical cancer screening: the utility of ecological anal ysis using the National Health Interview Survey. American Journal of Public Health, 88 (10), 1484-1489. Wennberg, J. E. (1990). Outcomes research, cost containment, a nd the fear of health care rationing. New England Journal of Medicine, 323 (17), 1202-4.

PAGE 188

177 White, J. E., Begg, L., Fishman, N. W., Guth rie, B., & Fagan, J. K. (1993). Increasing cervical cancer scr eening among minority el derly. Education and on-site services increase screening. Journal of Gerontological Nursing, 19 (5), 28-34. Winslow, C. E. A. (1920). The untilled field of public health. Modern Medicine, 2, 183-191. WHO Bulletin (1996). Prevalence and Incidence, 35, 783-784. World Health Organization (2002). Cervical cancer screenin g in developing countries: Report of a WHO consultation, v. Woodard, A. & Kawachi, I. (2000). Why reduce health inequalities? Journal of Epidemiology and Community Health, 54 (12), 923-9. zur Hausen, H., (2000). Papillomaviruses causi ng cancer: evasion from host-cell control in early events in carcinogenesis. Journal of the National Cancer Institute, 92, 690-698.

PAGE 189

178 Appendices

PAGE 190

179 Appendix A Panel of Experts John Large, Ph.D. Expertise in quantitative and statistical analysis Kofi Marfo, Ph.D. Expertise in measurement and theory, application of theory to practice Thomas Mason, Ph.D. Expertise in cancer epidemiology, principles of cancer screening and secondary data analysis Richard Roetzheim, M.D. Expertise in preventive medicine, primary care research, cancer screening practice and secondary data analysis using SAS Patricia Romily, M.D. Expert ise in diagnostic ultrasound and diagnostic radiology

PAGE 191

180 Appendix B The Stepwise Logistic Regressi on Procedure for Model Selection Model A: Adjusted Odds Ratio Estimates for the Socio-demograhic-(9) Covariates of Compliance with the 1999 Cervical Cancer Scr eening Guidelines of the ACS Variable estimate Adjusted OR 95% CI P-value Child-bearing age 0.42 1.52 1.33 – 1.73 <.0001 Black ethnic origin 0.54 1.71 1.42 – 2.08 <.0001 Hispanic ethnic origin 0.34 1.40 1.00 – 1.96 0.0449 Marital status -0.13 0.88 0.76 – 1.01 0.0719 Education level 0.21 1.24 1.08 – 1.41 <.0016 Employment level -0.09 0.92 0.80 – 1.10 0.2408 Income level 0.26 1.30 1.11 – 1.52 0.0009 State FIPS codes -0.01 1.00 1.00 – 1.00 0.1013 Have health coverage 0.86 2.36 2.00 – 2.80 <.0001 Cervical cancer screening advice 0.62 1.90 1.61 – 2.16 <.0001 1. Likelihood Ratio Test of Model A versus Model B with 4 degrees of freedom Chi-square p-value=0.0318 2. Hispanic ethnic origin, marital status, employment status and state FIPS codes not statistically significant, dropped from Model A 3. Model A retained with statistically significant covariates = Model B Model B: Adjusted Odds Ratio Estimates for Socio-demograhic-(6) Covariates of Compliance with the 1999 Cervical Cancer Screening Guidelines of the ACS Variable estimate Adjusted OR 95% CI P-value Child-bearing age 0.41 1.52 1.33 – 1.73 <.0001 Black ethnic origin 0.56 1.71 1.42 – 2.08 <.0001 Hispanic ethnic origin 0.33 1.40 1.00 – 1.96 0.0449 Education level 0.21 1.23 1.08 – 1.40 0.0018 Income level 0.26 1.30 1.11 – 1.52 0.0009 Have health coverage 0.86 2.36 2.00 – 2.80 <.0001 Cervical cancer screening advice 0.62 1.90 1.61 – 2.16 <.0001 1. Hispanic ethnic origin (p-value=0.0449) retained to control for the effects of Hispanic American women in the study sample 2. Likelihood Ratio Test of Model A versus Model B with 3 degrees of freedom Chi-square p-value=0.0786 3. Model B retained with RACEGRB = Model C

PAGE 192

181 Appendix B (Continued) The Stepwise Logistic Regressi on Procedure for Model Selection Model C: Adjusted Odds Ratio Estimates for Socio-demographic-(6) and Screening-(1) Covariates of Compliance with the 1999 Cervical Cancer Screening Guidelines of the ACS Variable estimate Adjusted OR 95% CI P-value Child-bearing age 0.71 2.03 1.74 – 2.37 <.0001 Black ethnic origin 0.52 1.68 1.39 – 2.03 <.0001 Hispanic ethnic origin 0.33 1.39 1.00 – 1.94 0.0523 Education level 0.21 1.23 1.08 – 1.40 0.0023 Income level 0.17 1.18 1.03 – 1.37 0.0203 Have health coverage 0.80 2.22 1.88 – 2.63 <.0001 Had mammogram 0.57 1.77 1.52 – 2.07 <.0001 Cervical cancer screening advice 0.67 1.96 1.69 – 2.28 <.0001 1. Had mammogram added 2. Likelihood Ratio Test of Model C versus Model B with 1 degree of freedom Chi-square p-value=0.0202 3. Model C retained with statistically significant covariates = Model D Model D: Adjusted Odds Ratio Estimates for Socio-demographic-(6) and Screening-(2) Covariates of Compliance with the 1999 Cervical Cancer Screening Guidelines of the ACS Variable estimate Adjusted OR 95% CI P-value Child-bearing age 0.70 2.00 1.72 – 2.35 <.0001 Black ethnic origin 0.59 1.76 1.45 – 2.13 <.0001 Hispanic ethnic origin 0.35 1.40 1.01 – 1.97 0.0395 Education level 0.18 1.23 1.08 – 1.40 0.0085 Income level 0.13 1.14 1.00 – 1.32 0.0675 Have health coverage 0.78 2.25 1.92 – 2.65 <.0001 Had mammogram 0.51 1.67 1.43 – 1.96 <.0001 Had breast exam 0.78 2.22 1.79 – 2.75 <.0001 Cervical cancer screening advice 0.64 1.96 1.69 – 2.28 <.0001 1. Had breast exam added 2. Likelihood Ratio Test of Model D versus Model C with 1 degree of freedom Chi-square p-value=0.0674 3. Income level not statistically significant, dropped from Model D 4. Model D retained with statistically significant covariates = Model E

PAGE 193

182 Appendix B (Continued) The Stepwise Logistic Regressi on Procedure for Model Selection Model E: Adjusted Odds Ratio Estimates for Socio-demographic-(5) and Screening-(2) Health Profile(1) Covariates of Compliance with the 1999 Cervical Cancer Screening Guidelines of the ACS Variable estimate Adjusted OR 95% CI P-value Child-bearing age 0.71 2.03 1.73 – 2.38 <.0001 Black ethnic origin 0.53 1.71 1.41 – 2.07 <.0001 Hispanic ethnic origin 0.31 1.36 0.97 – 1.91 0.0707 Education level 0.17 1.18 1.04 – 1.35 0.0114 Have health coverage 0.78 2.17 1.84 – 2.56 <.0001 Had mammogram 0.51 1.67 1.43 – 1.95 <.0001 Had breast exam 0.80 2.22 1.79 – 2.76 <.0001 Smoking status 0.26 1.29 1.12 – 1.49 0.0004 Cervical cancer screening advice 0.63 1.89 1.63 – 2.19 <.0001 1. Smoking status added 2. Likelihood Ratio Test of Model E versus Model D with 1 degree of freedom Chi-square p-value=0.0114 3. Model E retained with statistically significant covariates = Model F Model F: Adjusted Odds Ratio Estimates for Sociodemographic-(5), Screening-(2) and Health Profile(2) Covariates of Compliance with the 1999 Cervical Cancer Screening Guidelines of the ACS Variable estimate Adjusted OR 95% CI P-value Child-bearing age 0.67 1.96 1.68 – 2.30 <.0001 Black ethnic origin 0.61 1.85 1.52 – 2.24 <.0001 Hispanic ethnic origin 0.32 1.37 0.98 – 1.92 0.0643 Education level 0.14 1.15 1.01 – 1.31 0.0358 Have health coverage 0.77 2.16 1.83 – 2.55 <.0001 Had mammogram 0.52 1.69 1.44 – 1.97 <.0001 Had breast exam 0.80 2.23 1.79 – 2.77 <.0001 Smoking status 0.29 1.34 1.16 – 1.54 <.0001 Body Mass Index 0.38 1.47 1.28 – 1.67 <.0001 Cervical cancer screening advice 0.63 1.88 1.62 – 2.19 <.0001 1. Body Max Index added 2. Likelihood Ratio Test of Model F versus Model E with 1 degree of freedom gave Chi-square p-value=0.0357 3. Model F retained with statistically significant covariates = Model G

PAGE 194

183 Appendix B (Continued) The Stepwise Logistic Regressi on Procedure for Model Selection Model G: Adjusted Odds Ratio Estimates for Socio-demographic-(5), Screening-(2) and Health Profile(3) Covariates of Compliance with the 1999 Cervical Cancer Screening Guidelines of the ACS Variable estimate Adjusted OR 95% CI P-value Child-bearing age 0.57 1.76 1.50 – 2.07 <.0001 Black ethnic origin 0.62 1.85 1.52 – 2.25 <.0001 Hispanic ethnic origin 0.30 1.36 0.97 – 1.90 0.0771 Education level 0.11 1.11 0.98 – 1.27 0.1141 Have health coverage 0.78 2.19 1.86 – 2.59 <.0001 Had mammogram 0.60 1.81 1.55 – 2.13 <.0001 Had breast exam 0.83 2.28 1.84 – 2.84 <.0001 Smoking status 0.28 1.32 1.14 – 1.53 0.0002 Body Mass Index 0.37 1.45 1.27 – 1.66 <.0001 Had hysterectomy 0.46 1.58 1.34 – 1.86 <.0001 Cervical cancer screening advice 0.62 1.86 1.60 – 2.16 <.0001 1. Had hysterectomy added 2. Likelihood Ratio Test of Model G versus Model F with 1 degree of freedom Chi-square p-value=0.1141 3. Education level not statistically significant, dropped from Model G 4. Model G retained with statistically significant variables = Model H Model H: Final Model of Adjusted Odds Ratio Estimates for Covariates of Compliance with the 1999 Cervical Cancer Screening Guidelines of the ACS Variable estimate Adjusted OR 95% CI P-value Child-bearing age 0.57 1.78 1.50 – 2.08 <.0001 Black ethnic origin 0.60 1.83 1.51 – 2.22 <.0001 Hispanic ethnic origin 0.30 1.34 0.97 – 1.89 0.0847 Have health coverage 0.80 2.23 1.90 – 2.63 <.0001 Had mammogram 0.60 1.82 1.55 – 2.14 <.0001 Had breast exam 0.84 2.32 1.84 – 2.88 <.0001 Smoking status 0.30 1.35 1.17 – 1.55 <.0001 Body Mass Index 0.38 1.46 1.28 – 1.67 <.0001 Had hysterectomy 0.47 1.60 1.36 – 1.88 <.0001 Cervical cancer screening advice 0.63 1.90 1.61 – 2.17 <.0001

PAGE 195

184 Appendix B (Continued) The Stepwise Logistic Regressi on Procedure for Model Selection Model I: Adjusted Odds Ratio Estimates for Covariates of Compliance with the 1999 cervical cancer screening guidelines of the ACS with Interaction Terms Variable estimate Adjusted OR 95% CI P-value Child-bearing age 0.57 1.77 1.50 – 2.08 <.0001 Black ethnic origin 0.62 1.85 1.46 – 2.35 <.0001 Hispanic ethnic origin 0.21 1.23 0.83 – 1.83 0.3007 Have health coverage 0.80 2.23 1.89 – 2.63 <.0001 Had mammogram 0.60 1.82 1.55 – 2.14 <.0001 Had breast exam 0.84 2.32 1.87 – 2.89 <.0001 Smoking status 0.30 1.35 1.17 – 1.55 <.0001 Body Mass Index 0.38 1.46 1.28 – 1.67 <.0001 Had hysterectomy 0.47 1.60 1.36 – 1.88 <.0001 Interaction_Black ethnic origin 0.03 0.97 0.66 – 1.44 0.8768 Interaction_Hispanic ethnic origin 0.32 1.38 0.64 – 3.00 0.4162 Cervical cancer screening advice 0.62 1.86 1.58 – 2.19 <.0001 1. Likelihood Ratio Test of Model I versus Model H with 2 degrees of freedom Chi-square p-value=0.6998 2. Interaction_Black ethnic origin and interaction_Hispanic ethnic origin not statistically significant, dropped from Model I 3. Model I retained with statistically significant covariates = Model H

PAGE 196

185

PAGE 197

About the Author Chodaesessie Wellesley-Cole Morgan was awarded the degree of Bachelor of Science with Honors, from the University of Sierra Leone, in 1975. She was awarded the degree of Master of Public Health from the University of South Florida, and was admitted to the doctoral program in 1995. Her areas of interest include epidemiology, health policy and evaluation research. Ms. We llesley-Cole Morgan has won a number of academic scholarships and honors throughout her academic career, including membership, by nomination only, to the presti gious Delta Omega Society, the National Honor Society of Public Health. Ms. Wellesley-Cole Morgan, her husband Dr. Beale Morgan, and their children Yeatie and Yannick are a family with inte rnational lan. They are multicultural and multilingual, have traveled to, lived, studied, and worked in many countries of the World. Ms. Wellesley-Cole Morgan and her family have resided in the United States since 1992.