USF Libraries
USF Digital Collections

Re-engineering graduate medical education

MISSING IMAGE

Material Information

Title:
Re-engineering graduate medical education an analysis of the contribution of residents to teaching hospitals utilizing a model of an internal medicine residency program
Physical Description:
Book
Language:
English
Creator:
Elius, Ian M
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla.
Publication Date:

Subjects

Subjects / Keywords:
GME
Healthcare system
Linear programming
Skill mix
Proficiency
Dissertations, Academic -- Industrial and Management Systems -- Masters -- USF   ( lcsh )
Genre:
government publication (state, provincial, terriorial, dependent)   ( marcgt )
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Summary:
ABSTRACT: According to the Institute of Medicine (IOM), the U.S. health care delivery system does not provide consistent, high-quality medical care to all people all the time. As a significant component of the health care delivery system, the state of Graduate Medical Education in the United States has prompted much analysis in recent years due to the general view that desired and actual outcomes are increasingly at variance with each other. One area of focus has been the implications of change for provider credentialing and funding of graduate medical education. With this research we test the hypothesis that residents perform valuable work in the teaching hospitals where they undergo training, to inform the issue regarding provider credentialing for residents.We developed a framework to compare second-year residents (PGY2), physician assistants with one year of experience, and nurse practitioners with one year of experience to measurably address the interchangeability of providers. Data was collected by obtaining expert opinions on the proficiency of the three provider options (resident, physician assistant, nurse practitioner) in performing a set of tasks/procedures by surveying the program directors of Internal Medicine residency programs in the United States. The other residency programs at the University of South Floridas College of Medicine were also surveyed to obtain measurable performance on the service providers.Statistical tools were used to analyze the survey responses, aggregate patient data and salary data for each provider. The data analysis and summary indicated that residents displayed higher levels of proficiency than physician assistants and nurse practitioners for the tasks investigated.
Thesis:
Thesis (M.S.I.E.)--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 Ian M. Elius.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 75 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 - 001670388
oclc - 62324439
usfldc doi - E14-SFE0001258
usfldc handle - e14.1258
System ID:
SFS0025579:00001


This item is only available as the following downloads:


Full Text

PAGE 1

Re Engineering Graduate Medical Education: An Analysis of the Contribution of Residents to Teaching Hospitals Utilizing a Model of a n Internal Medicine Residency Program by Ian M. Elius A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science in Industrial Engineering Department of Industrial and Management Systems Engineering College of Engineering University of South Florida Co Major Professor: Jos L. Zayas Castro, Ph.D. Co Major Professor: Peter J. Fabri M.D. Michael X. Weng, Ph.D. Date of Approval: July 12 2005 Keywords: GME healthcare system linear programming, skill mix, proficiency Copy right 2005 Ian M. Elius

PAGE 2

DEDICATION To my mother, Dillia L.T. Elius, and the other members of my family.

PAGE 3

ACKNOWLEDGMENTS First and foremost, I would like to thank God for everything. I wish to express my deepest gratitude to Dr. Jos L. Zayas Castro for providing constant encouragement and support throughout this process, and my profound thanks to Dr. Peter J. Fabri for being my guide in the healthcare domain. Also, many thanks to Dr. Michael X. Weng for his time and ins ight in helping me to complete this work. Special thanks to Dr. Anita Callahan for introducing me to this area of research Dr. Geoffrey Okogbaa for his continued support, and Professor Dolores Gooding for her wisdom. Finally, this research could not have been undertaken and successfully completed without the understanding and patience of my family and friends They have my deepest appreciation

PAGE 4

i TABLE OF CONTENTS LIST OF TABLES ................................ ................................ ................................ ............. iii LIST OF FIGURES ................................ ................................ ................................ ............. v ABSTRACT ................................ ................................ ................................ ........................ vi CHAPTER ONE INTRODUC TION ................................ ................................ .................. 1 1.1 The Health Care Sector ................................ ................................ ....................... 1 1.2 The Health Care Supply Chain (HCSC) ................................ ............................. 1 1.3 Graduate Medical Education (GME) ................................ ................................ .. 2 1.3.1 GME Funding Mechanism ................................ ................................ .............. 2 1.4 Research Objectives ................................ ................................ ............................ 3 1.5 Thesis Organization ................................ ................................ ............................ 3 CHAPTER TWO LITERATU RE REVIEW ................................ ................................ ...... 5 2.1 Introduction ................................ ................................ ................................ ......... 5 2.2 Skill Mix ................................ ................................ ................................ ............. 5 2.2.1 Health Care ................................ ................................ ................................ ..... 6 2.3 Modes/Methodologies ................................ ................................ ......................... 8 2.3.1 Survey Design and Data Collection ................................ ................................ 8 2.3.2 Modeling ................................ ................................ ................................ ......... 9 2.4 Problem Statement and Objectives ................................ ................................ ..... 9 2.5 Research Scope ................................ ................................ ................................ 10 CHAPTER THREE METHOD OLOGY ................................ ................................ .......... 11 3.1 Introdu ction ................................ ................................ ................................ ....... 11 3.2 The Rationale of the Survey Method ................................ ................................ 12 3.3 Survey Instrument Elements ................................ ................................ ............. 14 3.3.1 Internal and External Validity ................................ ................................ ....... 16 3.3.2 Data Analysis Tools ................................ ................................ ...................... 17 3.4 Construction of the Model ................................ ................................ ................ 20 3.4.1 Introduction ................................ ................................ ................................ ... 20 3.4.2 LP Formulation ................................ ................................ ............................. 21 3.5 Validation of the Model ................................ ................................ .................... 23

PAGE 5

ii CHAPTER FOUR RESULTS AND DISCUSSION ................................ ........................ 24 4.1 Introduction ................................ ................................ ................................ ....... 24 4.2 Analysis and Verification of the Research Proposition ................................ .... 24 4.2.1 Sample Size ................................ ................................ ................................ ... 26 4.2.2 Univariate Analyses Int ernal Medicine ................................ ...................... 26 4.2.3 Multivariate Analyses Internal Medicine ................................ ................... 27 4.2.4 Qualitative Analysis ................................ ................................ ...................... 35 4.3 Other Residency Programs at USF College of Medicine ................................ 35 4.4 LP Model Results ................................ ................................ .............................. 36 4.5 Discu ssion ................................ ................................ ................................ ......... 39 CHAPTER FIVE CONCLUS IONS AND FUTURE RESE ARCH ................................ 43 5.1 Introduction ................................ ................................ ................................ ....... 43 5.2 Summary of Results ................................ ................................ .......................... 43 5.3 Contributions of Research ................................ ................................ ................. 44 5.4 Future Research ................................ ................................ ................................ 45 REFERENCES ................................ ................................ ................................ ................. 47 APPENDICES ................................ ................................ ................................ .................. 50 Appendix A Residency Program Director Survey Instrument ................................ .. 51 Appendix B Linear Programming Model and Output ................................ ............... 57 Appendix C Data Analysis of Residency Programs at USF College of Medicine .... 59 Appendix D Definitions and E/M Codes for Tasks/Procedures ................................ 64

PAGE 6

iii LIST OF TABLES Table 2.1 Skill Mix Determinants, Requirements and Possible Interventions ......... 6 Table 3. 1 Formal represen tation of Kruskal Wallis Test ................................ ........... 18 Table 4 .1 Summary of Survey Results ................................ ................................ ....... 25 Table 4.2 Descri ptive Statistics: Proficiency ................................ ............................. 25 Table 4.3 Kruskal Wallis Test : Proficiency versus Provider ................................ ..... 26 Table 4.4 Kruskal Wallis Test: Proficiency versus T ask ................................ ........... 27 Table 4.5 Kruskal Wallis Test: Proficiency versus Dimension ................................ 27 Table 4.6 ANOVA: Proficiency ver sus Provider, Task, Dimension ......................... 30 Table 4.7 General Linear Model: Proficiency ver sus Provider, Task, Dimension .... 31 Table 4.8 Nested ANOVA: Proficiency v ersu s Provider (Dimension), Task ........... 31 Table 4.9 Tukey Pai rwise Comparison for Provider ................................ .................. 32 Table 4.10 Tukey Pairwise Comparison for Task ................................ ........................ 33 Table 4.11 Tukey Pair wise Comparison for Dimension ................................ .............. 34 Table 4.12 Reg ression Model for Proficiency ................................ ............................. 34 Table 4.13 Ge neral Linear Model: Proficiency versus Provider, Task, Dimensi on (Residency Programs COMED) ................................ .......................... 36 Table 4.14 LP Model Inputs Healthcare System 1 ................................ ................... 36 Table 4.15 LP Model Results (Me an Proficiency) Healthcare System 1 ................. 37 Table 4.16 LP Model Results (Media n and Mode Proficiency) Healthcare System 1 ................................ ................................ ............................... 38

PAGE 7

iv Table 4.17 LP Model Inputs Healthcare System 2 ................................ ................... 38 Table 4.18 LP Model Results (Mean Proficiency) Healthcare System 2 ................. 39 Table 4.19 LP Model Results (Median and Mode Proficiency) Healthcare System 2 ................................ ................................ ............................... 39 Table 5.1 Summary of Re search Proposition and Analysis ................................ ....... 44

PAGE 8

v LIST OF FIGURES Figure 1.1 Health Care Supply Chain ................................ ................................ ........... 1 Figure 3.1 Reengineering GME Problem Description ................................ ................ 12 Figure 3.2 Proposition and Data Analysis Strategy ................................ .................... 14 Figure 3.3 General Factori al Design ................................ ................................ ........... 19 Figure 3.4 Nested Factorial Design ................................ ................................ ............. 20 Figure 4.1 Histogram of Resi duals ................................ ................................ .............. 28 Figure 4.2 Normal Probability Plot of Residuals ................................ ........................ 29 Figure 4.3 Normal Probability Plot of Raw Data ................................ ........................ 29

PAGE 9

vi RE ENGINEERING GRADUATE MEDICAL EDUCATION: AN ANALYSIS OF T HE CONTRIBUTION OF RES IDENTS TO TEACHING HOSPITALS UTILIZING A MODEL OF AN INTERN AL MEDICINE RESIDENCY PROGRAM Ian M. Elius ABSTRACT According to the Institute of Medicine (IOM), the U.S. health care delivery system does not provide consistent, high quality medical care to all people all the time. As a significant component of the health care delivery system, t he state of Graduate Medical Education in the United States has prompted much analysis in recent years due to the general view that desired and actua l outcomes are increasingly at variance with each other. One area of focus has been the implications of change for provider credentialing and funding of graduate medical education. With this research we test the hypothesis that residents perform valuable work in the teaching hospitals where they undergo training, to inform the issue regarding provider credentialing for residents. We developed a framework to compare second year residents (PGY2), physician assistants with one year of experience, and nurse practitioners with one year of experience to measurably address the interchangeability of providers. Data was collected by obtaining expert opinions on the proficiency of the three provider options (resident, physician assistant, nurse practitioner) in pe rforming a set of tasks/procedures by surveying the program directors of Internal Medicine residency programs in the United States. The other residency programs at the University of South Floridas College of Medicine were also surveyed to obtain measurab le performan ce on the service provide rs

PAGE 10

vii Statistical tools were used to analyze the survey res ponses, aggregate patient data and salary data for each provider. The data analysis and summary indicated that residents displayed higher levels of proficiency t han physician assistants and nurse practitioners for the tasks investigated. The proficiency values were utilized as parameters to build a mathematical programming model with the objective of m aximizing total proficiency by determining the optimal number of service provider s. The model was developed and tested using data from two healthcare system s This research demonstrates that residents perform productive work at teaching hospitals, within the scope of the tasks and dimensions evaluated. Additiona lly residents work capabilities were considered in the development of a model that can be scaled to investigate questions regarding skill mix.

PAGE 11

1 CHAPTER ONE INTRODUCTION 1.1 The Health Care Sector According to the Institute of Medicine (IOM), the U.S health care delivery system does not provide consistent, high quality medical care to all people all the time [1]. National health expenditures in the United States reached $1.6 trillion in 2002 or 14.9 percent of the nations gros s domestic product (G DP), increasing 9.3% from the previous year [2]. 1.2 The Health Care Supply Chain (HCSC) The graduate medical education component of the health care supply chain includes hospitals, providers (physicians, physician assi stants, nurse practitioners), Medi care, regulatory agencies, accreditation bodies as depicted by Figure 1.1. The IOM PROVIDERS Hospitals Physicians Residents Physician Assistants Nurse Practitioners PAYERS Medicare Medicaid Faculty Practice Plans DoD VA ACCREDITATION Residency Programs Member Organizations Patients Government Agencies Figure 1.1 Graduate Medical Education Component of Health Care Supply Chain

PAGE 12

2 has made several recommendations to redesign the healt h care delivery system of the United States. One of these recommendations addresses the preparation of the workforce i n health care by convening a summit of leaders in the health professions to develop strategies for: 1) restructuring clinical education, and 2) assessing the implications of change for provider credentialing, funding and sponsorship of education programs [ 3]. 1.3 Graduate Medical Education (GME) Graduate Medical Education (GME) refers to the period in a physician in trainings education after graduation from medical school, and serves as the preparation for the independent practice of medicine. This peri od is typically of four years duration, and physicians in training are referred to as residents Residency programs operate within each area of medical specialty at designated academic health centers or teaching hospitals [4]. The Accreditation Council fo r Graduate Medical Education (ACGME) is a private professional accreditation agency responsible for the accreditation of over 7,000 residency programs. ACGME stakeholders include residency programs, member organizations, patients, government agencies, and the general public. ACGME provides its stakeholders with the assurance that residency programs are compliant with an approved set of educational standards. The standards are developed and programs reviewed b y experts in each specialty who form the Residen cy Review Committees (RRC) [4]. 1.3.1 GME Funding Mechanism GME is currently funded from a number of different sources: Medicare, Medicaid, patient care revenue, faculty practice plans, grants, endowments, Department

PAGE 13

3 of Defense (DoD), Department of Vetera ns Affairs 1 (VA) The largest contributor to the cost of educating residents is Medicare, which provides 74% of the funding. Reimbursement from Medicare is categorized into Direct Graduate Medical Education (DGME) and Indirect Medical Education (IME) pay ments [5]. DGME payments are used to support: 1) overhead expenses for GME, 2) salaries and fringe benefits for residents, 3) some compensation for teaching physician time, and 4) costs of the administrative staff for GME. Total DGME payments to academic h ealth centers were $2.7 billion in 2000 [5]. IME payments are made to the academic health centers to reimburse additional costs for: 1) more complicated cases, 2) additional tests ordered by residents as part of the learning process, 3) uncompensated care, and 4) reduced patient care productivity by teaching staff members. IME payments totaled $5.1 billion to academic health centers in 2000 [5]. 1.4 Research Objectives The goals of this work are two fold: f irst, to discover if residents perform productive w ork by exploring the interchangeability of different types of service providers in an academic health center/teaching hos pital; and s econd, to recommend an optimal skill mix for a residency program based on the results of a model constructed from the data collection and analysis. 1.5 Thesis Organization The organization of this thesis is as follows: Chapter Two reviews the work of other authors in the area of skill mix and funding models; Chapter Three explains the 1 Medicaid operates in forty three states; VA and DoD operate their own programs distinct from the remainder of the U.S.

PAGE 14

4 concepts of statistical analysis, linear programming and survey design and validity used in this research ; Chapter Four provides data analysis results and discussion ; Chapter Five describes the conclusions, contributions and possible future work.

PAGE 15

5 CHAPTER TWO LITERATURE REVIEW 2.1 Introduction Graduate Medical Education (GME) in academic health centers (AHC) has been under increasing pressure from legislative and accreditation entities to improve the quality of educational outcomes while reducing cost s [19]. A number of alternatives have been proposed including: 1) restructuring the financing mechanism for GME, and 2) r eorganizing the roles of the health care team These would be accomplished by: 1) changing the Medicare allocation from an entitlement to an appropriations process, and 2) allow ing alternative health professionals to perform more advanced tasks to help achieve the work hours and workload goals for the resident [19, 20]. 2.2 Skill Mix The maximization of output of the industrial workplace is largely attributed to the theory of s cientific management, proposed by Frederick W. Taylor in 1903. This theory espoused the decomposition of specific tasks and the interchangeability of individuals performing those tasks [18]. Initially, the tenets of scientific management were primarily ap plied to lower skilled workers in mass production facilities but they are increasingly being applied to professi onals. This is due to a shift in the mode of professional practice from individual entrepreneurial activity to multi professional corporate site s, which has occurred over time to take advantage of technological advances and central

PAGE 16

6 administrative capabilities. Contemporary professionals are primarily employees of large scale, hierarchical firms in the fields of accounting, architecture, law, and h ealth care [18]. 2.2.1 Health Care An overview of skill mix in health care can be divided into three categories: 1) mix in nursing and other non medical health professions, 2) introduction of new types of workers, and 3) role overlap between doctors and other he alth professionals [21]. Table 2.1 illustrates some determinants, requirements and possible interventions related to skill mix. Table 2.1: Skill Mix Determinants, Requirements and Possible Interventions [21] Determinant Requirement Possible interventio ns Cost containment Improved management of organizational costs, specifically labor costs Reduce unit labor costs or improve productivity by altering staff mix or level Quality improvement Improved quality of care Improve use and deployment of staff skil ls to achieve best mix New health sector programs or initiatives Maximum health benefits of program implementation, by having appropriately skilled workers in place Determine the cost effective mix of staff required; enhance skills of current staff; intro duce new types of worker Health sector reform Cost containment, improvements in quality of care and performance, and responsiveness of health sector organizations Adjust staff roles; introduce new skills and new types of worker Changes in the legislative /regulatory environment Scope for changes in (or constraints on) role for different occupations, professions Adjust staff roles; introduce new skills and new types of worker In the first category, the most common model used is the qualified/unqualified m ix with technical aides or vocationally trained assistants used to perform simple nursing tasks. The results in these cases are mixed, with a majority indicating positive outcomes with respect to cost containment. It should be noted that the majority of th e literature in this area is published by supporters of health support staff [21]. In the second category, the most visible type of new worker has been the physician assistant whose role is

PAGE 17

7 uniquely to perform specific tasks previously performed solely by physicians. There is a high growth trend in the occupation with 63,033 employed in 2002 and a projected 93,827 to be employed in 2012 [22]. In the third category, role overlap between doctors and nurses has been studied extensively with the resulting view that nurses should take on more advanced responsibilities. In many cases the outcomes were that nurses provided a more cost effective alternative to doctors, and patient satisfaction reports were higher due to the extra time nurses spent with the patients [21]. More specific to skill mix with respect to residents, a survey conducted in 1995 by the Council of Teaching Hospitals indicated that 62% of its members employed physician assistants and assistant registered nurse practitioners to undertake tasks pre viously performed by residents [20]. Hospitals were seeking alternative health professionals to residents based on the projected shift in the nations physician supply by the Council for Graduate Medical Education (COGME) at that time and the consequent re duction in the number of residents that would be available [20]. A meta analysis of the literature conducted in 2003 sought to quantify and categorize the activities that residents engaged in at a teaching hospital. The findings indicated that residents time were allotted as follows: 36% on inpatient care, 15% on teaching and learning, 35% on tasks of marginal or no educational value, and 16% on other or miscellaneous activities [20]. It should be noted at this point that grouping the inpatient care and teaching and learning categories together, and then grouping the marginal and miscellaneous categories will yield a 51% 51% 2 separation that supports 2 Percentage exceeds 100% due to rounding error.

PAGE 18

8 a major assumption of this research work of a 50% 50% split between work and study for the res idents. 2.3 Modes/Methodologies 2.3.1 Survey Design and Data Collection Survey instruments have long been used as data collection tools for various types of research studies, including those undertaken in the health care domain. In particular, the self administered questionnaire has been used to obtain data from geographically diverse populations and also as a supplement to other data collection techniques. The major source of concern regarding survey instruments relate to instrument reliability and vali dity. Reliability refers to the consistency of the data obtained, and validity refers to the accuracy of the data [25]. A crucial aspect of the instrument design is the selection of a measurement scale to record the responses for each survey item. The v isual analog scale (VAS) is a type of rating scale that allows the respondent to rank his/her preference and indicate this preference on a scale or line. The VAS derives its theoretical underpinnings from two sources: 1) the decision sciences/economics dis ciplines, which interpret the VAS scores as a measurable value function representing the strength of preferences under uncertainty distinct from a utility function, and 2) the psychology/psychophysics disciplines, which focus on the effect of stimuli and r esponse modes on judgment [28]. A comparison between the VAS and the 7 point Likert scale showed that VAS had a larger response base than the Likert scale after standardizing on a 10 point scale. However, the variability was greater with VAS and the diffe rence in responses was not statistically significant. Therefore, the two methods are comparable to represent res ults [27]. Another

PAGE 19

9 comparison including a VAS, a 12 point Borg scale, and a 5 point Likert scale concluded that the best overall scale was the VAS for reproducibility and sensitivity [26]. 2.3.2 Modeling The field of engineering has an opportunity to provide value to health care delivery by exploring the following capabilities: 1) developing better metrics, 2) identifying proven tools and methodo logies for application in health care, and 3) developing quantitative models for more thorough examination and optimization of system performance [23]. One possible area of emphasis is the utilization of mathematical programming models applied to vario us aspects of health care. It is becoming increasingly important to efficiently allocate healthcare resources, and mathematical programming can be particularly useful in budget allocations, scheduling clinical studies and assigning medical personnel [15]. 2.4 Problem Statement and Objectives The literature addresses the need for re engineering GME, and the complexities of the current GME funding mechanism [19]. Additionally, research efforts have been undertaken to provide the aggregate quantity and group ing of activities that residents perform in teaching hospitals [20]. It has also been demonstrated that alternative health professionals have been utilized to substitute for residents in performing tasks at a teaching hospital at a time when the physician workforce outlook was uncertain in the United States [21]. There appears to be a gap in the literature and corresponding knowledge on the topic of specific interchangeability among residents and alternative health professionals,

PAGE 20

10 which requires creatin g a framework to facilitate a standard measure for each type of service provider for specific tasks. Survey methodology is utilized to obtain the data in this research effort, and the literature shows that these techniques are established in the health care domain as we ll as in other fields [24]. Additionally linear programming is demonstrated as a tool with increased possibilities of application in the health care sector. Traditional areas include nurse scheduling, revenue management, hospital bed capa city modeling, reducing patient wait times, and optimizing inventory levels [15,16,17]. It appears from the literature that linear programming has not been used to model skill mix among providers in a teaching hospital. This research effort will ut iliz e this modeling approach with its inherent flexibility to allow various scen arios to be easily evaluated while yield ing optimal results, including scalability and transferability. 2.5 Research Scope The research addresses the creati on of a framework for data collection on three types of service providers, the analysis of survey responses to describe the differenc es among the providers, and the develop ment of a residency program model to determine th e optimal skill mix Survey methodology is used for the data collection phase of the research, and the findings and resulting conclusions are made with respect to residents and alternative health professionals in Internal Medicine residency programs.

PAGE 21

11 CHAPTER THREE METHODOLOGY 3.1 Introduction Residents g ain proficiency in advanced medical training through repetitive performance of routine tasks and exposure to complex cases. Other medical personnel would undertake these necessary tasks if residents were unavailable, distinguishing task completion from ed ucation. This leads to the notion that residents responsibilities can be categorized as work or study. We propose that under the 80 hour weekly limit rule instituted by ACGME, 40 hours is allotted to work and 40 hours is allotted to study. The product ive work done by the residents may be compensated separately. As the residents progress through the residency program, they may achieve certification to perform various procedures and thus acquire the status of a professional. In our study, we compare the proficiency of a second year resident (PGY2), a Physician Assistant (PA) with one year of experience, and a Nurse Practitioner (NP) with one year of experience. The components of the research design are presented in Figure 3.1, beginning with the probl em statement and ending with projected contributions.

PAGE 22

12 Problem: Healthcare costs are spiralling upward in the United States. A portion of this cost is an entitlement paid by Medicare for the training of physicians in teaching hospitals. There is much debate about the financial model that should be used to support physicians-in-training. Medicare finances 75% of cost of GME for training purposes. Sub-Problem: There is a lack of knowledge regarding the amount contributions that residents make to the teaching hospitals where they undergo training. Hypothesis: Residents perform valuable work in the teaching hospitals where they undergo training. The training that residents undergo and the work performed are confounded. Method to test hypothesis: Compare the interchangeability of health care providers (residents, PA's, NP's) for a finite number of tasks to quantify the work contributions of residents in a teaching hospital setting. Assumptions: Work is necessary (ie. must be performed by some entity/resource/provider). There is a 50%-50% split between valuable work and training/learning/study of the time residents spend at the teaching hospitals Residents (Second Yr), Physician Assistants (1 Yr Exp), Nurse Practitioners (1 Yr Exp) are comparable. Instruments: Survey questionnaire. Aggregate patient data categorized per task. Salary data for each health care provider from a teaching hospital. Tools: Statistical Analysis (Descriptive Statistics, Analysis of Variance (ANOVA)). Mathematical Programming (LP). Contributions: Demonstrate that residents perform productive work at teaching hospitals. Provides a model that considers resident work capabilities to determine staffing needs in teaching hospitals. Establishes a foundation for many questions regarding skill mix to be explored. Figure 3.1: Re eng ineering GME Research Design The analytical framework utilized to undertake this research endeavor was the survey approach, combined with the quantitative tools that are used exte nsively in experimental design. 3.2 The Rationale of the Survey Method The survey is a system for collecting information from or about people to describe, compare, or explain their knowledge, attitudes or behavior. To establish the necessary rigor, sev en components must be included in the survey method: setting

PAGE 23

13 objectives for information collection, survey design, preparing a valid and reliable instrument, administering the survey, survey data management and analysis, results reporting [24]. As outline d in Chapter One, the research questions relating to this work are: 1) to discover if residents perform productive work by exploring the interchangeability of different types of service providers in an academic health center/teaching hospital, and 2) to re commend an optimal skill mix for a residency program based on the results of a model constructed from data collection and analysis. The proposition, primary data analysis strate gy, and hypotheses are provided in Figure 3.2.

PAGE 24

14 Proposition 1 The proficiency of residents, physician assistants and nurse practitioners are measurable and can be equated for specific procedures/tasks. ANOVA For 5 tasks measured in 3 dimensions: 15 hypotheses. Hypothesis 1 For Task 1 along Dimension 1: Proficiency of residents, physician assistants and nurse practitioners are equal ANOVA H 0 : R = PA = NP ; H 1 : R ? PA ? NP ; Hypothesis 2 For Task 1 along Dimension 2: Proficiency of residents, physician assistants and nurse practitioners are equal ANOVA H 0 : R = PA = NP ; H 1 : R ? PA ? NP ; Hypothesis 3 For Task 1 along Dimension 3: Proficiency of residents, physician assistants and nurse practitioners are equal ANOVA H 0 : R = PA = NP ; H 1 : R ? PA ? NP ; Hypothesis 4 For Task 2 along Dimension 1: Proficiency of residents, physician assistants and nurse practitioners are equal ANOVA H 0 : R = PA = NP ; H 1 : R ? PA ? NP ; Hypothesis 5 For Task 2 along Dimension 2: Proficiency of residents, physician assistants and nurse practitioners are equal ANOVA H 0 : R = PA = NP ; H 1 : R ? PA ? NP ; Hypothesis 6 For Task 2 along Dimension 3: Proficiency of residents, physician assistants and nurse practitioners are equal ANOVA H 0 : R = PA = NP ; H 1 : R ? PA ? NP ; Hypothesis 7 For Task 3 along Dimension 1: Proficiency of residents, physician assistants and nurse practitioners are equal ANOVA H 0 : R = PA = NP ; H 1 : R ? PA ? NP ; Hypothesis 8 For Task 3 along Dimension 2: Proficiency of residents, physician assistants and nurse practitioners are equal ANOVA H 0 : R = PA = NP ; H 1 : R ? PA ? NP ; Hypothesis 9 For Task 3 along Dimension 3: Proficiency of residents, physician assistants and nurse practitioners are equal ANOVA H 0 : R = PA = NP ; H 1 : R ? PA ? NP ; Hypothesis 10 For Task 4 along Dimension 1: Proficiency of residents, physician assistants and nurse practitioners are equal ANOVA H 0 : R = PA = NP ; H 1 : R ? PA ? NP ; Hypothesis 11 For Task 4 along Dimension 2: Proficiency of residents, physician assistants and nurse practitioners are equal ANOVA H 0 : R = PA = NP ; H 1 : R ? PA ? NP ; Hypothesis 12 For Task 4 along Dimension 3: Proficiency of residents, physician assistants and nurse practitioners are equal ANOVA H 0 : R = PA = NP ; H 1 : R ? PA ? NP ; Hypothesis 13 For Task 5 along Dimension 1: Proficiency of residents, physician assistants and nurse practitioners are equal ANOVA H 0 : R = PA = NP ; H 1 : R ? PA ? NP ; Hypothesis 14 For Task 5 along Dimension 2: Proficiency of residents, physician assistants and nurse practitioners are equal ANOVA H 0 : R = PA = NP ; H 1 : R ? PA ? NP ; Hypothesis 15 For Task 5 along Dimension 3: Proficiency of residents, physician assistants and nurse practitioners are equal ANOVA H 0 : R = PA = NP ; H 1 : R ? PA ? NP ; Figure 3.2: Propositi on and Data Analysis Strategy There are four types of survey instruments: self administered questionnaires, interviews, structured record interviews, and structured observations [24]. S elf administered questionnaires were the type of survey instrument uti lized in this research and a sample is shown in Appendix A. 3.3 Survey Instrument Elements This research required expert opinions on the proficiency of the three provider options (resident, PA, NP) in performing a set of tasks/procedures. The survey

PAGE 25

15 inst rument was developed, with proficiency measured on a visual analog scale from 0 100 with 100 representing the proficiency of an attending physician. The tasks/procedures relate to five evaluation and management (E/M) codes 3 that are defined in Appen dix D. The task /procedure titles follow: Inpatient Admission Inpatient Consultation Emergency Department Services Critical Care Evaluation Discharge Day Management [Hospital Discharge Services] The proficiency of each task was measured along the dimension s of: Time (to completion) Quality (of outcome) Supervision (level required) o Independence was specifically measured in the questionnaire to maintain congruity in direction of increasing performance on the scale. All Internal Medicine programs in the Unit ed States and all residency programs at the USF College of Medicine were surveyed. The survey instruments were distributed through the Graduate Medical Education (GME) Office at the USF College of Medicine. 3 The Evaluation and Management codes were first introduced in the 1992 CPT (C urrent Procedural Terminology), and were jointly developed by the AMA (American Medical Association) and HCFA (Health Care Financing Administration). The mutual goal was to provide Physicians and claim reviewers with advice on how to prepare or review docu mentation for evaluation and management services, and to increase accuracy and consistency in reporting levels of service furnished [11].

PAGE 26

16 Completed surveys were returned to the GME Offi ce. Surveys were distributed on a second occasion to the Interna l Medicine programs to improve participation. Data related to the set of tasks/procedures indicated earlier was obtained from the USF Physicians Group for Tampa General Hospital and Moffitt C ancer Center by specifying the pertinent codes. Specifically, the number of tasks/procedures performed at each of these locations was provided on a quarterly basis for the one year. Also, salary information for each service provider type was provided. 3.3.1 Int ernal and External Validity A descriptive factorial design was selected as the basis of the survey method. In this design the dependent variable was the program directors expert opinions of provider proficiency. The independent variables or factors were: 1) provider (resident, physician assistant, or nurse practitioner), 2) task (inpatient admission, inpatient consultation, emergency department services, critical care evaluation, or discharge day management), 3) Dimension (time, quality, level of supervis ion). This type of survey design has potential for invalidity in the following respects: internal (selection), and external (interactive effects of selection, reactive effects of testing, reactive effects of innovation) [25]. With regard to internal valid ity issues, selection of the survey participants (i.e. program directors of Internal Medicine residency programs) was based on the criteria that only they possessed the knowledge and organizational responsibility to rate the proficiency of each provider. A dditionally, each program director had an equal, nonzero chance of participating because all program directors were surveyed. Regarding external validity issues: interactive effects of selection were not created due to the descriptive/non experimental natu re of the survey

PAGE 27

17 design; reactive effects of testing were not displayed because no pre measures were done, therefore no portion of the participant population was sensitized; reactive effects of innovation were not created leading to uncharacteristic behavi or due to the descriptive/non experimental nature of the survey design. 3.3.2 Data Analysis Tools Data analysis is defined as the examination, categorization, tabulation, or otherwise recombination of evidence to address the initial propositions of a study [9 ]. Non parametric and parametric statistical tests were used as data analysis tools on the survey results. The proficiency values were treated as continuous variables, based on the use of a visual analog scale for collection purposes. The independent varia bles (service provider, task, and dimension) were treated as categorical variables, which require no special treatment to generalize known properties of regression to models of analysis of variance and covariance [12]. The Kruskal Wallis test can be appli ed in the one factor ANOVA case. It is a non parametric test for the situation where the ANOVA normality assumptions may not apply, and was used in this work to test the difference in means independent of the normality assumptions associated with the ANOVA tests [14] L et n i ( i = 1, 2,, k ) represent the sample sizes for each of the samples ( k groups) in the data. For the service provider and dimension factors, k = 915; for the task factor, k = 549. Next, rank the combined sample. Then compute R i = the sum o f the ranks for group i Then the Kruskal Wallis test statistic is: ) 1 ( 3 ) 1 ( 12 1 2 + + = = n n R n n H k i i i (1)

PAGE 28

18 This statistic, shown in equat ion 1, approximates a chi square distribution with k 1 degrees of freedom if the null hypothesis of equal populations is true. The minimum sample size must be at least 5 for the approximation to be valid. We reject the null hypothesis of equal population m eans if the test statistic H is greater than c 2 a k 1 where c 2 is the chi square function. A more formal description is shown in Table 3.1. Table 3.1: Formal representation of Kruskal Wallis Test H 0 : k m m m = = = ... 2 1 H A : j i m m ; for at least one set of i and j Test Statistic: ) 1 ( 3 ) 1 ( 12 1 2 + + = = n n R n n H k i i i Significance Level: a typically set to 0.05. Critical Region: H > c 2 a k 1 where c 2 is the chi square function. Conclusion: R eject the null hypothesis if the test statis tic lies in the critical region Analysis of variance (ANOVA) is used to distinguish the main and interaction effects of categorical independent variables on an interval dependent variable [14]. The general fac torial design and nested factorial designs were used as data analysis tools as shown in Figures 3.3 and 3.4.

PAGE 29

19 y ijkl = m + t i + b j + g k + ( t b ) ij + ( t g ) ik + ( b g ) jk + ( t b g ) ijk + e ijkl where, i = 1, 2,, a j = 1, 2,, b k = 1, 2,, c l = 1, 2,, n i = service providers; a=3; 1 = residents 2 = physician assistants 3 = nurse practitioners j = tasks (procedures); b=5; 1 = Inpatient Admission 2 = Inpatient Consultation 3 = Emergency Department Services 4 = Critical Care Evaluation 5 = Discharge Day Management [Hospital Discharge Services] k = dimensions; c=3; 1 = time 2 = quality 3 = level of supervision l = replicates; n=61; Figure 3.3: General Factorial Design

PAGE 30

20 y ijkl = m + t i + b j + g k(i) + ( t b ) ij + ( b g ) jk(i) + e ijkl where, i = 1, 2,, a j = 1, 2,, b k = 1, 2,, c l = 1, 2,, n i = service providers; a=3; 1 = residents 2 = physician assistants 3 = nurse practitioners j = tasks (procedures); b=5; 1 = Inpatient Admission 2 = Inpatient Consultation 3 = Emergency Department Service s 4 = Critical Care Evaluation 5 = Discharge Day Management [Hospital Discharge Services] k = dimensions; c=3; 1 = time 2 = quality 3 = level of supervision l = replicates; n=61; Figure 3.4: Nested Factorial Design 3.4 Construction of the Model 3.4.1 Introduction The results of the data analysis were utilized as parameters for the relevant variables necessary to build a mathematical programmin g model. The linear model served as a d ecision making tool to determine the optimal combination of service provider types in an Internal Medicine residency prog ram. The service providers being evaluated we re second year residents (R), physician assistants with one year of

PAGE 31

21 experience (PA), and nurse practitioners with o ne year of experience (NP). P hysician assis tants and nurse practitioners are independently certified professionals hired by hospitals (teaching and non teaching) to perform specific tasks. Five of these tasks, which ha ve previously been identified, we re utilized in the data collection portion of th is work. We make the assumption that service providers contribute an equal proportion of their time to tasks that are not included in the model. 3.4.2 LP Formulation The objective of the model is to maximize proficiency while meeting the cost and other requirements of the internal medicine residency program. Values of combined proficiency (P C ) for each type of service provider were extracted from the analysis of the data collected. The combined proficiency values are the mean of the proficiencies across three dimensions (time, quality, level of supervision) for each type of service provider. Consequently, the objective function of the LP model follows : Max P C = P C1 R + P C2 PA + P C3 NP (2) The constraints of the model we re constructed from various characteristics of a functioning Internal Medicine Re sidency Program at a healthcare system S 1 R + S 2 PA + S 3 NP Budget per year (3) Equation 3 represents the financial constraints of the residency program. S i where i=1,,3, represent the salarie s of the service providers including benefits. R Maximum # of resident slots (4) R 4 AP (5) Equation 4 and 5 represent the regulatory and accreditation constraints. Equation 4 denotes the number of resident slots allocated to a residenc y program that Medicare is

PAGE 32

22 willing to fund. Equation 5 characterizes the ratio of residents to attending physicians (AP) that are allowed by the accreditation body for residency programs (ACGME). There is no specific number indicated for a general Internal Medicine residency program. A majority of general residency programs have a 4:1, resident to attending physician ratio. Some specialties of Internal Medicine have a 1.5:1, resident to attending physician ratio. U 1 R + U 2 PA + U 3 NP max # of tasks perfor med per year (6) Equation 6 represents the physical capacity constraints. U i where i=1,,3, denotes the utility of the service provider that is computed by dividing the number of minutes worked in a year by the result of the division of the standard tim e an attending physician takes to perform a procedure/task (30 minutes) by the time proficiency of that provider. An example will illustrate: Number of minutes worked in a year = (40 hrs/wk 50 wks 60 mins/hr) = 120, 000 minutes Proficiency of resid ent to attending physician = 75/100 = 0.75 Standard time an attending physician takes to perform a procedure = 30 minutes U 1 = 120000 / (30 / 0.75) = 3000 procedures/yr R 0, PA 0, NP 0 (7) Equation 7 represents the non negativity constraints, so that a feasible solution would not be returned with a negative value for any type of service provider. Additionally, the variables for residents, physician assistants and nurse practition ers were set to return general integer values.

PAGE 33

23 3.5 Validation of the Model The model was validated by utilizing one set of data from a healthcare system in Tampa Florida to build the model, then testing the model outcomes using a data set from another ac ademic health center

PAGE 34

24 CHAPTER FOUR RESULTS AND DISCUSSI ON 4.1 Introduction The process of data collection from Program Directors of Internal Medicine in the United States, and all residency programs at the USF College of Medicine was completed over an e ight (8) month period. During the data collection period, the Internal Medicine Program Directors were surveyed on two occasions and all other residency programs at USF were surveyed once. The second survey to the Internal Medicine residency programs was sent to the non respondent Program Directors, in order to generate increased participation. The survey instrument was sent to 389 Internal Medicine Programs, 98 (25.2%) total responses were received with 61 (15.7%) responses completed. At the USF College of Medicine, the survey instrument was sent to 45 residency programs (including specialties, and sub specialties) with 11 (24.4%) total responses received and 5 (11.1%) responses completed. These responses were analyzed using statistical tests, specifical ly the Kruskal Wallis and Analysis of Variance (ANOVA) tests to support the research proposition. 4.2 Analysis and Verification of the Research Proposition The research proposition statement follows: The proficiency of residents, physician assistants, an d nurse practitioners are measurable and can be equated for specific tasks/procedures.

PAGE 35

25 To support this proposition a total of fifteen (15) hypotheses were made: one hypothesis for each of the five (5) tasks across the three (3) dimensions. For example, the hypothesis for Inpatient Admission is that each service provider has the same level of proficiency in terms of time, quality, and level of supervision required. Table 4.1 shows the summary of survey results with the mean, median and mode values for each t ype of service provider per task and dimension. Table 4.2 shows the descriptive statistics for the proficiency variable in the complete data set for the Internal Medicine residency programs. Table 4.2 Descriptive Statistics: Proficiency Variable N Mean Median TrMean StDev SE Mean Proficiency 2745 61.185 64.000 62.234 21.380 0.408 Variable Minimum Maximum Q1 Q3 Proficiency 0.000 100.000 50.000 75.000 Survey Question R PA NP R PA NP R PA NP Time Task 1 71 58 56 75 62 56 75 63 50 Time Task 2 64 49 47 65 50 50 75 50 50 Time Task 3 69 58 55 73 63 53 75 75 50 Time Task 4 68 48 45 69 50 47 50 50 50 Time Task 5 73 67 68 75 71 71 75 75 75 Quality Task 1 79 61 59 76 63 56 75 50 50 Quality Task 2 74 52 51 75 52 50 75 50 50 Quality Task 3 74 59 56 75 63 59 75 75 50 Quality Task 4 76 50 48 75 50 50 75 50 50 Quality Task 5 77 70 72 76 75 75 75 75 75 Supervision Task 1 76 55 53 75 56 54 75 50 50 Supervision Task 2 71 48 46 75 50 50 75 53 50 Supervision Task 3 72 54 51 75 54 50 75 50 50 Supervision Task 4 70 45 42 75 47 47 75 25 25 Supervision Task 5 76 69 69 75 74 74 75 75 75 Grand Mean/Median/Mode 73 56 55 75 58 54 75 50 50 mean median mode Table 4. 1 Summary of Survey Results

PAGE 36

26 4.2.1 Sam ple Size With a response rate of 61 from a population of 389 Internal Medicine Program Directors surveyed, the confidence interval was computed as 11.54 (~ 12) with a confidence level of 95%. For the USF College of Medicine residency program responses of 5 from a population of 45, the confidence interval was computed as 41.79 (~ 42) with a confidence level of 95%. 4.2.2 Univariate Analyses Internal Medicine The Kruskal Wallis test was used to evaluate the level of significance for proficiency in terms of service provider type, task, and dimension respectively. Tables 4.3, 4.4 and 4.5 show the results of each test. Table 4.3 Kruskal Wallis Test: Proficiency versus Provider Kruskal Wallis Test on Proficiency Provider N Median Ave Rank Z 1 915 75.00 1812.2 20.53 2 915 58.00 1184.5 8.81 3 915 54.00 1122.3 11.72 Overall 2745 1373.0 H = 424.34 DF = 2 P = 0.000 H = 425.58 DF = 2 P = 0.000 (adjusted for ties) The service providers are statistically significantly different from each other with a p value of 0.000, with residents (provider 1) having the highest ranking followed by physician assistants and nurse practitioners in descending order.

PAGE 37

27 Table 4.4 Kruska l Wallis Test: Proficiency versus Task Kruskal Wallis Test on Proficiency Task N Median Ave Rank Z 1 549 65.00 1421.6 1.61 2 549 56.00 1160.6 7.02 3 549 63.00 1349.1 0. 79 4 549 54.00 1152.6 7.28 5 549 75.00 1781.1 13.49 Overall 2745 1373.0 H = 230.01 DF = 4 P = 0.000 H = 230.69 DF = 4 P = 0.000 (adjusted for ties) The tasks are statistically significantly d ifferent from each other with a p value of 0.000, with Discharge Day Management [Hospital Discharge Services] (task 5) having the highest ranking followed by Inpatient Admission, Emergency Department Services, Inpatient Consultation and Critical Care Evalu ation in descending order. Table 4.5 Kruskal Wallis Test: Proficiency versus Dimension Kruskal Wallis Test on Proficiency Dimension N Median Ave Rank Z 1 915 63.00 1283.2 4.20 2 915 67.00 1474.2 4.73 3 915 64.00 1361.6 0.53 Overall 2745 1373.0 H = 26.85 DF = 2 P = 0.000 H = 26.93 DF = 2 P = 0.000 (adjusted for ties) The dimensions are statistically significantly different from each other with a p value o f 0.000, with quality (dimension 2) having the highest ranking followed by level of supervision and time in descending order. 4.2.3 Multivariate Analyses Internal Medicine The Analysis of Variance (ANOVA) tests were performed to analyze the simultaneous effect of all of the factors on proficiency. The validity of the assumptions required for the ANOVA to be an exact test of the hypothesis of no difference in means was checked. These assumptions are: 1) the observations are adequately described by the

PAGE 38

28 fixed ef fects model in Figures 3.3 and 3.4, and 2) the errors are normally independently distributed with mean zero and constant but unknown variance s 2 [NID (0, s 2 )]. The normality assumption was checked by plotting a histogram of the residuals, which resulted in a plot similar to a sample from a normal distribution centered at zero indicating normality as shown in Figure 4.1. Additionally, as depicted in Figures 4.2 and 4.3 normal probability plots of the residuals and the raw data were constructed with straig ht lines approximated in both plots indicating normality. Figure 4.1 Histogram of Residuals R e s i d u a l F r e q u e n c y 3 5 0 1 7 5 0 0 1 7 5 3 5 0 5 2 5 7 0 0 2 0 0 1 5 0 1 0 0 5 0 0 H i s t o g r a m o f t h e R e s i d u a l s ( r e s p o n s e i s P r o f i c i e n c y )

PAGE 39

29 A v e r a g e : 6 1 1 8 5 1 S t D e v : 2 1 3 8 0 1 N : 2 7 4 5 K o l m o g o r o v S m i r n o v N o r m a l i t y T e s t D + : 0 0 3 9 D : 0 0 7 8 D : 0 0 7 8 A p p r o x i m a t e P V a l u e < 0 0 1 0 5 0 1 0 0 0 0 1 0 1 0 5 2 0 5 0 8 0 9 5 9 9 9 9 9 P r o b a b i l i t y P r o f i c i e n c y N o r m a l P r o b a b i l i t y P l o t Figure 4.2 Normal Probability Plot for Residuals 0 5 0 1 0 0 1 5 1 0 2 0 3 0 4 0 5 0 6 0 7 0 8 0 9 0 9 5 9 9 D a t a P e r c e n t A D C o r r e l a t i o n 3 1 2 1 0 9 7 9 G o o d n e s s o f F i t N o r m a l P r o b a b i l i t y P l o t f o r P r o f i c i e n c y L S X Y E s t i m a t e s 9 5 % C I M e a n S t D e v 6 1 1 8 5 1 2 0 9 5 1 0 L S X Y E s t i m a t e s Figure 4.3 Normal Probability Plot o f Raw Data

PAGE 40

30 The full factorial ANOVA in Table 4.6 shows that the main factors provider, task, and dimension are statistically significant, with test statistics for each of the factors supporting the results of the univariate analyses. Table 4.6 ANOVA: P roficiency versus Provider, Task, Dimension Factor Type Levels Values Provider fixed 3 1 2 3 Task fixed 5 1 2 3 4 5 Dimension fixed 3 1 2 3 Analysis of Variance for Proficiency Source DF SS MS F P Provider 2 181686 90843 257.12 0.000 Task 4 94871 23718 67.13 0.000 Dimension 2 11070 5535 15.67 0.000 Error 2736 966675 353 Total 2744 1254302 MANOVA for Provider s = 1 m = 0.0 n = 1367.0 Criterion Test Statistic F DF P Wilk's 0.84179 257.115 ( 2, 2736) 0.000 Lawley Hotelling 0.18795 257.115 ( 2, 2736) 0.000 Pillai's 0.15821 257.115 ( 2, 2736) 0.000 Roy's 0.18795 MANOVA for Task s = 1 m = 1.0 n = 1367.0 Criterion Test Statistic F DF P Wil k's 0.91063 67.129 ( 4, 2736) 0.000 Lawley Hotelling 0.09814 67.129 ( 4, 2736) 0.000 Pillai's 0.08937 67.129 ( 4, 2736) 0.000 Roy's 0.09814 MANOVA for Dimensio n s = 1 m = 0.0 n = 1367.0 Criterion Test Statistic F DF P Wilk's 0.98868 15.667 ( 2, 2736) 0.000 Lawley Hotelling 0.01145 15.667 ( 2, 2736) 0.000 Pillai's 0.01132 15.667 ( 2, 2736) 0.000 Roy's 0.01145 The full factorial ANOVA in Table 4.7 shows that the main factors (provider, task, and dimension) are statistically significant. Also, the provider/task and provider/dimension interactions are significant. The task/dimension interaction and

PAGE 41

31 provider/task/dimension interactions are not significant. Given these results, a nested factorial ANOVA was computed in Table 4.8 with dimension nested within provider. Using this statistica l model, all factors are significant. Table 4.7 General Linear Model: Proficiency versus Provider, Task, Dimension Factor Type Levels Values Provider fixed 3 1 2 3 Task fixed 5 1 2 3 4 5 Dimensio n fixed 3 1 2 3 Analysis of Vari ance for Proficie ncy using Adjusted SS for Tests Source DF Seq SS Adj SS Adj MS F P Provider 2 181686.1 181686.1 90843.1 262.17 0.000 Task 4 94870.7 94870.7 23717.7 68.45 0.000 Dimensio n 2 11070.5 11070.5 5535.2 15.97 0.000 Provider*Task 8 24755.7 24755.7 3094.5 8.93 0.000 Provider*Dimensio n 4 3837.8 3837.8 959.5 2.77 0.026 Task* Dimensio 8 1643.5 1643.5 205.4 0.59 0.785 Provider*Task*Dimensio n 16 889.2 889.2 55.6 0.16 1.000 Error 2700 935548.5 935548.5 346.5 Total 2744 1254302.0 Table 4.8 Nested ANOVA: Proficiency versus Provider (Dimension), Task Factor Type Levels Values Provider fixed 3 1 2 3 Dimensio n (Provider) fixed 9 1 2 3 1 2 3 1 2 3 Task fixed 5 1 2 3 4 5 Analysis of Vari ance for Proficie ncy using Adjusted SS for Tests Source DF Seq SS Adj SS Adj MS F P Provider 2 181686 181686 90843 257.76 0.000 Dimensio n (Provider) 6 14908 14908 2485 7 .05 0.000 Task 4 94871 94871 23718 67.30 0.000 Error 2732 962837 962837 352 Total 2744 1254302 Tukeys test was used for all pairwise mean comparisons showing that resident s have a higher proficiency than physician assistants and nurse practitioners, as shown in Table 4.9. Additionally, physician assistants and nurse practitioners are not significantly different from each other in terms of proficiency.

PAGE 42

32 Table 4.9 Tukey P airwise Comparison for Provider Tukey 95.0% Simultaneous Confidence Intervals Response Variable Proficiency All Pairwise Comparisons among Levels of Provider Provider = 1 subtracted from: Provider Lower Center Upp er ---+ --------+ --------+ --------+ -2 18.46 16.40 14.35 ( --* -) 3 20.06 18.00 15.95 ( -* -) ---+ --------+ --------+ --------+ -18.0 12.0 6.0 0.0 Provider = 1 subtracted from: Level Difference SE of Adjusted Provider of Means Difference T Value P Value 2 16.40 0.8777 18.69 0.0000 3 18.00 0.8777 20.51 0.0000 Provider = 2 subtracted from: Level Difference SE of Adjusted Provider of Means Difference T Value P Value 3 1.597 0.8777 1.819 0.1633 Th e Tukey pairwise comparison test was also used for analyzing tasks and dimensions. As shown in Table 4.10, Discharge Day Management (task 5) had the highest proficiency, followed by Inpatient Admission (task 1) and Emergency Department Services (task 3), t hen followed by Inpatient Consultation (task 2) and Critical Care Services (task 4). Table 4.11 shows that quality (dimension 2) had the highest overall proficiency, followed by time (dimension 1) and level of supervision (dimension 3).

PAGE 43

33 Table 4.10 Tukey Pairwise Comparison for Task Tukey 95.0% Simultaneous Confidence Intervals Response Variable Proficiency All Pairwise Comparisons among Levels of Task Task = 1 subtracted from: Task Lower Center Upper -+ -------+ --------+ --------+ ---2 10.17 7.078 3.982 ( -* -) 3 5.48 2.384 0.712 ( -* -) 4 11.56 8.466 5.370 ( --* -) 5 4.96 8.053 11.149 ( -* -) -+ --------+ --------+ --------+ ---10 0 10 20 Task = 1 subtracted from: Level Difference SE of Adjusted Task of Means Difference T Value P Value 2 7.078 1.135 6.239 0.0000 3 2.384 1.135 2.102 0.2193 4 8.466 1.135 7.462 0.0000 5 8.053 1.135 7.098 0.0000 Task = 2 subtracted from: Level Difference SE of Adjusted Task of Means Difference T Value P Value 3 4.694 1.135 4.137 0.0003 4 1.388 1.135 1.223 0.7377 5 15.131 1.135 13.337 0.0000 Task = 3 subtracted from: Level Difference SE of Adjusted Task of Means Difference T Value P Value 4 6.082 1.135 5.361 0.0000 5 10.437 1.135 9.200 0.0000 Task = 4 subtra cted from: Level Difference SE of Adjusted Task of Means Difference T Value P Value 5 16.52 1.135 14.56 0.0000

PAGE 44

34 Table 4.11 Tukey Pairwise Comparison for Dimension Tukey 95.0% Simultaneous Confid ence Intervals Response Variable Proficiency All Pairwise Comparisons among Levels of Dimension Dimension = 1 subtracted from: Dimension Lower Center Upper -------+ --------+ --------+ -------2 2.296 4.3530 6.410 ( ---* ----) 3 1.864 0.1923 2.249 ( ----* ---) -------+ --------+ --------+ -------3.5 0.0 3.5 Dimension = 1 subtracted from: Level Difference SE of Adjusted Dimension of Means Difference T Value P Value 2 4.3530 0.8788 4.9534 0.0000 3 0.1923 0.8 788 0.2189 0.9739 Dimension = 2 subtracted from: Level Difference SE of Adjusted Dimension of Means Difference T Value P Value 3 4.161 0.8788 4.735 0.0000 Regression analysis was us ed to develop an empirical model for proficiency as depicted in Table 4.12, however the predictive value of the model is severely limited due to the qualitative nature of the main factors. Table 4.12 Regression Model for Proficiency The regression equatio n is Proficiency = 74.6 9.00 Provider + 1.47 Task + 0.096 Dimension Predictor Coef SE Coef T P VIF Constant 74.579 1.595 46.75 0.000 Provider 9.0005 0.4670 19.27 0.000 1.0 Tas k 1.4718 0.2696 5.46 0.000 1.0 Dimensio 0.0962 0.4670 0.21 0.837 1.0 S = 19.98 R Sq = 12.8% R Sq(adj) = 12.7% PRESS = 1097253 R Sq(pred) = 12.52% Analysis of Variance So urce DF SS MS F P Regression 3 160157 53386 133.74 0.000 Residual Error 2741 1094145 399 Total 2744 1254302

PAGE 45

35 4.2.4 Qualitative Analysis The survey instrument containe d sixteen (16) questions, with the last question being open ended in nature to elicit descriptive responses from the participants. The participants were asked to suggest any other possible factors to evaluate proficiency of the service providers in additio n to time, quality and level of supervision. There were a range of responses, however, the highest frequency responses related to patient satisfaction and professionalism. 4.3 Other Residency Programs at USF College of Medicine The data obtained from the surv ey responses at the USF College of Medicine residency programs were analyzed with the same tools utilized on the Internal Medicine residency program survey data. Assumptions of normality were tested; then the Kruskal Wallis and ANOVA methods were used to t est for differences in means among the factors; followed by the Tukey comparison test to specifically determine which factors differ significantly from each other. The full results of the analyses are shown in Appendix C. The univariate analyses (Kruskal Wallis tests) for proficiency versus each factor (provider, task, dimension) each showed a statistical significant difference from each other. The multivariate analyses supported these differences, as shown by the ANOVA test results in Table 4.13.

PAGE 46

36 Table 4.13 General Linear Model: Proficiency versus Provider, Task, Dimension (Residency Programs COMED) Factor Type Levels Values Provider fixed 3 1 2 3 Task fixed 5 1 2 3 4 5 Dimension fixed 3 1 2 3 Analysis of Variance for P roficiency, using Adjusted SS for Tests Source DF Seq SS Adj SS Adj MS F P Provider 2 4641.3 4641.3 2320.7 6.37 0.002 Task 4 3756.1 3756.1 939.0 2.58 0.039 Dimension 2 2378.8 2378.8 1189.4 3.27 0.040 Provider*Task 8 3447.2 3447.2 430.9 1.18 0.312 Provider*Dimension 4 276.2 276.2 69.1 0.19 0.944 Task*Dimension 8 1748.3 1748.3 218.5 0.60 0.777 Provider*Task*Dimension 16 898.8 898.8 56.2 0.15 1.000 Error 180 65560.8 65560.8 364.2 Total 224 82707.6 The ANOVA anal ysis indicates that the main factors (provider, task, dimension) are significantly different, however, the interaction terms are not significantly different. 4.4 LP Model Results The LP model generated optimal values for the number of service provider types g iven the constraints of the internal medicine program The inputs for the model of healthcare system 1 are shown in Table 4.14 and the model results are displayed in Table 4.15 Residents are consistently recommended in greater numbers than physician ass istants and nurse practitioners for an optimal skill mix. The output from the Lindo linear program is shown in Appendix B. Table 4.14 LP Model Inputs Healthcare System 1 Resident Physician Assistant Nurse Practitioner Salaries + Benefits 4 /yr $50, 165 $ 83, 820 $82, 550 # Tasks /provider/yr 2 840 2 320 2 240 Medicare Slots 10 Attending Physicians 5 Provider Budget/yr $14,200,000 Max # Tasks /yr 24,461 4 Indicates average provider salaries for specific healthcare system for year 2004, plus 27% benefits.

PAGE 47

37 Table 4.15 LP Model Results (Mean Proficiency) Healthcare System 1 Service Provider Resi dent Physician Assistant Nurse Practitioner Combined Proficiency Budget ($) Tasks (#) Proficiency 73 56 55 Model Skill Mix 7 1 1 622 517,525 24,440 Current Skill Mix 10 0 1 785 584,200 30,640 The model generates the combination of providers cap able of performing the tasks while maximizing combined proficiency. The binding constraint for the model is the maximum number of tasks to be performed annually. In other words, the number of tasks is the only fact that affects the level of skill mix, and the current budget is more than what is needed. Therefore, t he current skill mix row of Table 4.15 shows a higher combined proficiency, budget allocation and number of tasks performed than the model results It is important to note at this point that t he model only considers the optimal skill mix for the five tasks selected. The service providers currently in the residency program, in excess of the number generated by the model, are performing other tasks not considered in the model or pursuing educatio nal objectives (in the case of residents). The binding constraint (maximum number of tasks) was relaxed to consider the combined proficiency and budget allocation values. The upper limit of the binding constraint is determined by facility and safety issues which were outside the scope of this research. What if analyses were also performed to explore changes to the optimal solution given changes to parameters in the LP. The objective function coefficients were changed from mean values of proficiency to medi an and mode proficiency values. The summarized results are shown in Table 4.16 The combined proficiency values for both alte rnative scenarios are higher than the result obtained when using the mean proficiency

PAGE 48

38 values for each provider, primarily due to sl ightly higher value s of proficiency for the resident in each case. The maximum number of tasks is the binding constraint in each case; however, the budget allocation is l ower for the scenario with mode values because a physician assistant is replaced with a nurse practitioner who has slightly lower compensation Table 4.16 LP Model Results (Median and Mode Proficiency) Healthcare System 1 Service Provider Resident Physician Assistant Nurse Practitioner Combined Proficiency Budget ($) Tasks (#) Proficienc y (Median) 7 5 5 8 5 4 Model Skill Mix 7 1 1 6 37 517,525 24,440 Proficiency (Mode) 75 5 0 50 Model Skill Mix 7 0 2 625 516,255 24,360 The model was tested by utilizing the characteristics from another healthcare system. The new parameters for the model inputs are shown in Table 4.17, and the results for Healthcare System 2 are shown in Table 4.18. The corresponding what if analyses are summarized in Table 4.19. Healthcare System 2 has a smaller residency program than Healthcare System 1, theref ore the number of resident slots, attending physician s, provider budget, and maximum number of tasks are reduced in magnitude. Table 4.17 LP Mo del Inputs Healthcare System 2 Resident Physician Assistant Nurse Practitioner Salaries + Benefits/yr $50,165 $83,820 $82,550 # Tasks /provider/yr 2,840 2,320 2,240 Medicare Slots 4 Attending Physicians 2 Provider Budget/yr $5,600,000 Max # Tasks /yr 5,322 The results generated by the model are shown in Table 4.18 in comparison with the current sk ill mix scenario. Two providers are reduced from the total number currently

PAGE 49

39 contributing to the Internal Medicine residency program. The maximum number of tasks continues to be the binding constraint, but is relaxed to determine the combined proficiency an d budget allocation required for the current skill mix. Table 4.18 LP Model Results (Mean Proficiency) Healthcare System 2 Service Provider Resident Physician Assistant Nurse Practitioner Combined Proficiency Budget ($) Tasks (#) Proficiency 73 56 55 Model Skill Mix 1 1 0 129 133,985 5,160 Current Skill Mix 4 0 0 292 200,660 11,360 In Table 4.19, t he alternative scenarios used for Healthcare System 2 result ed in identical budget allocations and number of tasks completed because the skill mix generated by the model remained the same in both cases. The combined proficiency for the scenario with median values is higher than for the original model results using mean values because of the higher proficiency values for residents and physician assis tants. The higher proficiency value for residents in the scenario with mode values is not sufficient to offset the lower proficiency of physician assistants in that scenario. Table 4.19 LP Model Results (Median and Mode Pr oficiency) Healthcare System 2 Service Provider Resident Physician Assistant Nurse Practitioner Combined Proficiency Budget ($) Tasks (#) Proficiency (Median) 75 58 54 Model Skill Mix 1 1 0 133 133,985 5,160 Proficiency (Mode) 75 50 50 Model Skill Mix 1 1 0 125 133,985 5 ,160 4.5 Discussion The findings of the analyses performed on the data collected through the surveys, and the optimal skill mix recommended by the model support the research proposition.

PAGE 50

40 The univariate and multivariate statistical tests show a signif icant difference among the types of service providers, with pairwise comparisons indicating that residents have a higher proficiency than physician assistants and nurse practitioners (with a 95% confidence level ). Also, physician assistants and nurse prac titioners are not significantly different from each other indicating their interchangeability for the tasks investigated with no difference in proficiency levels. The full factorial ANOVA showed significant interaction between provider and task, and prov ider and dimension. The provider and task interaction indicates that the effect of the provider on proficiency is dependent on the task being performed Additionally the effect of the provider on proficiency is dependent on the dimension being evaluated. The latter interaction is further supported statistically by the nested factorial analysis with dimension nested within provider and task as main factors. There is no significant interaction between task and dimension, indicating independence between thes e two factors. The results of the LP model are consistent with the factorial analys es, as residents are generated in greater numbers than the other service providers with the objective of maximizing proficiency. Additionally it should be noted that res idents have the lowest unit cost among the service providers therefore the model is validated logically by selecting the highest proficiency and lowest cost provider within the constraints of the program. The model provides value to decision makers in ho spitals by generating skill mix options based on the combined proficiency values of service providers. The binding constraint is the number of tasks to be performed; however, the flexibility of the model

PAGE 51

41 allows this constraint to be relaxed to explore scen arios unique to each residency program. In Healthcare System 1, it is observed that a phys ician assistant was added to the skill mix, but that the total number of providers decreased by 2. The what if scenarios generate d the same total number of provider s as the primary model however, a nurse practitioner replace d a physician assistant when the pr oficiencies were the same This result may be due to the model selecting the lower cost option. In Healthcare Sy stem 2, a physician assistant was also added to the skill mix, but t he total number of providers decreased by 2. The what if scenarios provide d the sa me results as the primary model for this Healthcare System. It is important to reiterate that only the five tasks investigated in this research are c onsidered in the model to determine the optimal number of service providers The seemingly excess numbers of service providers currently in the residency programs perform tasks that are not included in this work. Therefore, the assumption that each servi ce provider contributes the same proportion of his/her time to tasks not included in the model is necessary for consistency. With the research proposition supported by the above data analyses and model, it can be implied that residents, physician assista nts and nurse practitioners are interchangeable specifically for the five tasks selected in an internal medicine program. This measurable interchangeability enables hospital administrators to view residents as resources comparable to the other service prov iders, and to develop models to commercially justify their presence in a teaching hospital beyond the educational objectives of the residency program.

PAGE 52

42 The analysis of the data from the other residency programs at the USF College of Medicine yielded simil ar results to the results from the internal medicine residency programs. There was a significant difference between providers, with residents demonstrating higher proficiency than physician assistants and nurse practitioners using pairwise compari son. Addi tionally there were no significant interactions between provider and task and provider and dimension. These results are promising for the notion of generalizing the framework to model all residency programs. The next chapter summarizes the research contr ibutions, and outlines the possible future research directions from the foundation established by this work.

PAGE 53

43 CHAPTER FIVE CONCLUSIONS AND FUTU RE RESEARCH 5.1 Introduction Health care delivery in the United States has become an issue of the highest national priority, and strategies for reform are being sought in different components of the health care system. The issue of provider credentialing is a critical element in the effort to re engineer graduat e medical education, which will potentially have a signifi cant impact on the health care system T his research was undertaken to analyze the contributions of residents to a teaching hospital in terms of work, and t o provide the foundation for a new framework of provider credentialing and funding mechanism for GM E. Specifically, the goals of this work were two fold: f irst, to discover if residents perform productive work by exploring the interchangeability of different types of service providers in an academic health center/teaching hospital; and second to reco mmend an optimal skill mix for a residency program based on the results of a model constructed from the data collection and analysis. 5.2 Summary of Results The primary conclusion of this research endeavor is that residents do perform work as measured by the set of tasks utilized and the dimensions evaluated, in comparison with physician assistants and nurse practitioners. Table 5.1 shows the results of the research proposition and hypothesis tests.

PAGE 54

44 Table 5.1 Summary of R esearch Proposition and A nalysis Propo sition The proficiency of residents, physician assistants, and nurse practitioners are measurable and can be equated for specific tasks/procedures. Supported Factor Type Proficiency(Rank) Provider Resident 73 ( 1 5 ) Physician Assistant 56 (2 ) Nurse Practitioner 55 (2 ) Task Inpatient Admission 70 (2) Inpatient Consultation 63 (3) Emergency Department Services 67 (2) Critical Care Evaluation 63 (3) Discharge Day Management 73 (1) Dimension Time 65 (2) Quality 70 (1) Level of Supervision 67 (2) 5.3 Contributions of Research This research contributes to re engineering graduate medical education by providing a framework of analysis to quantify how well work is performed by service providers, by demonstrating that residents perform productive w ork at teaching hospitals, and by developing a model of an internal medicine program that generates the optimal 5 Values i n parentheses indicate relative ranking of factors. Factors of the same rank are not significantly different from each other.

PAGE 55

45 skill mix of service providers. The measurable comparison of residents to commercial providers have implications for the manner in which residen ts are viewed by teaching hospitals, and creates a framework for provider credentialing and re evaluating the GME funding mechanism. The use of linear programming as a tool to model residency programs for skill mix optimality is novel. Linear programming h as previously been utilized in the health care sector in the more traditional areas of staff scheduling and inventory management. LPs provide a level of flexibility for adaptin g to changing parameters among residency programs. 5.4 Future Research There are di rections to explore in pursuing extensions and refinements to this research endeavor. First, the framework of analysis for quantifying work performed by residents can be expanded. The tas ks/procedures investigated may be expanded to be more relevant to the specific residency program being explored. Also, the initial additions to the dimensions of evaluation can be obtained from the most frequent responses to the last question on the survey instrument: patient satisfaction and professionalism Second, pr oficiency values for each type of service provider, task and dimension can be obtained by survey methodology and/or other data collection techniques for other residency programs Third, the LP model of the internal medicine residency prog rams can be applie d to other residency programs. T his research effort took a critical first step in redefining residents role in a teaching hospital by measuring the work being performed. The efforts to extend this

PAGE 56

46 research in different directions will increase the likel ihood of re engineering graduate medical education in a manner that will positively impact the health care system.

PAGE 57

47 REFERENCES [1] Institute of Medicine. Informing the Future: Critical Issues in Health. 2nd ed Washington, DC: The National Academies Press, 2003. [2] "Highlights National Health Expenditures, 2002." 17 Sep 2004. Centers for Medicare and Medicaid Services. U.S. Department of Health and Human Services. 17 Oct 2004 . [3] Institute of Me dicine. Crossing the Quality Chasm: A New Health System for the 21st Century. Washington, DC: National Academy Press, 2001. [4] "Accreditation Council for Graduate Medical Education (ACGME)." ACGME: Accreditation Council for Graduate Medical Education. Accred itation Council for Graduate Medical Education (ACGME). 19 Oct 2004
PAGE 58

48 [11] Current procedural terminology (Standard ed. : 1998) Current procedural terminology : CPT / American Medical Association Chicago, IL : American Medical Association, c1998 [12] Powers, Daniel A., and Yu Xie. Statistical Methods for Categorical Data Analysis Orlando: Academic Press, 2000. [13] "Probability and Statistics for Engineers and Scientists", 2nd ed., Walpole and Myers, MacMillian, 1978. [14] Montgo mery, D. (2001). Design and analysis of experiments 5th ed. New York: John Wiley & Sons, Inc.. [15] Earnshaw, Stephanie R., and Susan L. Dennett. "Integer/Linear Mathematical Programming Models." Pharmocoeconomics 21.12 (2003): 839 851. [16] Grunow, Martin, Hans Otto Gunther, and Gang Yang. "Development of a Decision Support Model for Scheduling Clinical Studies and Assigning Medical Personnel." Health Care Management Science 7 (2004): 305 317. [17] Jaumard, Brigitte, Frederic Semet, and Tsevi Vovor. "A generalized li near programming model for nurse scheduling." European Journal of Operational Research 107 (1998): 1 18. [18] Crain, Marion. "The Transformation of the Professional Workforce." Chicago Kent Law Review 79 (2004): 543 615. [19] Knapp, Richard M. "Complexity and Unce rtainty in Financing Graduate Medical Education." Academic Medicine 77 (2002): 1076 1083. [20] Abrass, Christine K., Ruth Ballweg, Margaret Gilshannon, and John B. Coombs. "A Process for Reducing Workload and Enhancing Residents' Education at an Academic Medic al Center." Academic Medicine 76 (2001): 798 805. [21] Buchanan, James, and Mario R. Dal Poz. "Skill mix in the health care workforce: Reviewing the evidence." Bulletin of the World Health Organization 80.7 (2002): 575 580. [22] "Industry Occupation Employment Mat rix: Occupation Report." U.S. Department of Labor Bureau of Labor Statistics U.S. Bureau of Labor Statistics. 16 May. 2005
PAGE 59

49 No&ResortButton=No&Base=2002&Proj=2012&SingleSelect=29107104 11&Type=Occupation&Number=10>. [23] "Engineering the Delivery of Health Care: Priorities for Application and Research ." Engineering the Delivery of Health Care: Priorities for Application and Resea rch National Academy of Engineering of the National Academies. 16 May. 2005 . [24] Fink, Arlene. The Survey Handbook 2nd ed. Thousand Oaks: Sage Publications, 2003 [25] Fink, Arlene. How to Design Survey Studies 2nd ed. Thousand Oaks: Sage Publications, 2003. [26] Grant, Stan, Tom Aitchison, Esther Henderson, and Jim Christie. "A Comparison of the Reproducibility and the Sensitivity to Change of Visual Analogue Scales, Borg Scales, and Likert Sc ales in Normal Subjects During Submaximal Exercise*." Chest 116.5 (1999): 1208 1217. [27] Guyatt, G H, M Townsend, L B Berman, and J L Keller. "A comparison of Likert and visual analogue scales for measuring change in function.." Journal of Chronic Diseases 40 (1987): 1129 1133. [28] Torrance, George W., David Feeny, and William Furlong. "Visual Analog Scales: Do They Have a Role in the Measurement of Preferences for Health States?." Medical Decision Making 21.4 (2001): 329 334.

PAGE 60

50 APPENDICES

PAGE 61

51 Appendix A Res idency Program Director Survey Instrument Introduction / Instructions Dear Program Directors: The following 16 questions seek your assessment of the performance of three classes of healthcare providers relative to the performance of an attending physician. The healthcare providers are: 1) Residents (PGY2), 2) Physician's Assistants (1 Yr Experience), and 3) Nurse Practitioners (1 Yr Experience). Fifteen (15) of the questions will ask for your assessment based on three variables (Time, Quality, Independence) for five specific tasks directly related to E/M codes. Please indicate your responses by placing a vertical mark along the visual analog scale adjacent to the healthcare provider in question. For example: Question 1 The amount of TIME required to conduct an INPATIENT ADMISSION? [E/M Codes: 99221 99223, 99234 99236] Resident (PGY2) 0 100 Physician's Assistant (1 Yr Exp.) 0 100 Nurse Practitioner (1 Yr Exp.) 0 100 The above would indicate responses of 68 for a Resident, 62 for a Physician's Assistant, and 72 for a Nurse Practitioner on a 0 to 100 scale with an Attending Physician rated at 100. The final question is open-ended, and thus requires a written response. THANK YOU FOR YOUR VITAL CONTRIBUTION TO THIS STUDY.

PAGE 62

52 Appendix A (Continued) How would you rate the proficiency of the following healthcare providers (relative to an attending physician as 100% proficient) in: Q1. The amount of TIME required to conduct an INPATIENT ADMISSION? [E/M Codes: 99221 99223, 99234 99236] Resident (PGY2) 0 100 Physician's Assistant (1 Yr Exp.) 0 100 Nurse Practitioner (1 Yr Exp.) 0 100 Q2. The amount of TIME required to conduct an INPATIENT CONSULTATION? [E/M Codes: 99251 99255] Resident (PGY2) 0 100 Physician's Assistant (1 Yr Exp.) 0 100 Nurse Practitioner (1 Yr Exp.) 0 100 Q3. The amount of TIME required to perform EMERGENCY DEPARTMENT SERVICES? [E/M Codes: 99281 99285] Resident (PGY2) 0 100 Physician's Assistant (1 Yr Exp.) 0 100 Nurse Practitioner (1 Yr Exp.) 0 100 Q4. The amount of TIME required to perform CRITICAL CARE EVALUATION? [E/M Code: 99291] Resident (PGY2) 0 100 Physician's Assistant (1 Yr Exp.) 0 100 Nurse Practitioner (1 Yr Exp.) 0 100

PAGE 63

53 Appendix A (Continued) How would you rate the proficiency of the following healthcare providers (relative to an attending physician as 100% proficient) in: Q5. The amount of TIME required to perform DISCHARGE DAY MANAGEMENT [HOSPITAL DISCHARGE SERVICES]? [E/M Code: 99238] Resident (PGY2) 0 100 Physician's Assistant (1 Yr Exp.) 0 100 Nurse Practitioner (1 Yr Exp.) 0 100 Q6. The QUALITY of the outcome in conducting an INPATIENT ADMISSION? [E/M Codes: 99221 99223, 99234 99236] Resident (PGY2) 0 100 Physician's Assistant (1 Yr Exp.) 0 100 Nurse Practitioner (1 Yr Exp.) 0 100 Q7. The QUALITY of the outcome in conducting an INPATIENT CONSULTATION? [E/M Codes: 99251 99255] Resident (PGY2) 0 100 Physician's Assistant (1 Yr Exp.) 0 100 Nurse Practitioner (1 Yr Exp.) 0 100

PAGE 64

54 Appendix A (Continued) How would you rate the proficiency of the following healthcare providers (relative to an attending physician as 100% proficient) in: Q8. The QUALITY of the outcome in performing EMERGENCY DEPARTMENT SERVICES? [E/M Codes: 99281 99285] Resident (PGY2) 0 100 Physician's Assistant (1 Yr Exp.) 0 100 Nurse Practitioner (1 Yr Exp.) 0 100 Q9. The QUALITY of the outcome in performing CRITICAL CARE EVALUATION? [E/M Code: 99291] Resident (PGY2) 0 100 Physician's Assistant (1 Yr Exp.) 0 100 Nurse Practitioner (1 Yr Exp.) 0 100 Q10. The QUALITY of the outcome in performing DISCHARGE DAY MANAGEMENT [HOSPITAL DISCHARGE SERVICES]? [E/M Code: 99238] Resident (PGY2) 0 100 Physician's Assistant (1 Yr Exp.) 0 100 Nurse Practitioner (1 Yr Exp.) 0 100

PAGE 65

55 Appendix A (Continued) How would you rate the proficiency of the following healthcare providers (relative to an attending physician as 100% proficient) in: Q11. The level of INDEPENDENCE (capability of working without supervision) demonstrated in conducting an INPATIENT ADMISSION? [E/M Codes: 99221 99223, 99234 99236] Resident (PGY2) 0 100 Physician's Assistant (1 Yr Exp.) 0 100 Nurse Practitioner (1 Yr Exp.) 0 100 Q12. The level of INDEPENDENCE (capability of working without supervision) demonstrated in conducting an INPATIENT CONSULTATION? [E/M Codes: 99251 99255] Resident (PGY2) 0 100 Physician's Assistant (1 Yr Exp.) 0 100 Nurse Practitioner (1 Yr Exp.) 0 100 Q13. The level of INDEPENDENCE (capability of working without supervision) demonstrated in performing EMERGENCY DEPARTMENT SERVICES? [E/M Codes: 99281 99285] Resident (PGY2) 0 100 Physician's Assistant (1 Yr Exp.) 0 100 Nurse Practitioner (1 Yr Exp.) 0 100

PAGE 66

56 Appendix A (Continued) How would you rate the proficiency of the following healthcare providers (relative to an attending physician as 100% proficient) in: Q14. The level of INDEPENDENCE (capability of working without supervision) demonstrated in performing CRITICAL CARE EVALUATION? [E/M Code: 99291] Resident (PGY2) 0 100 Physician's Assistant (1 Yr Exp.) 0 100 Nurse Practitioner (1 Yr Exp.) 0 100 Q15. The level of INDEPENDENCE (capability of working without supervision) demonstrated in performing DISCHARGE DAY MANAGEMENT [HOSPITAL DISCHARGE SERVICES]? [E/M Code: 99238] Resident (PGY2) 0 100 Physician's Assistant (1 Yr Exp.) 0 100 Nurse Practitioner (1 Yr Exp.) 0 100 Q16. What other factors do you believe can be used to distinguish between a resident, physician's assistant and nurse practioner (other than time, quality, and independence)?

PAGE 67

57 Appendix B Linear Programming Model and Output Healthcare System 1 m ax 73 R + 56 PA + 55 NP subject to 50165 R + 83820 PA + 82550 NP <= 1 42 00000 R <= 10 R <= 20 2840 R + 2320 PA + 2240 NP <= 24461 R >= 0 PA >= 0 NP >= 0 end gin R gin PA gin NP OBJECTIVE FUNCTION VALUE 1) 622.0000 VARIABLE VALUE REDUCED COST R 7.000000 73.000000 PA 1.000000 56. 000000 NP 1.000000 55.000000 ROW SLACK OR SURPLUS DUAL PRICES 2) 482475.000000 0.000000 3) 53.000000 0.000000 4) 409.000000 0.000000 5) 21.000000 0.000000 6) 7.000000 0.000000 7) 1.000000 0.000000 8) 1.000000 0.000000 NO. ITERATIONS= 26

PAGE 68

58 Appendix B (Continued) Healthcare System 2 max 73 R + 56 PA + 55 NP subject to 50165 R + 83820 PA + 82550 NP <= 5600000 R <= 4 R <= 8 2840 R + 2320 PA + 2240 NP <= 5322 R >= 0 PA >= 0 NP >= 0 end gin R gin PA gin NP OBJECTIVE FUNCTION VALUE 1) 129.0000 VARIABLE VALUE REDUCE D COST R 1.000000 73.000000 PA 1.000000 56.000000 NP 0.000000 55.000000 ROW SLACK OR SURPLUS DUAL PRICES 2) 5466015.000000 0.000000 3) 3.000000 0.000000 4) 7.000000 0.000000 5) 162.000000 0.000000 6) 1.000000 0.000000 7) 1.000000 0.000000 8) 0.000000 0.0000 00 NO. ITERATIONS= 13

PAGE 69

59 Appendix C Data Analysis of Residency Programs at USF College of Medicine Univariate Analyses Kruskal Wallis Test: Proficiency versus Provider Kruskal Wallis Test on Proficiency Provider N Median Ave Rank Z 1 75 67.00 136.4 3.81 2 75 63.00 102.2 1.77 3 75 63.00 100.5 2.04 Overall 225 113.0 H = 14.55 DF = 2 P = 0.001 H = 14.59 DF = 2 P = 0.001 (adjusted for ties) K ruskal Wallis Test: Proficiency versus Task Kruskal Wallis Test on Proficiency Task N Median Ave Rank Z 1 45 63.00 108.7 0.50 2 45 63.00 113.5 0.06 3 45 67.00 129.0 1.84 4 45 54.00 86.8 3.01 5 45 67.00 127.0 1.61 Overall 225 113.0 H = 12.28 DF = 4 P = 0.015 H = 12.31 DF = 4 P = 0.015 (adjusted for ties) Kruskal Wallis Test: Proficiency versus Dimension Kruskal Wallis Test on Proficiency Dimension N Median Ave Rank Z 1 75 58.00 101.0 1.96 2 75 64.00 126.2 2.14 3 75 63.00 111.9 0.18 Overall 225 113.0 H = 5.65 DF = 2 P = 0.059 H = 5.66 DF = 2 P = 0.059 (adjusted for ties)

PAGE 70

60 Appendix C (Continued) Multivariate Analyses A p p r o x i m a t e P V a l u e < 0 0 1 D + : 0 0 8 1 D : 0 1 5 4 D : 0 1 5 4 K o l m o g o r o v S m i r n o v N o r m a l i t y T e s t N : 2 2 5 S t D e v : 1 9 2 1 5 4 A v e r a g e : 5 6 9 5 5 6 9 0 8 0 7 0 6 0 5 0 4 0 3 0 2 0 1 0 9 9 9 9 9 9 5 8 0 5 0 2 0 0 5 0 1 0 0 1 P r o b a b i l i t y P r o f i c i e n c y N o r m a l P r o b a b i l i t y P l o t General Linear Model: Proficiency versus Provider, Task, Dimension (Reside ncy Programs COMED) Factor Type Levels Values Provider fixed 3 1 2 3 Task fixed 5 1 2 3 4 5 Dimension fixed 3 1 2 3 Analysis of Variance for Proficiency, using Adjusted SS for Tests Source DF Seq SS Adj SS Adj MS F P Provider 2 4641.3 4641.3 2320.7 6.37 0.002 Task 4 3756.1 3756.1 939.0 2.58 0.039 Dimension 2 2378.8 2378.8 1189.4 3.27 0.04 0 Provider*Task 8 3447.2 3447.2 430.9 1.18 0.312 Provider*Dimension 4 276.2 276.2 69.1 0.19 0.944 Task*Dimension 8 1748.3 1748.3 218.5 0.60 0.777 Provider*Task*Dimension 1 6 898.8 898.8 56.2 0.15 1.000 Error 180 65560.8 65560.8 364.2 Total 224 82707.6

PAGE 71

61 Appendix C (Continued) Tukey Pairwise Comparison for Provider Tukey 95.0% Simultaneous Confidence Inte rvals Response Variable Proficie All Pairwise Comparisons among Levels of Provider Provider = 1 subtracted from: Provider Lower Center Upper ----+ --------+ --------+ --------+ 2 16.76 9.400 2.040 ( ---------* --------) 3 17.21 9.853 2.493 ( ---------* --------) ----+ --------+ --------+ --------+ 14.0 7.0 0.0 7.0 Pro vider = 1 subtracted from: Level Difference SE of Adjusted Provider of Means Difference T Value P Value 2 9.400 3.117 3.016 0.0082 3 9.853 3.117 3.162 0.0052 Provider = 2 subtracted from: Level Difference SE of Adjusted Provider of Means Difference T Value P Value 3 0.4533 3.117 0.1455 0.9884

PAGE 72

62 Appendix C (Continued) Tukey Pairwise Comparison fo r Task Tukey 95.0% Simultaneous Confidence Intervals Response Variable Proficie All Pairwise Comparisons among Levels of Task Task = 1 subtracted from: Task Lower Center Upper --------+ --------+ --------+ -----2 9.54 1.556 12.651 ( -------* --------) 3 6.38 4.711 15.807 ( -------* -------) 4 18.05 6.956 4.140 ( -------* -------) 5 7.63 3.467 14.562 ( -------* -------) --------+ --------+ --------+ ------12 0 12 Task = 1 subtracted from: Level Difference SE of Adjusted Task o f Means Difference T Value P Value 2 1.556 4.023 0.387 0.9952 3 4.711 4.023 1.171 0.7679 4 6.956 4.023 1.729 0.4190 5 3.467 4.023 0.862 0.9105 Ta sk = 2 subtracted from: Level Difference SE of Adjusted Task of Means Difference T Value P Value 3 3.156 4.023 0.784 0.9349 4 8.511 4.023 2.115 0.2181 5 1.911 4.023 0.475 0.9895 Task = 3 subtracted from: Level Difference SE of Adjusted Task of Means Difference T Value P Value 4 11.67 4.023 2.900 0.0338 5 1.24 4.023 0.309 0.9980 Task = 4 subtracted from: Level Difference SE of Adjusted Task of Means Difference T Value P Value 5 10.42 4.023 2.590 0.0764

PAGE 73

63 Appendix C (Continued) Tukey Pairwise Comparison f or Dimension Tukey 95.0% Simultaneous Confidence Intervals Response Variable Proficiency All Pairwise Comparisons among Levels of Dimension Dimension = 1 subtracted from: Dimension Lower Center Upper ----+ -------+ --------+ --------+ 2 0.600 7.960 15.32 ( -------* -------) 3 3.614 3.747 11.11 ( --------* -------) ----+ --------+ --------+ --------+ 8.0 0.0 8.0 16.0 Dimension = 1 subtracted from: Level Difference SE of Adjusted Dimension of Means Difference T Value P Value 2 7.960 3.1 17 2.554 0.0307 3 3.747 3.117 1.202 0.4534 Dimension = 2 subtracted from: Level Difference SE of Adjusted Dimension of Means Difference T Value P Value 3 4.213 3.117 1.352 0.3685

PAGE 74

64 Appendix D Definitions and E/M Codes for Tasks/Procedures Task E/M Codes Definition Inpatient Admission 99221 99223 99234 99236 Initial hospital care, per day, for the evaluation and management of a patient req uiring: A detailed/comprehensive history A detailed/comprehensive examination Medical decision making of low/moderate/high complexity Observation or inpatient hospital care for the evaluation and management of a patient including admission and discharge o n the same date requiring the above (bulleted list). Inpatient Consultation 99251 99255 Initial inpatient consultation for a new or established patient requiring: A problem focused/detailed / comprehensive history A problem focused/detailed / comprehens ive examination Medical decision making of low/moderate/high complexity Emergency Department Services 99281 99285 The provision of unscheduled episodic services to patients who present for immediate medical attention. Emergency department visit for the evaluation and maintenance of a patient requiring the above (bulleted list from Inpatient Consultation).

PAGE 75

65 Appendix D (Continued) Task E/M Codes Definition Critical Care Evaluation 99291 Critical care is the direct delivery by a physician(s) of medical c are for a critically ill or injured patient . the care of such patients involves decision making of high complexity to assess, manipulate, and support central nervous system failure, circulatory failure, shock like conditions, renal, hepatic, metabolic or respiratory failure, postoperative complications, overwhelming infection, or other vial system functions to treat single or multiple vital organ system failure or to prevent further deterioration. Discharge Day Management 99238 The time spent for fin al discharge of a patient. This includes, as appropriate, final examination of the patient, discussion of the hospital stay, instructions for continuing care to all relevant caregivers, and preparation of discharge records, prescriptions, and referral form s.


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 001670388
003 fts
005 20051216093342.0
006 m||||e|||d||||||||
007 cr mnu|||uuuuu
008 051122s2005 flu sbm s000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0001258
035
(OCoLC)62324439
SFE0001258
040
FHM
c FHM
049
FHMM
090
T56 (Online)
1 100
Elius, Ian M.
0 245
Re-engineering graduate medical education
h [electronic resource] :
b an analysis of the contribution of residents to teaching hospitals utilizing a model of an internal medicine residency program /
by Ian M. Elius.
260
[Tampa, Fla.] :
University of South Florida,
2005.
502
Thesis (M.S.I.E.)--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 75 pages.
520
ABSTRACT: According to the Institute of Medicine (IOM), the U.S. health care delivery system does not provide consistent, high-quality medical care to all people all the time. As a significant component of the health care delivery system, the state of Graduate Medical Education in the United States has prompted much analysis in recent years due to the general view that desired and actual outcomes are increasingly at variance with each other. One area of focus has been the implications of change for provider credentialing and funding of graduate medical education. With this research we test the hypothesis that residents perform valuable work in the teaching hospitals where they undergo training, to inform the issue regarding provider credentialing for residents.We developed a framework to compare second-year residents (PGY2), physician assistants with one year of experience, and nurse practitioners with one year of experience to measurably address the interchangeability of providers. Data was collected by obtaining expert opinions on the proficiency of the three provider options (resident, physician assistant, nurse practitioner) in performing a set of tasks/procedures by surveying the program directors of Internal Medicine residency programs in the United States. The other residency programs at the University of South Floridas College of Medicine were also surveyed to obtain measurable performance on the service providers.Statistical tools were used to analyze the survey responses, aggregate patient data and salary data for each provider. The data analysis and summary indicated that residents displayed higher levels of proficiency than physician assistants and nurse practitioners for the tasks investigated.
590
Adviser: Jose L. Zayas-Castro, PhD.
Co-adviser: Peter J. Fabri, MD
653
GME.
Healthcare system.
Linear programming.
Skill mix.
Proficiency.
690
Dissertations, Academic
z USF
x Industrial and Management Systems
Masters.
773
t USF Electronic Theses and Dissertations.
4 856
u http://digital.lib.usf.edu/?e14.1258