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 001680991
007 cr mnu|||uuuuu
008 051207s2004 flu sbm s000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0000606
Moore, D. Helen.
Evaluation of the prognostic criteria for medicare hospice eligibility
h [electronic resource] /
by D. Helen Moore.
[Tampa, Fla.] :
b University of South Florida,
Thesis (Ph.D.)--University of South Florida, 2004.
Includes bibliographical references.
Text (Electronic thesis) in PDF format.
System requirements: World Wide Web browser and PDF reader.
Mode of access: World Wide Web.
Title from PDF of title page.
Document formatted into pages; contains 99 pages.
ABSTRACT: This work evaluates Medicare Hospice Benefit (MHB) eligibility standards that are referenced throughout this work as either "Medicare prognostic criteria," or "Local Medical Review Policies." Following the Chapter 1 overview of prognosis in end-stage disease, association between the Medicare clinical predictors and survival outcomes in dementing, cardiovascular and cerebrovascular illnesses are described in Chapter 2. Chapter 3 examines the prognostic belief systems of multidisciplinary hospice personnel. Chapter 4 seeks to improve the predictive performance of the Medicare prognostic criteria for dementia. The fifth and final chapter critiques the Medicare prognostic criteria from conceptual, methodological, and applied perspectives and suggests related research and policy directions. The Chapter 2 sample comprised 453 medical records of terminally ill persons; Chapter 4 sample, 187 medical records.Thirty-seven hospice personnel comprised the respondent sample in the Chapter 3 study. Chapter 2 assesses the scientific validity of federally sanctioned Medicare "severe illness/end-stage illness" demarcations in three non-cancer disease catregories. Calculation of measures of predictive validity revealed striking and consistent imbalances of false negative and false positive errors across the three diagnostic categories studied, suggesting inequitable distribution of the costs and benefits of regulatory reform among public health payers, consumers and providers. Chapter 3 qualitatively examines the belief systems of experienced hospice personnel regarding physical and non-physical time-to-death influences in end-stage disease. Non-physical survival influences were believed by these expert informants to have more survival impact in non-cancer as opposed to cancer end-stage diseases, and at remote as compared to imminent death proximities.Chapter 4 demonstrates that dropping one of the three prognostic criteria for dementia (the medical complications criteria) may improve predictive validity. This finding demonstrates that, in dementing illnesses at least, functional debility may better identify 6-month survival prognosis and thus hospice eligibility, than the composite Medicare prognostic criteria. The merit of parsimony in objective definitions of terminality is implied. Chapter 5 critiques the Medicare prognostic criteria, and suggests policy alternatives that are both prognostically- and non-prognostically-based. Peripheral findings of this work and suggestions for future end-of-life research conclude the dissertation.
Adviser: Mortimer, James A.
Medicare Hospice Benefit.
x Aging Studies
t USF Electronic Theses and Dissertations.
Evaluation of the Prognostic Criteria for Medicare Hospice Eligibility by D. Helen Moore A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy School of Aging Studies College of Arts and Sciences University of South Florida Major Professor: James A. Mortimer, Ph.D. R. Clifford Blair, Ph.D. Kathleen A. Egan, MA Kay Perrin, Ph.D. Ronald Schonwetter, M.D. Brad Stuart, M.D. Date of Approval: March 16, 2004 Keywords: prognosis, terminal illness, Medicare Hospice Benefit dementia, heart, stroke Copyright 2004, D. Helen Moore
Dedicated in Honor of Mur Dorothy Helen Moore Martin, of Georgetown, Kentucky LOVELIEST of trees, the cherry now Is hung with bloom along the bough, And stands about the woodland ride Wearing white for Eastertide. Now, of my threescore years and ten, Twenty will not come again, And take from seventy springs a score, It only leaves me fifty more. And since to look at things in bloom Fifty springs are little room, About the woodlands I will go To see the cherry hung with snow. A. E. Housman, from A Shropshire Lad 1896.
Acknowledgements This dissertation research was made possible in part through a pre-doctoral stipend awarded by The Hospice of the Florida Suncoast (HFSC). Ms. Mary Labyak, HFSC President, provided research vision and facilitation at every level, as well as the invaluable collaboration of her staff, incl uding Vice President Kathleen Egan, Director Katherine Brandt, Medical Records Specialist Michael Condon, Nurse Researchers Gay Lynn Horvath and Linda Hummer and numerous instrumental and dedicated others. Homer Jack Moore, MD, offered medical consultation; other family members, including daughter Abigail Susik, Snuggler Susik-Moore, Jacob Moore, MD, Jackson E. Moore, Chris Vinsonhaler, Marian Wingo, Emanuel H. Martin IV, Michael Susik, Charles and Betty Ann Hambrick and Ruby and Martel Gilpin lent loving support and encouragement. Esteemed classmates Carla VandeWeerd, Ph.D. and Hongbin Chen, MD, PhD, were on hand throughout with assistance and wise counsel. Dr. Chen reviewed the manuscript, as did Amanda G. Smith, MD, Regina Velasco, DO, Patricia Geasa, RN, and Sandra Mutolo, MSW, LCSW. Special gratitude is due Linda Wilson Beavers, Robert Sprinkle and Yougui Wu, P h.D., for professional programming, database and statistical methods assistance. My deepest gratitude and respect is reserved for Major Professor James A. Mortimer, Ph.D., who demanded nothing less than good science.
i Table of Contents List of Tables iii Abstract vi Chapter One: Introduction 1 The Medicare Hospice Benefit 3 The Dilemma of Prognostic Accuracy 4 Clinical Factors that Af fect Prognostic Accuracy 4 Non-Clinical Factors that Affect Prognostic Accuracy 6 Approaches to End-of-life Prognostication 6 Medicare Prognostic Criteria Origin 7 Medicare Prognostic Criteria Controversy 8 Conclusion 9 Chapter Two: Evaluation of the Local Medical Review Policies for Medicare Hospice Eligibility in Advanced Demen tia, Stroke and Heart Disease 10 Abstract 10 Background 10 Methods 10 Results 10 Conclusions 11 Introduction 11 Methods 18 Results 21 Discussion 24 Chapter Three: Time-to-Death in Far Advanced Disease: The Belief Systems of Multidisciplinary Hospice Personnel 30 Abstract 30 Introduction 31 Methods 31 Research Protocol 31 The Sample 32 Focus Group Interviews 33 Data Analysis 34 Results 34 Physical Time-to-Death Beliefs of Hospice Personnel 43 Patient-specific Time-to-Death Beliefs of Hospice Personnel 44 Patient Attitudes, Cogni tions and Behaviors 44 Quality of Life 44 Stress 45 Contextual and Environmenta l Time-to-Death Beliefs of 46
ii Hospice Personnel Social Relationships 46 Caregiver Factors 47 Milieu of Care 48 Discussion 48 Limitations of Research 50 Further Research on Terminal Prognosis 50 Chapter Four: Can the Predictive Performance of the Dementia Criterion for Medicare Hospice Eligibility Be Improved? 52 Abstract 52 Introduction 52 Methods 55 Results 56 Predictive Performance of Differ ent Combinations of Medicare 58 Prognostic Criteria Discussion 58 Further Research 62 Conclusion 63 Chapter Five: Medicare Local Medical Review Policies (LMRPs): Concepts, and Consequences 64 Abstract 64 Local Medical Review Policies: Â“C linical Guidelines or Policies?Â” 64 LMRP Assumptive, Methodological, and Logistical Limitations 65 Assumptive Limitations 65 Methodological Limitations 67 Logistical Limitations 67 LMRPs: Background, Costs and Benefits 68 Policy-based Alternatives to the Medicare Prognostic Criteria for Medicare Hospice Eligibility 71 Suggestions for Systems-wide Research in Terminal Illness 71 Conclusion 71 Dissertation Conclusion 75 References 76 About the Author End Page
iii List of Tables Table 1: NHPCO general guidelin es for determining prognosis 13 Table 2: Prognostic criteria for Medicare hospice eligibil ity by diagnostic category 15 Table 3: Characteristics of the sample by diagnosis at hospice admission 22 Table 4: Case/Control comparison of sample at hospice admission 23 Table 5: Odds ratios and measures of predictive validity by LMRP diagnostic category 25 Table 6: Focus group composition, hospice personnel 33 Table 7: Hospice-related survival beliefs of hospice personnel 35 Table 8: General prognostic beliefs of hospice personnel 36 Table 9: Patient-specific be liefs of hospice personnel 37 Table 10: Physical beliefs of hospice personnel 38 Table 11: Spiritual belief s of hospice personnel 39 Table 12: Caregiver-related beliefs of hospice personnel 40 Table 13: End-of-life prognostic factors recorded by hospice personnel in a wr itten exercise 41 Table 14: Medicare prognostic criteria for dementia 53 Table 15: Characteristics of dementia pa tients on first admission to hospice care 57 Table 16: Comparison of odds ratios and measures of predictive validity for the dementia prognostic criteria: Single and composite criteria 59 Table 17: NHPCO parameters of hospice e ligibility 66 Table 18: Disease-specific disparity in Local Medical Review Policies fulfillment requirements 66 Table 19: Potential outcomes of the Local Medical Review Policies for Medicare hospice eligibility 72
iv Table 20: Health system outcomes of the prognostic criteria for Medicare hospice eligibility 73
v Evaluation of the Prognostic Criteria for Medicare Hospice Eligibility Abstract This work evaluates Medicare Hospice Benefit (MHB) eligibility standards that are referenced throughout this work as either Â“Medicare prognostic criteria,Â” or Â“Local Medical Review Policies.Â” Following the Chapter 1 overview of prognosis in end-stage disease, association between the Medicare clinical predictors and survival outcomes in dementing, cardiovascular and cerebrovascular illnesses are described in Chapter 2. Chapter 3 examines the prognostic belief systems of multidisciplinary hospice personnel. Chapter 4 seeks to improve the predictive performance of the Medicare prognostic criteria for dementia. The fifth and final chapter critiques the Medicare prognostic criteria from conceptual, methodological, and applied perspectives and suggests related research and policy directions. The Chapter 2 sample comprised 453 medical records of terminally ill persons; Chapter 4 sample, 187 medical records. Thirty-seven hospice personnel comprised the respondent sample in the Chapter 3 study. Chapter 2 assesses the scientific validity of federally sanctioned Medicare Â“severe illness/end-stage illnessÂ” demarcations in three non-cancer disease catregories. Calculation of measures of predictive validity revealed striking and consistent imbalances of false negative and false positive errors across the three diagnostic categories studied, suggesting inequitable distribution of the costs and benefits of regulatory reform among public health payers, consumers and providers. Chapter 3 qualitatively examines the belief systems of experienced hospice personnel regarding physical and non-physical time-to-death influences in end-stage
vi disease. Non-physical survival influences were believed by these expert informants to have more survival impact in non-cancer as opposed to cancer end-stage diseases, and at remote as compared to imminent death proximities. Chapter 3 highlights the enormous complexity of time-to-death influences as well as the importance of non-physical influences on duration of survival in end-stage disease. Chapter 4 demonstrates that dropping one of the three prognostic criteria for dementia (the medical complications criteria) may improve predictive validity. This finding demonstrates that, in dementing illnesses at least, functional debility may better identify 6-month survival prognosis and thus hospice eligibility, than the composite Medicare prognostic criteria. The merit of parsimony in objective definitions of terminality is implied. Chapter 5 critiques the Medicare prognostic criteria, and suggests policy alternatives that are both prognosticallyand non-prognostically-based. Peripheral findings of this work and suggestions for future end-of-life research conclude the dissertation.
1 Chapter One Introduction This work uses a variety of methods to address systemic U.S health care policy issues that concern individuals with lifethreatening, non-cancer diseases. Specifically evaluated are three sets of disease severity indicators applied in non-cancer diagnostic categories to define Medicare Hospice Benefit eligibility. Administratively known as Â“Local Medical Review PoliciesÂ” (LMRPs) and more generically as Â“Medicare prognostic criteria,Â” these clinical standards have an effect on admission, re-certification and discharge determinations for the nearly 1 in 4 Americans who die each year in hospice care settings (1). To this investigatorÂ’s best knowledge, this dissertation work comprises the single most comprehensive evaluation of the Medicare prognostic criteria to date. Results confirm previously reported findings on the predictive deficits of the Medicare prognostic criteria, and provide new details on the outcomes of these criteria in simulations of regulatory usages. The work further reveals the complex and often contradictory range of factors that can influe nce the timing of death. Further, it suggests that the Medicare prognostic criterion for dementia currently applied across the U.S to approve or deny hospice admission can be improved through simple modification. The dissertation findings may be of interest to the many parties and entities that stand to be affected by the Medicare prognostic criteria, before, during and after the actual provision of hospice care. Such parties include present and potential consumers of palliative, hospice care services (patients and their families), referring physicians,
2 organizational providers of hospice care, those who can articulate policy concerns and, optimistically, those with the power to effectuate regulatory and reimbursement change. Several factors have motivated this 4-year investigation. First, in 1999-2000, heart disease, stroke and dementia, the diagnostic foci of this study, respectively represented the first, third and eighth all-age death causes in the United States (2). Furthermore, cardiac and dementing diseases are among the top five non-cancer causes of death in hospice patients (1). Nearly 54 percent of all Medicare hospice patients served in 2002 were diagnosed with cancer upon admission, 46 percent with life-threatening diseases of non-cancer origin (1). Of all enrollees that year, 10 percent were diagnosed with heart disease, seven percent with dementia (2). The second, and perhaps most important motivating factor for research relates to the unprecedented reductions in hospice median lengths of stay that have paralleled Medicare prognostic policy instigation. As reiterated within the body of the work, the Medicare prognostic criteria may have influenced patient selection processes in favor of observably over less observably critically and terminally ill individuals, to the disadvantage of Medicareeligible patients with certain non-cancer diagnoses. Although it is unlikely that such a causal link can be empirically established, the evidence that suggests such a link is highly suggestive (1). Third, in the present era of health care cost containment, the Medicare/Medicaid program consumes an annual budget of over 2 billion dollars (3). As death in America is increasingly defined by chronic, non-cancer illnesses (4) and as the population that seeks palliative end-of-life treatment grows (1), it is important to evaluate the costs and benefits of health care policy innovation.
3 The Medicare Hospice Benefit The Medicare Hospice Benefit (MHB) provides medical, psychological, social and spiritual interventions to dying patients and uniquely includes the patientÂ’s family caregiver in the unit of care. Palliative care in hospice settings has been Medicare Part A reimbursable since 1982. Third-party reimbursement for hospice is uniquely comprehensive and nearly all-inclusive, covering home care, acute care, respite care, prescription drugs, allied therapies, and psychosocial and spiritual interventions (5). Legal requirements of eligibility for this public service are: (1.) a terminal diagnosis, i.e., a life-threatening disease for which no cure is anticipated; (2.) a limited survival prognosis, i.e. six-months or less survival assuming normal disease course, and (3.) benefit election, i.e. patient/family choice of palliative, non-curative options over regular Medicare Part A benefits (6). Currently, there are two initial 90-day benefit periods in Medicare hospice followed by an unlimited number of 6-day periods. A physician must re-certify that a patient has six months or less to live before each benefit period (1). Cancer was and is the most common diagnosis in hospice, but the proportion of cancer patients in Medicare hospice decreased from 75 percent in 1992 to 58 percent in 2000 (7). Under MHB statutory provisions, car e can be provided for up to 210 days or sometimes longer. The majority of Medicare hospice beneficiaries receive the bulk of their care in their homes from family caregivers (1). Other beneficiaries receive hospice health care services in nursing homes, hospitals or other inpatient facilities. Seven out of ten hospice patients are dependent in basic self-care skills such as bathing, dressing or eating; about 70 percent are doubly incontinent; four out of five have mobility
4 limitations; and half use oxygen (7) Without a doubt, the hospice-eligible population comprises one of the most impaired and vulnerable groups in America. The Dilemma of Prognostic Accuracy Terminally ill persons, given physiciansÂ’ certification of 6-month or less survival prognosis, represent potential MHB-eligibles. From historical (8), contemporary (9) and scientific perspectives (10), however, accurate terminal status determination is notoriously inaccurate and may remain ever so, particularly in non-cancer diseases and at more remote, 6-month proximities from death. Interestingly, it has been shown that survival in hospice care settings varies substantially according to diagnosis (11). In one landmark study, relatively longer durations of survival in non-cancer categories were observed among hospice patients with dementia and pulmonary diseases (12). Prediction of length of life remaining is one of the most complex and daunting tasks in medicine, hinging on consideration of a complex, interrelated and dynamically shifting array of contextual, patientand disease-specific factors. Reliable prognostication will remain undefined, and the thresholds of terminality ambiguous, until the determinants of death are more comprehensively understood. As an introduction to prognostic issues, the possible range of such factors is briefly reviewed. Clinical Factors that Affect Prognostic Accuracy Age, race and sex, factors linked to the timing of mortality by a large and diverse literature (13), are notably absent from Medicare prognostic formulations. Although this omission may reflect the need to avoid discrimination-based protest in national health
5 care policy, it is also likely that age, race and sex play a role in life expectancy at birth that are not significantly related to short term survival after hospice admission. Many other patient-specific factors are similarly unaccounted for by the Medicare criteria, such as the occurrence of idiosyncratic patient responses to certain hospice treatments that are ambiguously palliative (diuretics, vasodilators and ACE inhibitors) (14), and the extreme functional heterogeneity that characterizes older adult populations (15). Furthermore, although the Medicare prognostic criteria do incorporate measures of general physical debility (presence of pressure sores, 10% weight loss) and severe infection (pneumonia/septicemia/recurrent fever/urinary tract infections), they do not account for dual diagnoses, co-morbidities and intercurrent illnesses, the composite survival effect of which remains unknown. This is a particular concern given that the acknowledged cause of death in older adult populations is multifactorial (16). Prototypical death trajectories provide a good example of the biologic confounds of terminal prognosis. A short period of unm istakable deterioration typifies cancer diseases; periodic exacerbation of long-term disabled status defines end-of-life circumstances in non-cancer, chronic organ systems failures. In most cancer diagnoses, fatal decline progresses rapidly and predictably downward (17); non-cancer fatal trajectories are comparatively erratic (18). For example, end-stage dementia and cerebrovascular illnesses are prototypically marked by variable and unpredictable Â“plateaus of stability,Â” the durations of which are difficult to predict. Impaired consciousness on hospice admission is a reported risk factor for stroke mortality (19). The pattern of decline in congestive heart failure stands in sharp contrast to decline seen in dementia, a condition in which death occurs due to overwhelming physiologic failure.
6 Typically a cardiac patient appears no more ill during the weeks that immediately precede death than in previous phases of illness (9), a factor that obviously confounds prognosis. As referenced above, patients severely ill with congestive heart failure may survive for many years if offered symptom control treatments that include vasodilators (14). Non-Clinical Factors that Affect Prognostic Accuracy A positive association between ease and accuracy of survival prediction and patient nearness to death is a prognostic fundamental. This is so because, in cancer and non-cancer illnesses alike, deathÂ’s imminent approach (days, hours) is clearly marked by a cascade of organ system failures (20). Prognosis of time to death from temporal perspectives that are Â“intermediateÂ” (2-3 months) or Â“remoteÂ” (6-month) is much less straightforward, however, perhaps due in part to the influence of psychosocial and contextual factors that, according to the Chapter 3 informants, strongly influence endstage survival duration in non-cancer diseases and at more remote death proximities. As one example of psychosocial effects on health outcomes, severely impaired stroke patients with little hope for recovery tend to experience shorter survival than those with more hopeful attitudes (21). Finally, mortality in advanced chronic and severely debilitating illnesses may be determined by factors as basic as the quality of custodial care rendered or as esoteric as patient-perceived quality of life (Chapter 3).
7 Approaches to End-of-life Prognostication Beyond traditional physician clinical judgments, several models have been developed to prognosticate survival, including the Acute Physiological and Chronic Health Evaluation (APACHE) (2224) and the SUPPORT model (25-27). These tools represent a modest advancement over clinical judgment (25), but are less appropriate for use with individual patients than for research conducted with large population-based samples. They were developed for use with acutely ill hospitalized patients and constructed based on studies of patients that received standard medical therapy in traditional medical settings. For these reasons, and because 6-month survival outcome measures were generally not employed, such models do not provide viable options for hospice-based survival estimates. Regression-based studies of hospice mortality pointers are equivocal due to small sample sizes and failure to account for co-morbid and patient heterogeneity confounds (28-32). A third approach to end-of-life prognosis compares the severity of patient symptoms to pre-selected clinical points thought to possess prognostic significance (33). This final Â“threshold-basedÂ” or Â“staging theoryÂ” method was first developed in response to the need for improved cancer prognosis (34). This is the Centers for Medicare and Medicare Services (CMS) chosen approach as embodied by the Medicare prognostic criteria. Of note is the fact that the Medicare criteria are far more restrictive than are the eligibility parameters federally specified in the founding Medicare Hospice Benefit statutes (6).
8 Medicare Prognostic Criteria Origin Criteria for evaluating the timeliness of hospice enrollment for patients with endstage, non-malignant diseases were first proposed in 1993 (35) by a hospice physician and for selected non-malignant diagnoses, clinical guidelines were formally developed and published by the National Hospice and Palliative Care Organization (NHPCO) in 1995. The Medicare prognostic criteria were appropriated nearly verbatim from a second edition of the NHPCO monograph entitled Â“Guidelines for Determining Prognosis in Non-cancer Diseases,Â” commonly referenced as the Â“NHPCO GuidelinesÂ” (36). Terminality parameters as set forth in the Guidelines represent composites of previously published mortality pointers that are organized according to degenerative organ systems processes. The authors of the Guidelines clearly represent their work as a starting point for ongoing end-of-life prognostic research, with the caveat that the Guidelines should be adapted based on further research. Medicare Prognostic Criteria Controversy The appropriateness of the Medicare prognostic criteria, administratively known as Â“Local Medical Review Policies (LMRPs),Â” can and has been contested on several grounds (37). First, mounting quantitative evidence shows little association between the LMRPs and short-term survival outcomes in hospice populations (38). Second, although Medicare hospice was founded upon holistic, physical, psychosocial and spiritual treatment principles, the LMRP criteria exclude non-physical markers of disease progression. Third, although the consequences of regulatory reform have yet to be fully understood, the LMRPs appear to be disadvantageous to persons less obviously and
9 imminently terminally ill, as evidenced by a 20 percent decline in the national average length of hospice stay from 1992 to 1998 (39). In 2001, the most recent year for which comprehensive data are available, median length of stay was 21.5 days; 34 percent of hospice patients died within 7 days or less from admission, while only 6 percent died in 180 days or over (1). The decline in the mean number of per-beneficiary hospice days was most conspicuously apparent among non-cancer populations. Hospice lengths of stay for this group declined by 38 percent while cancer stays declined by 14 percent from 1992 through 1998 (39). Inordinately short lengths of hospice stay primarily disadvantage patients and providers (40); for example, short stays preclude patients from receiving the full value of hospice services (41), a benefit aimed in part at enhancing the quality of life during the dying process and providing support for the primary patient caregiver. From the organizational perspective, short-term stays typically entail acute and much more expensive care, a reality that disrupts the economic feasibility of hospice care provision. In the special case of delayed referrals, Medicare beneficiaries may be admitted Â“at the brink of deathÂ” or not at all. Many hospices have reported that patients are increasingly referred to them within days of death (42), a circumstance that undercuts the statutory Medicare Hospice Benefit mandate. Conclusion This work is an in-depth examination of the Medicare prognostic criteria for cardiac disease, dementia and stroke. It was undertaken to attain an increased understanding of the criteria from medical, applied and policy perspectives. In its sanction of prognostically-based the Local Medical Review Policies, Medicare
10 administrators sought to introduce consistent payment standards for legitimate hospice enrollment. The physician-researchers who authored the Â“Guideline s Â” assumed the use of their clinical parameters in concert with practitioner judgment and ongoing patient involvement (43). Although the disarticulation between these separate goals may not be readily apparent, the consequences of regulatory reform require careful analyses.
11 Chapter Two: Evaluation of the Local Medical Review Policies for Medicare Hospice Eligibility in Advanced Dementia, Stroke and Heart Disease Abstract Background Medicare fiscal intermediaries judge the eligibility of patients for the Medicare Hospice Benefit (MHB) using Â“Local Medical Review PoliciesÂ” or LMRPs. Because access to quality end-of-life care in the U.S is influenced by such policies, it is important to evaluate their validity and regulatory impact. Methods To evaluate the predictive validity and classificatory errors of the LMRP criteria, a retrospective case-control study was undertaken at a single, large, Medicare-certified hospice in Florida comparing 207 Medicare benefi ciaries with primary diagnoses of heart disease, dementia, or acute or chronic stroke who experienced long-term survival post MHB enrollment (>180 days) with 246 patients matched on primary diagnosis and admission site of care (residence or nursing home) who experienced short-term survival (< 180 days). Results Only the dementia criteria were significantly associated with short-term duration of survival (OR = 2.97, 95% CI = 1.52 to 5.82, p <.005); the heart disease (OR = 1.65, 95% CI = 0.92 to 2.97; p = 0.08) and acute stroke criteria (OR = 3.60, 95% CI = 0.35 to
12 34.71, p = 0.17) were not. The percentage of patients who did not meet LMRP criteria but died within 6 months ranged from 55.8% for dementia to 75.4% for acute stroke. The percentage of patients who met LMRP criteria but survived more than 6 months ranged from 8.6% for acute stroke to 21% for dementia and heart disease. Conclusions Local Medical Review Policies for the MHB misclassify large numbers of hospice patients based on verified durations of survival. Introduction Accurate prediction of duration of survival is important for informed decisionmaking in the care of gravely ill patients. One example is the need to judge a patientÂ’s terminal status for Medicare Hospice Benefit (MHB) eligibility, legally defined as a terminal diagnosis and six-month life expectancy, assuming normal disease course (1). End-of-life prognostication remains a precari ous science (27-29, 32, 44-46), particularly in diagnostic categories other than cancer (10, 12, 46). Despite this, prognostically-based policies have been enacted as claims review tools for MHB reimbursement. In 1998, the Centers for Medicare and Medicaid Services (CMS) adopted clinical protocols for prediction of six-month morta lity in selected non-cancer, chronic diseases. These were disseminated to the five regional CMS-intermediary agencies, which enacted them into regulatory policies over a three-year period (1998-2001). Administrative records auditors use these protocols, termed Â“Local Medical Review PoliciesÂ”(LMRPs) (47), to screen for long-surviving (>six months) MHB recipients who may have been inaccurately certified for hospice enrollment. Through such claims review processes,
13 legitimacy of physician certifications of terminal status have been approved or challenged, patient eligibility for Medicare hospice confirmed or denied, and organizational payment claims reimbursed or denied. Previous studies of the LMRP criteria have focused on their validity for identification of short-term MHB enrollees. Schonwetter et al. (48) examined 104 chronically ill hospice patients who survived less than 6 months following admission and found that only 35% fulfilled the LMRP criteria. Fox et al. (38) studied 923 hospital patients who survived less than 6 months; of these only 277 (30%) were correctly identified as short-term survivors using simulated LMRP versions. Neither Schonwetter nor Fox investigated patients who fulfilled the criteria but experienced long-term (> 6 months) durations of survival. Luchins et al. (49) prospectively identified a shortsurviving subgroup of hospice patients with deme ntia through applications of FAST stage 7C criteria (50-52), key components of the LMRP criteria for dementia. These investigators concluded that the FAST criteria might not be feasible for survival prognosis, given that dementia symptoms do not invariably progress ordinally, conclusions supported in a replication of study results (53). To the investigatorÂ’s knowledge, this evaluation is the first to comprehensively evaluate measures of predictive validity, including errors of classification for the heart disease, dementia and stroke (Table 1) and general (Table 2) sets of LMRP criteria.
14 Table 1. Prognostic criteria for Medicare Hospice Benefit eligibility* by diagnostic category Diagnostic category Required criteria Optional criteria Dementia (1.) FAST Stage 7 or higher AND impaired function (canÂ’t ambulate, dress or bathe without help assistance + incontinent urine and stool + no meaningful communication) AND (2.) one of six medical complications (aspiration pneumonia; upper urinary tract infection; multiple decubitus, stage 3-4; septicemia; fever recurrent with antibiotics; 10% weight loss in 6 months) Heart disease (1.) optimally treated with two ejection fraction < 20% medications (diuretics, vasoarrhythmia dilators, usually ACE inhibitors) history of cardiac AND (2.) symptoms of CHF at arrest or syncope
15 rest and on exertion (NYHAIV) brain embolism resuscitation HIV disease Acute stroke coma >3 days OR coma upper urinary tract infection with myoclonus OR dysphagia decubitus, stage 3-4 that prevents sufficient septicemia nutrition fever with antibiotics 10% weight loss in 6 months. Chronic stroke (1.) FAST Stage 7 or higher AND impaired function (canÂ’t aspiration pneumonia ambulate, dress or bathe w/o upper urinary tract assistance + incontinent urine infection and stool + no meaningful decubitus, stage 3-4 communication) OR (2.) septicemia Karnofsky 40 OR poor fever w/antibiotics nutritional status 10% weight loss in 6 months. Palmetto Regional Home Health and Hospice Intermediary, formerly known as Palmetto Government Benefit Administrators, Local Medical Review Policies, 1/29/1998
16 Table 2. *NHPCO general guidelines for determining prognosis The patient should meet all of the following criteria: I. The patientÂ’s condition is life limiting, and the patient and/or family have been informed of this determination. A. A Â“life limiting conditionÂ” may be due to a specific diagnosis, a combination of diseases, or there may be no specific diagnosis defined. II. The patient and/or family have elected treatment goals directed toward relief of symptoms rather than cure of the underlying disease. III. The patient has either of the following: A. Documented clinical progression of disease, which may include: 1. Progression of the primary disease process as listed in disease-specific criteria, as documented by serial physician assessment, laboratory, radiologic or other studies. 2. Multiple Emergency Department visits or inpatient hospitalizations over the prior six months.
17 3. For homebound patients receiving home health services, nursing assessment may be documented. 4. For patients who do not qualify under 1, 2 or 3, a recent decline in functional status may be documented. a. Functional decline should be r ecent, to distinguish patients who are terminal from those with reduced baseline functional status due to chronic illness. Clinical judgment is required for patients with a terminal condition and impaired status due to a different non-terminal disease, e.g., a patient chronically paraplegic form spinal cord injury who is recently diagnosed with cancer. b. Diminished functional status may be documentd by either : 1. Karnofsky Performance Status of less than or equal to 50% 2. Dependence in at least three of the following six Activities of Daily Living (ADLÂ’s). i. Bathing ii. Dressing
18 iii. Feeding iv. Transfers v. Continence of urine and stool vi. Ability to ambulate independently to bathroom B. Documented recent impaired nutritional status related to the terminal process. 1. Unintentional, progressive weight loss of greater than 10% over the prior six months. 2. Serum albumin less than 2.5 gm/dl may be a helpful prognostic indicator, but should not be used in isolation from other factors in I-III above. *National Hospice Organization Standards and Accreditation Committee, Medical Guidelines Task Force. Medical Guidelines for Determining Prognosis in Selected NonCancer Diseases. 2nd ed. Arlington, Va: National Hospice Organization;1996.
19 Methods The original, 1998 versions of three of the sets of LMRP criteria implemented by the CMS administrator for the Southeastern region of the United States (54) were evaluated. These are very similar to the corresponding sets of LMRP criteria enacted by the other four national Medicare intermediary agencies (54-57). The study was conducted at a single, freestanding, not-for-profit hospice located in West Central Florida with an average 1998 daily census of approximately 1,200 patients. The 1998 case mix profile at this site (45% non-cancer/55% cancer) closely paralleled national hospice case-mix profiles in that year (43% non-cancer/57% cancer) (39). An 18-month study enrollment interval (1/1/97-6/30/98) was selected that largely preceded the national period for staggered LMRP implementation (1/29/98 -10/1/00). This was done in order to give the policies a fair test based on assessment of physicianreferred individuals who would have been admitted to Medicare hospice if the policies did not exist. Because the rarity of long-term durations of survival at the study hospice rendered a cohort design infeasible, a case-control design was adopted. Long-surviving (>180 days) MHB recipients admitted within the specified enrollment interval (cases) were compared to a random sample of short-surviving ( 180 days) recipients bearing identical primary ICD-9 codes (controls). Long-term durations of survival were verified based on nearly four years of follow-up through March 31, 2001, the end date of study observation. All deceased, discharged and surviving MHB recipients admitted to the study site during the enrollment interval with primary ICD-9 codes for cardiac, pulmonary, cerebrovascular and dementing diseases (n = 1,123) were identified. Eleven patients
20 discharged prior to 181 days of hospice care were dropped because it was not possible to determine whether they died before or after 180 days. The remaining medical records were dichotomized into control (< 180 days) or case (>180 days) survival categories. Cases were matched on primary ICD-9 code and admission site of care (residence or nursing home) to one or more randomly selected controls. The majority of pulmonary cases (n = 63) were missing lab values for hypoxemia and hypercapnia, data essential for assessment of criteria fulfillment, because these diagnostic tests were not required previous to policy initiation. Therefore, this ICD-9 category was dropped from the analysis. The final sample (n = 453) consisted of 207 long-surviving cases and 246 short-surviving controls. Two nurse abstractors with clinical and research backgrounds abstracted data from archived medical records. Demographic and duration of survival data were electronically abstracted from the study site administrative database and/or from the Social Security Death Index (SSDI). Local ne wspaper obituary files were searched if the date of death could not be SSDI verified. The LMRP criteria were formatted into assessment instruments for pilot testing on a randomly selected sample of medical records (n = 40). Following the pilot study, hard-copy prototype instruments were revised and operationalized into a Microsoft Access 2000Â™ application to provide disease-specific data entry screens with checks for completion and errors. Abstraction was restricted to a specific set of forms (transfer records, clinical summary checklists, interdisciplinary team admission assessments and plan of care narratives) completed or available at the time of initial MHB enrollment. Abstractors
21 were cautioned against review of re-certification or other data recorded after the initial intake. For patients experiencing more than one enrollment (n = 8) during the observation period, only initial intake data were abstracted. Karnofsky scores (58-59), an optional measure of global disablement relevant to the LMRP for stroke, are narratively recorded in study site records and were missing for 5.1% of the total sample. To reduce potential observational bias, the abstractors were cross-trained in uniform LMRP interpretation, subject to inter-rater reliability tests and blinded to case/control survival outcome. During the training period, conferral between abstractors, investigator and hospice and non-hospice physicians was permitted. During chart review proper, assistance with clinical interpretation was limited to individual abstractor consultation with the hospice Medical Director. For every case and control record in the sample, each LMRP criterion was assessed according to a Â”fulfilled/failed to fulfillÂ” standard. Along with the appropriate set of disease-specific criteria, the National Hospice and Palliative Care OrganizationÂ’s (NHPCO) Â“general guidelinesÂ” (Table 2) were concurrently applied to every sample record. Assessment of fulfillment status was carried out on a disease-by-disease basis. Inter-rater reliability was computed using the simple kappa coefficient following independent and simultaneous abstractor assessments of criteria fulfillment in the 11th through 20th record in each diagnostic grouping. The strength of the association between fulfillment of the overall prognostic criteria and short-term duration of survival was assessed by simple logistic regression. Within each diagnostic grouping, patients were classified into 2 x 2 tables to calculate measures of predictive validity. Because there were more short-term survivors in the cohort than those selected for study, positive and
22 negative predictive values and false positive and negative rates were adjusted for the sampling fraction of controls. This was done by multiplying the number of controls selected who met and did not meet criteria by the total number of short-term survivors divided by the number of controls who were ra ndomly selected. Statistical Applications Software (SAS)Â™ 8.2 (SAS Institute, Inc., Cary, NC) was used for all analyses. Results Demographic and clinical characteristics of the sample are displayed in Table 3. The mean (SD) patient age was 85.5 8.2 years; 60.3% received hospice care in nursing homes (heart disease 45.2%; stroke 51.9%; dementia 79.7%) as opposed to residential care settings. Individuals selected and non-selected as controls from among the original cohort of short-term survivors did not differ signifi cantly with respect to age, sex, functional status or site of care for any of the diagnostic categories. Cases and controls within each diagnostic category did not differ significantly on age, gender or site of care (Table 4). Assessors completely agreed on fulfillment of dementia and stroke criteria (kappa = 1.0). The kappa for fulfillment of the heart disease criteria was 0.6, indicating moderate agreement.
23 Table 3. Characteristics of the sample by diagnosis at hospice admission Acute & Chronic Total Heart Stroke Dementia Sample (n=210) (n=56) (n=187) (n=453) Mean (SD) Age 85.0(8.7) 84.9(9.2) 86.3(7.3) 85.5(8.2 ) Karnofsky score 30.3(10.4) 23.3( 8.8) 25.6(8.1) 27.5(9.7) Percentage Female 54.8 76.0 79.1 67.5 Institutional site of care 45.2 51.9 79.7 60.3 Weight loss 10% or more in 6-months 52.2 45.7 78.1 61.9 Incontinence of urine and stool 34.3 94.0 98.3 68.1 Dyspnea 64.3 24.1 23.5 42.4 Karnofsky < 30 65.2 88.2 87.6 77.2 Bedridden 32.9 74.1 53.0 46.2
24 Table 4. Case/control comparison of the sample at hospice admission Heart Stroke Dementia Total Sample n=210 n=56 n=187 n=453 Case Control Case Control Case Control Case Control n=93 n=117 n=23 n=33 n=91 n=96 n=207 n=246 Age [mean(+SD)] 84.7(+9.9) 85.1(+7.6) 84.8(+9.1) 85.1( +9.2) 86.2(+8.0) 86.5(+6.6) 85.4(9.0) 85.67(7.5) % female 57.0 52.0 61.0 85.0 84.0 76.0 69.0 66.0 % nursing home site of 41.9 47.9 52.2 48.5 73.6 85.4 57.0 62.6 care
25 Using short-term survival ( 6 months) as the outcome variable, odds ratios for fulfillment of the Medicare LMRP prognostic criteria are displayed in Table 5. Only in patients with dementia was the association between fulfillment of prognostic criteria and short-term ( 6 months) survival statistically significant (OR = 2.97, 95% CI = 1.52 to 5.82, p<0.05). The association in heart disease approached, but did not attain statistical significance (OR = 1.65, 95% CI = 0.92 to 2.97; p = 0.08). The association in acute stroke was not significant (OR = 3.60, 95% CI = 0.35 to 34.71, p = 0.17). The odds ratio between the prognostic criteria for chronic stroke and short-term survival could not be calculated, because all chronic stroke patients met the criteria irrespective of duration of survival. One hundred percent of the sample met the NHPCO general guidelines tested. Therefore, odds ratios and measures of predictive validity were not calculated for these criteria. Across the three sets of disease-specific LMRP criteria tested, overall rates of classification error were high and false negativ e error rates were consistently much higher than corresponding false positive error rates. False negative rates ranged from 56% in dementia to 70% in heart disease and 75% in acute stroke (Table 4). In contrast, false positive errors ranged from 9% in acute stroke to 20-21% in heart disease and dementia. Discussion Hospice is one of the fastest growing Medicare services (39). Increasing admission of patients with non-cancer diagnosis in part motivated the Centers for Medicare and Medicaid Services (CMS) to institute Local Medical Review Policies (LMRP) for non-cancer claims review (60). We investigated the association of the
26 Table 5. Odds ratios and measures of predictive validity by LMRP diagnostic category. Odds ratios 95% False #False % Met prognostic confidence positive negative Diagnosis criteria interval P -value *Sensitivity Â†Specificity Â‡PPV Â§NPV rate rate Dementia n=187 64.2 2.97(1.52 Â– 5.82) .005 76.04 48.35 78.98 44.16 21.02 55.84 Heart n=210 55.7 1.65(.92 2.97) .075 61.21 51.06 79.60 29.67 20.40 70.33 Acute stroke n=29 79.3 3.60( ** .35 Â– 34.71) Â†Â†. 168 85.71 37.50 91.69 24.58 8.63 75.42 Chronic stroke n=27 100.00 Zero cells, chi square not Zero cells, unable to calculate measures of predictive validity a valid test Sensitivity probability that short-term survivors meet the criteria. Â†Specificity probability that long-term survivors do not meet the criteria. Â‡Positive predictive value (PPV) probability that those who meet the criteria survive 6 months or less. Â§Negative predictive value (NPV) probability that those w ho do not meet the criteria survive longer than 6 months False positive rate percent of those meeting criteria who were long-term survivors #False negative rate percent of those not meeting the criteria who were short term survivors ** Cornfield limits not accurate; exact limits used Â†Â†Fischer exact results used due to small sample size
27 criteria specified in these policies with actual survival and classification error frequencies. With the exception of dementia, the LMRP criteria were not significantly associated with short vs. long term duration of survival. These findings raise substantial concerns about the use of LMRP criteria as the gold standard upon which the approval or denial of Medicare Hospice Benefit (MHB) reimbursement claims are based. The LMRPs represent modified versions of disease-specific sets of severity indicators that were first published in The Medical Guidelines for Determining Prognosis in Selected Non-cancer Diseases (36). The Â“GuidelinesÂ“ were constructed based on expert opinion combined with previously existing data on clinical markers associated with short-term ( 6-month) survival, and were developed by the National Hospice and Palliative Care Organization (NHPCO), the nationÂ’s largest hospice membership group.. Although the two instruments similarly profile the physical signs of terminal status, they are not similar in content, structure, and intended usage (31,32). The NHPCO Guidelines define conditions of hospice referral by providi ng general and disease-specific protocols, but do not attempt to codify the formulation of end-of-life prognoses. The Medicare policies specify the number and combinations of required criteria and are applied as screening tools in official chart audit processes to protect Medicare against improper use and payments. The LMRP criteria notably narrow and restrict MHB eligibility as defined in the founding MHB Federal statutes (6). Sets of criteria to reliably distinguish the terminal phases (six months or less of life expectancy) of advanced and progressive diseases would serve important clinical, practical and humanitarian goals in hospice and in many other health care settings (33). However, the LMRP criteria investigated appear to have limited scientific merit. With
28 the exception of dementia, neither this study nor those of others (37,38,49,53) have shown these criteria to be significantly associated with short-term ( 6-month) duration of survival outcomes. Although predictive models may improve survival estimation for groups of patients (48,23, 61, 62), wide variability exists in actual survival durations of individual patients (9,38). As is apparent from these findings, app lication of the LMRP criteria can yield considerable misclassification. A consistent pattern of relatively low rates of false positive error was observed in conjunction with much higher corresponding rates of false negative error.. In this policy context, the practical consequence of high false negative errors is erroneous classification as MHB-ineligible of individuals who would survive less than six months. The consequence of false positive errors is erroneous classification as MHB-eligible of individuals who survive longer than six months.. Relative to the costs and benefits of regulatory innovation, false negative errors most disadvantage patients, families and providers; false positive errors most disadvantage payers of publically-funded health care services. A conspicuous annual reduction in hospice median lengths of stay has paralleled national LMRP implementation. From 1992 to 1998, the national average length of stay in hospice declined 20 percent, from 74 to 59 days (39). The decline in the median number of hospice days used per beneficiary was particularly striking among patients with diagnoses other than cancer; length of stay among this patient group declined by 38 percent, while cancer patientsÂ’ stays declined by 14 percent (39). In 2002, the most recent year for which data are available, median length of stay in US Medicare hospice programs was 26 days (1). Decreasing length of stay trends may suggest altered post-
29 LMRP decision-making behaviors on the pa rt of referring physicians and hospice providers, who may favor hospice certification and enrollment of patients whose condition is most reflective of LMRP criteria. Sixty percent of the sample received hospice care in nursing homes, a proportion substantially higher than the 36 percent of MHB recipients who resided in nursing homes on a national basis in 2000 (39). This discrepancy is largely explained by two study design features. First, only patients with non-cancer diagnoses were selected, and second, long-term survivors were over-sampled. Patients with non-cancer diagnoses are disproportionately represented among nursing home populations (39). Of all patients treated at the study site in 2000, approximately 38 percent resided in nursing home settings at admission, similar to the above referenced national percentage. We elected to study hospice populations, rather than those rather than terminally ill, community-based individuals who might show a different association between fulfillment of Local Medical Review Policies and life expectancy. Given the regulatory simulation objective of the study, this may not pose a limitation. We utilized a hospice sample for two additional reasons. First, we sought to hold constant any selection processes that might predispose certain groups to opt for palliative care in lieu of traditional end-of-life health care services. Second, it has been noted that patient quality of life may improve or stabilize in hospice care settings and that prognosis may become less well defined (41). Potential reporting bias presents another area of concern. Given that a large proportion of sample patients suffered cognitive impairment, much of the clinical source data were based on surrogate rather than self-report. Further, certain of the LMRP
30 criteria, e.g., nutritional status, require at patients or surrogates recall patient patterns of weight loss that occurred over the 6 months that precede hospice admission. Reliance on surrogate report is an accepted practice in investigations that rely on secondary sources, however, and was unavoidable in this study. In addition, there is no reason to expect that the quality of data for cases and controls would differ, so any bias introduced is unlikely to be differential. The kappa for fulfillment for heart disease criteria was only moderate (6), which may have led to a weaker association between the criteria fulfillment and case/control status for this disease group. Finally, the findings represent the survival outcomes of patients at a single large, Florida hospice. Replication in hospices with different organizational structures and in nationally representative samples is required.
31 Chapter Three Time-to-Death Factors in Far Advanced Dis ease: The Belief Systems of Experienced, Multidisciplinary Hospice Personnel Abstract Prognostic beliefs of experienced, multi-disciplinary hospice personnel on longterm (6-month), physical and non-physical prognostic influences in far advanced diseases were identified through content analysis of focus group data. Belief was consistent across disciplinary boundaries (nurses, social workers, chaplains, home health aids and physicians) that duration of survival in end-stage disease is primarily influenced by physical factors. Consensus of belief add itionally existed that non-physical factors additionally influence longevity in terminal illness, but more so in non-cancer relative to cancer diseases and at remote (months) versus more imminent (days, weeks) death proximities. Beyond diagnosis, progression and severity of disease, quality of life, stress level, social support, caregiver traits and the milieu of care were identified as particularly important patient time-to-death influences. Although hospice experts explain time to death primarily physically, they additionally report the effects of a complex and dynamically shifting array of patient-specific and contextual factors. Prognosis in advanced diseases may best be informed by physical and non-physical factors.
32 Introduction Development of reliable formulations for long-term (6-month) time-to-death estimation remains a highly desirable but elusive medical goal (9), particularly in noncancer diseases (32). Attempts to codify the more traditional and subjective approach to this dilemma, namely physician clinical judgment, have met with limited success (10). The few indices that attempt to stratify general patients into risk groups for long-term mortality have a number of limitations. Most apply to hospitalized patients (25, 61) and all require complex calculations and data not routinely available to health care personnel in applied clinical settings. Only a few include functional status measures despite its association with mortality in older hospitalized patients (63). The prognostic formulations in use toda y typically exclude non-physical factors, despite the widespread acceptance of their health effects. The goal of this descriptive, exploratory study was a increased understa nding of physical and non-physical factors that may influence prognostic accuracy in end-stage disease. The study was undertaken as a preliminary step in a broader research agenda that evaluates the performance of so called Â“Medicare prognostic criteriaÂ” for the establishment of hospice eligibility. Methods Research Protocol Five discipline-specific focus group sessions of 90-minutes each were conducted with experienced, multi-disciplinary staff with direct patient-care responsibilities at a single large, non-profit hospice located in Florida. Through a pilot testing and revision process, a three-part interview protocol (i ntroductory and closing statements, sets of
33 queries, and instructions for a pen-and-paper exercise) was developed to ensure adherence to the principles of qualitative research (66). Focus group discussion was restricted to time-to-death factors believed to influence the survival duration of newly admitted Medicare hospice beneficiaries, or those initially certified by two physicians to be within a 6 months or less proximity to death. The Sample Hospice personnel from dissimilar health care disciplines (registered nurses, social workers, chaplains, home health aides and physicians) with a minimum two years of direct patient care experience were recru ited for voluntary study participation. Of the 37 total respondents, 23 (61.16%) were female and 100% were Caucasian (Table 6). Individual focus group size ranged from 4 physicians to 11 registered nurses. Mean number of years of hospice experience ranged from a low of 4.5 years (social workers) to a high of 9.4 years (physicians) with a 6.9 years mean level of experience for the total sample.
34 Table 6. Focus group composition, hospice personnel Number of Mean years Discipline participants Male/Female hospice experience Nurses 11 0/11 6.8 Social Workers 11 3/8 4.5 Home Health Aides 6 3/3 9.0 Chaplains 5 4/1 4.6 Physicians 4 4/0 9.4 Total 37 14/23 6.9 Focus Group Interviews Two co-moderators and a trained and experienced moderator facilitated each focus group session. A co-moderator who read a scripted statement on research goals introduced each session. The other co-moderator then in turn phased all queries (physical; psycho-social; spiritual; environmental; caregiver) exactly as scripted: Â“What _______ factors do you consider most influential in determining time-to-death among patients newly admitted (terminally diagnosed and assigned a 6-month or less survival prognosis) to hospice care?Â” A concluding written exercise was also introduced as scripted: Â“List your opinions on the top three physical or non-physical factors that most strongly influence time to death in newly-admitted hospice patients.Â” All focus group discussions were audiotaped, transcribed verbatim and later checked for accuracy against co-moderator hand-written notes.
35 Data Analysis A three-person moderator/co-moderator team applied consensus-based analysis to cross-disciplinary focus group data (65). Each analyst individually applied coding and categorization (66) to each of the five tran scripts; transcript notations were then compared in a moderator/co-moderator group process to identify recurrent, crossdisciplinary themes of discourse. To provide for test-retest reliability, the process of analysis was documented and archived. A full transcript of the raw data is stored on computer disk. Results Analysis revealed cross-disciplinary consensus on a more powerfully influential role of non-physical mortality factors in non-cancer relative to cancer diseases, and at remote (months) versus imminent (days, weeks) death proximities. Care has been taken herein to present findings within a framework of existent medical evidence and to include all study data, including that which may appear contradictory or paradoxical. Verbatim transcript excerpts are displayed in Tables 7-12.
36 Table 7. Hospice-related time-to-death beliefs hospice personnel (Commentator discipline is signified by RN=nurse; SW=psychosocial professional; C=chaplain; CNA=certified nurse assistant; and P=physician) Hospice survival benefits Â“I think with hospice care you finally get your caregiver, change in the environment, psychosocial support, spiritual support. Just as important, the patient gets his medications on a regular basis, maybe for the first time. Someone is showing him how to use them, and the patient lives longer than expected. Hospice actually provides a spectacular service that allows people to be really comfortable, maybe for the first time.Â” P Â“I think hospice people have the capacity to come in when the family is obviously upset because of the process and the reality of the events. Â…And hospice, because of lack of turnover, confidence of staff, the support staff gives each other, is pretty much able to maintain a concerned, matter of fact attitude.Â” P Â“It could be an indication that this person may go sooner if they are particularly lonely. But at the same time, when we get these folks with hospice we see them rebound because they have people who are tending to their needs, caring for them. SW Â“It depends on the diagnosis. It seems like our cardiac people, real elderly cardiac people, actually thrive when hospice care arrives. The COPD people, the people who are left at home alone.Â” And the dementia people, they just go on and on.Â” RN
37 Table 8. General time-to-death beliefs of hospice personnel (Commentator discipline is signified by RN=nurse; SW=psychosocial professional; C=chaplain; CNA=certified nurse assistant; and P=physician) Survival outcomes tend to be patient-specific Â“Because everybody is unique, it (decline) goes on at a different rate.Â” SW Â“That is always the problem with setting up guidelines. It is so iffy. It is relative to the individual at the time.Â” C Â“...it (prognosis) is a nonlinear process. It is the attempt to use actuarial data to apply to a single patient with a number of imponderables and variables.Â” P Survival influences are dynamic Â“Â…All of this stuff can change, the caregiver issues, the environment, the psychosocial aspects and the spiritual. So two weeks from when you did that (made the prognosis) the net balance may have changed.Â” P Â“Any intervention will change the prediction. Any change in the environment. There are lots of ways of changing it.Â” P Â“My point is that during six months period, the estimate of survival you made initially may be so affected by these other changes that it is meaningless.Â” P Survival influences vary by diagnosis Â“They may last nine months, they may last that year, in poor shape, granted, but their will may be a little str onger than their illness at that time, if you are talking about a cardiac or a COPD versus a cancer. We have two different kinds of patients we are talking about.Â” SW Â“ IÂ’m not sure what the reason is, but patients with non-neoplastic diagnoses like chronic lungers seem to do better after they come through the front door of hospice than when they donÂ’t. It seems like just on the same regimen (as cancer patients) they do better.Â” P
38 Table 9. Patient-specific time-to-death beliefs of hospice personnel (Commentator discipline is signified by RN=nurse; SW=psychosocial professional; C=chaplain; CNA=certified nurse assistant; and P=physician) Â“W ill to liveÂ” Â“We have all seen people who should have been dead months agoÂ…Their will to live is keeping them here, their body is notÂ…Their sheer determination to keep going for whatever reason is keeping them here.Â” SW Â“We hear a lot of them say on admission, Â‘Well, I have to live until Â– Â‘I will see that baby,Â’ Â‘I promised her I would walk her down the aisle. I have to be thereÂ…Â’ And come hell or high water, they usually are.Â” RN Â“Will to dieÂ” Â“Â…When they realize that they are just going to get weaker and weaker, they just say Â‘IÂ’m not going to do this, I am just going to die.Â” SW Â“ Depression hastens death. I think itÂ’s a sense of Â‘If this is all there is, I donÂ’t want to be hereÂ’.Â” C Self-defined Quality of life Â“I think in certain situations hospice extends life and in certain situations, it makes the quality of life much better. Maybe in all situations it makes the quality of death better.Â” P Â“Part of the quality of life issue is normalcy. When normalcy declines, people lose those things through which they once defined themselves.Â” SW Â“I think choice equals quality of life and quality of life equals choice. When a decrease of choice happens, a decrease in the quality of life happens.Â” SW
39 Table 10. Physical time-to-death beliefs of hospice personnel (Commentator discipline is signified by RN=nurse; SW=psychosocial professional; C=chaplain; CNA=certified nurse assistant; and P=physician) Â“Â…If they are telling me Â“Gee, I was out walking around the block last week and now I can barely get out of bed, IÂ’m not eating, IÂ’m not drinking.Â” They are really telling me that they know what is going on.Â” RN Â“Within each disease category there are factors which measure the disease, which are important. It basically translates into time in bed and vital signs. And if the weight loss is ten pounds per month, they canÂ’t live very long.Â” P Â“Pain control, far and above for most patients, is the most important thing. Get the pain under control, and then the other things come into play. No matter how much somebody loves you, if you have bone cancer and are screaming in pain, nothing will do.Â” C Â“From a medical standpoint, when peopleÂ’s pain is out of control, very rarely will they be able to Â“let goÂ” until they, either through medication or psychosocial intervention can become relaxed and then they can go.Â” SW Â“I donÂ’t think there is any doubt that (with skillful control of pain) it (death)) may come sooner, but I donÂ’t think that is the issueÂ…Â” P
40 Table 11. Spiritual time-to-death beliefs of hospice personnel (Commentator discipline is signified by RN=nurse; SW=psychosocial professional; C=chaplain; CNA=certified nurse assistant; and P=physician) Spiritual beliefs Â“They are resolved about whatever is next. They may be resolved about the spiritual beliefs they have or they may be resolved that this is it, there is nothing else. So the question is, are they resolved, or are they in fear, unresolved?Â” SW Â“I had one patient who lived so long because of his uncomfortableness with what was going to happen in the hereafter. If we can get to a place where whatever we believe in, whether it is heaven, paradise or reincarnation, and we can accept that belief spiritually, then it is easier to let go of this life.Â” C Â“If they are atheistic, they are not going anywhere for awhileÂ… I have only two who were really true atheist and the patient hung on and hung on. Because there is nothing to go to and they have something here, and there is nothing anywhere else.Â” RN Â“...
41 Table 12. Caregiver time-to-death beliefs of hospice personnel (Commentator discipline is signified by RN=nurse; SW=psychosocial professional; C=chaplain; CNA=certified nurse assistant; and P=physician) Â“...she is elderly, has had lots of health problems. When she greeted me at the door with a walker, then I knew I had a problem. The man just had a hip operation so he is having a hard time getting out of the bed.Â” CNA Â“You go in and you say to this little caregiver, Â‘Why are these MondayÂ’s meds (medications) when it is Wednesday?Â’ and she will say, Â‘Oh, I forgotÂ’.Â” CNA Â“She never had a child. She never had to give any care. She didnÂ’t care if her husband was taking a bath or not. She didnÂ’t care if he ate or not.Â” CNA Â“I think when a caregiver is stressed, it goes right back to the patient. If you are giving the patient stress, it is going to shorten their lives.Â” CNA Â“Anxiety level of the caregiver. The minute I went into the house she grabbed me. She said Â‘I canÂ’t take care of him.Â’ Â‘I donÂ’t want him to dieÂ’.Â” CNA Â“One of the things is the attitude of the caregiver toward the patient. Are they providing care because they want to, because they have to, out of duty, are they hostile about doing it?Â” P Â“Culture. Culture often dictates the type of care and the extent of care that will be given by the familyÂ…A lot ofÂ…from their own spirituality in terms of such things as DNRs, living wills, feeding tubes, withholding nutr ition and these types of thingsÂ…. Maybe the point is that they are weak and that they need food so we are going to make sure that they eatÂ…. Maybe narcotics are badÂ…. It is drugs, no I donÂ’t want to give them that. That is our culture.Â” C
42 Paper and pencil responses have been thematically categorized in Table 13. The majority of written responses concern management of death anxiety and late-life developmental tasks through coping mechanisms that are religious, spiritual and/or philosophical in nature. The only physical survival effects listed by these respondents were diagnosis and overall level of disease severity. Table 13. Time-to-death factors recorded by multi-disciplinary hospice personnel in a written exercise Physical factors Non-cancer diagnoses Physical debility General non-physical factors Patient psycho/social/spiritual posture. Psycho/social/spiritual environment. Patient cognitions, attitudes, mood state factors Will PatientÂ’s will to live. Will to die. PatientÂ’s will to die Â“WillÂ” (strong desire to control what is wanted). Will to live or die. PatientÂ’s will/attitude towards illness/dying. Acceptance Acceptance of death will to die. Verbalization of acceptance and peacefulness to die Acceptance of diagnosis. Patients attitude. PatientÂ’s attitude toward dying. Patient expects to die in certain time frame 6 months.
43 Quality of Life Decreased quality of life having choices. Loss of control. Loss of choice and will. Quality of life. Loss of perceived control. Loss of control and choices. Depression Lack of hope and meaning at that point of prognosis. Loss of hope. Hopelessness. Loss of meaning. Lack of goals to continue living. Loss of desire to fight illness any longer Depression both that of patient and careg iver including lack of meaning, lack of support, sense of isolation Social support factors Social support Emotional ties. Lack of support system being alone. Â“FamilyÂ” support and attitude. Family/caregiver dynamics. Support system. Lack of supportive care environment. Limited support systems. Loss of support. Lack of interpersonal relationships. Relationships of people involved in care. Emotional climate surrounding patient. Lack of caregivers. Lack of support. Closure in interpersonal and practical affairs Resolution of end-of-life issues. No Â“unfinished business.Â” Completion of unsettled issues. Bothersome issues have been resolved. Unfinished business. Patient reports he/she is Â“ready to go,Â” all goals/tasks have been completed. Loved ones will be cared for. Patient feels confident that ones left behind will be O.K. at the time of death. Financial.
44 Permission to die Primary caregiver has accepted patientÂ’s terminal diagnosis. Family caregiver tells patient Â“itÂ”s O.K. to go, IÂ’ll be O.K.Â” Family permission to leave (die). Patient and family acceptance of end of life. Spiritua l factors Reconciliation, forgiveness PatientÂ’s issues re: forgiveness of self and others. Reconciliation with God, family or others forgiveness, independence. Reconciliation of life issues (includes patientÂ’s discussion with deceased persons) Peace, acceptance through spiritual/religious belief system Inner peace Whether patient has made peace with his dying psychosocial/spiritual acceptance Belief system surrounding death spiritual preparedness. Spiritual belief vs. Non-belief. Belief that they are going to a better place. Belief system. Patients belief system or lack of same. Open discussion with family members regarding belief in future life. Willingness to discuss their religious spiritual issues. Patients with strong faith or those with none seem more able to face the unknown. Patient is O.K. with dying feels has had a good life, believes in some kind of after life Patient reports he/she is at peace has lived a good life, will go to a better place. Spiritual. Spiritual comfort Patients view and meaning of death. Patient and caregiver find peace in near death experiences. Visualizing the calling Â“to goÂ” from other deceased. Physical Time-to-Death Beliefs of Hospice Personnel Respondents perceived time-to-death judgments to be much more precarious in non-cancer than cancer diseases, a belief epidemiologically verified (12). Furthermore, imminently impending death (death within days) was considered relatively easy to
45 predict through reliance on clinical signs of organ system shutdown, but the range of time-to-death markers at long-term (6-month) proximities were characterized more than once as Â“imponderable.Â” These respondents linked prognostic accuracy with clinical knowledge of patients obtained within that patientÂ’s own care milieu, a belief supported by the traditions of clinical medical practice. Although physical factors were deemed the primary s time-to-death determinants in advanced diseases, dissimilar hospice survival curves were observed to occur in patients with similar diagnostic and physiological profiles. Physical control of pain through pharmacological methods was paradoxically referenced as both supporting prolonged survival and facilitating life/death transitions. Patient-specific Time-to-Death Beliefs of Hospice Personnel Patient Attitudes, Cognitions and Behaviors Patient verbal expressions and/ behavioral manifestations of Â“will to liveÂ” or Â“will to dieÂ” attitudes were described as weak but consequential survival influences in noncancer but not cancer diseases. PatientsÂ’ Â“will to surviveÂ” intents were most often discussed in previously reported end-of-life contexts (67), such as patient desire to live to experience upcoming visits, holidays or ceremonial occasions of deep personal significance and/or patient resolve to attain pre-death closure practical, psychological, spiritual and/or interpersonal affairs (68). The reality of such effects on survival is supported by (69-70) Â“dip-peakÂ” death patterns that are known to cross-culturally bracket events of broad social importance.
46 Quality of Life Improved quality of life (QOL) has been identified as a chief benefit of palliative over traditional end-of-life care (71). Quality of life patient-appraised as unsatisfactory was associated by these respondents with previously described late-life syndromes that include Â“weariness with life,Â” Â“loss of the will to live,Â” and late life clinical depression (72-75). Quality of life more positively appraised by patients was believed to be a powerful motivator for a continued personal struggle for survival (76). Stress The tremendous challenges and burdens inherent in conditions of advanced old age, severe illnesses, and limited survival prognoses were discussed by research participants, burdens described as encompassing limited function, chronic pain, alteration of personally significant life roles, conflictual or unresolved or relationship issues the reality of impending loss of the Self and others, and the developmental tasks of life completion and closure. Â“Death or existential anxietyÂ” thus described is in accord with the psychoanalytic belief systems of Erikson (77), who maintained that the central late life developmental challenge concerns the maintenance of psychological equilibrium or Â“integrity versus despair.Â” Individual management of stressful late life challenges through religiouslyor spirituallyoriented channels was linked by these respondents with improved well-being (78) and improved mental and physical health. Further, these informants linked religious preoccupations of a less positive nature, those for example that might be focused on after-life retribution, with height ened patient anxiety. Late life anxiety
47 thus intensified was paradoxically described as both a life-prolonging and deathhastening agent. According to respondent beliefs as analyzed herein, the survival effects of patient religious beliefs vary according to death proximity. For example, at more remote 6 month proximities, self-comforting religious or spiritual ideations and/or behaviors were described as important contributors to positively appraised patient QOL, health maintenance and prolonged survival. At more imminent death proximities, however, similar thoughts and actions were more commonly linked with facilitated and hastened life/death transitions. In sum, these data suggest that hospice personnel perceive end-stage health status and longevity to be subject to modification through cognitive/behavioral channels. While these findings are not clear-cut, they are consonant with similarly mixed literature reports. For example, Jarvis and Northcott found associations of religion with survival (79), whereas Christakis (40) did not. The general association between religious and spiritual behaviors and improved health and wellbeing is well known (80-84). The negative health consequences of stress are also established (85-86), and intriguingly, prayer has been associated with improved health outcomes (87). Contextual Time-to-Death Beliefs of Hospice Personnel Social Relationships Supportive social relationships were described as quality of life essentials, their absence an ominous mortality risk factor. This was held to be true even given contrary
48 patient social habits or preferences. Related research shows that social isolation places patients at increased mortality risk (88-90); cardiac patients appear to be differentially disadvantaged by this factor (91). Caregiver Factors Focus group members believed that caregiver-specific traits, including certain disordered mood states (, depressed, extremely anxious) and care delivery styles described as sub-optimal (hostile, withholding, inept) have an effect on patient time to death in hospice settings. The association most frequently referenced linked impaired caregiver physical or cognitive function with negative patient survival outcomes. Because over 50 percent of MHB beneficiaries are aged 75 years and over (1), their care at home is, according to these respondents, largely provided by spouses who are themselves elderly, and oftentimes significantly hindered by cognitive or physical health problems. Crippling arthritis, heart conditions, back problems that impede lifting and hearing impairments were among the caregiver problems discussed by focus group members. A related study on caregiving noted a rapidly increasing extent of ill health as the age of the Â“carerÂ” increases; in this study over three quarters of the careers over the age of 75 years reported some previous ill health. Finally, an absent or neglectful caregiver was universally held by focus group members to place patients at premature death risk. Although caregiving has been intensively studied over the past decade, surprisingly little is known about the association of explicit caregiver variables and survival outcomes. Higher stress levels among caregiver populations has been established (92), however, as have lower levels of life satisfaction (93), greater incidence
49 of depression (94) and poorer health (95) in comparison to non-caregivers. According to these respondents, because hospice care is usually provided in the patientÂ’s home, or those of relatives or friends, the caregiverÂ’s physical and psychological abilities are key patient quality of life and survival factors. Milieu of Care Focus group members strongly endorsed a hospice versus conventional care survival advantage On the one hand proving a hospice survival effect is methodologically problematic and remains to be empirically established. On the other, hospice patient and caregiver satisfaction levels exceed those reported in conventional medical settings (96-97). Effective manageme nt of physical and psychological distress is fundamental to the hospice model of care. Discussion Accurate prediction of the course of end-stage disease becomes increasingly urgent in terminal conditions. Patients and families need reliable survival estimates to facilitate end-of-life planning; physicians must rely on prognosis for appropriately timed hospice referrals. Despite the humanitarian and administrative relevance of this topic, little is known about Â“long-term mortality predictors,Â” or time-to-death factors operative within 6-month as opposed to a days or weeks timeframe. Hospice care settings are ideal for the study of both long-term and more immediate time-to-death factors In the focus group discussions, non-physical, patient-specific, cognitive/behavioral factors (will to live, quality of life, stress and anxiety) and contextual factors (social support, caregiver, environmental) were believed to be
50 important time-to-death influences, but much more cogently so in non-cancer as compared to cancer diseases. Focus group members were convinced, irrespective of disciplinary background, that hospice care interventions benefit quality of life and late life survival, and particularly so for patients with non-cancer diseases. If or why this might be so is unknown, but rationales may be found in Â“burden of illnessÂ” theories. The burden of illness in dementia, heart disease, stroke and other chronic, life-threatening conditions is particularly high for both patients and caregivers. This is because the consequences of chronic illnesses are persistent and recurring over many yearsÂ’ duration and significantly limit individual ability to perform routine activities of daily living. Thus, in addition to medical services, people with chronic conditi ons often experience Â“weariness with life,Â” and may require intervention that are social, psychological and/or rehabilitative in nature. Furthermore, increased burden of illness factors have been associated with higher incidence of clinical depression (94), and psychosocial factors are known to modify the association between disability and depression in older adults (98). At the risk of oversimplification, hospice may sufficiently improve the quality of life to tilt the balance toward protracted individual struggle for surviv al. As previously referenced according to hospice experts, any Â“will to liveÂ” survival effect may be essentially null in cancer diseases, but significantly important to the prolongation of non-cancer survival. Stress is known to alter biomarkers (91), and as previously referenced; hospice may confer superior advantages that support effective stress management. The health effects of many of the time-to-death factors qualitatively identified are empirically established, but prognostic usefulness by and large remains unfounded.
51 Based on these analyses, it may be concluded that prognosis from months-long perspectives is considerably more complex than reflected in codified tools such as the Â“Medicare Prognostic criteriaÂ” that screen for 6-month survival, requiring attention to influences other than physical disease severity. The relative lack of evidence showing a role for non-physical, patient-specific and contex tual factors in remote prognosis of death leaves the contribution of such factors largely unknown. The value of the focus group findings may therefore lie less in prognostic implication, but may point to the need for psychosocial interventions aimed at improving patient and caregiver well-being and health status and in medical settings. The value of training in life completion and closure issues for both patients and caregivers is suggested, as are the development of techniques specifically targeted at the management of death anxiety and late-life stress. Maximization of interpersonal connectedness and social support in healthcare settings is strongly suggested as therapeutic. Such interventions and others, including increased awareness of and treatment for late-life anxious and depressive disorders are likely to be beneficial regardless of their implications for duration of survival. Limitations of the Research Staff or patient demographics that may be unique may limit generalizability of the findings. Hospice volunteers were unfortunate ly omitted from the respondent pool even though they contribute 13% of all clinical hours to hospice patients and families (1). The study design did not allow for differentiation of terms commonly used by respondents such as Â“loss of the will to liveÂ” Â“will to die,Â” Â“readiness to die,Â” and Â“acceptance of death.Â” For example, it was not possible from these analyses to deduce whether the
52 described Â“will to dieÂ” attitudes of terminally ill individuals are expressions of mental disorder or of rational self-determination. Further Research on Prognosis An accuracy comparison of various methods for time-to-death estimation might prove instructive, including clinical judgments of physicians, team-based consensus of multidisciplinary hospice teams, and formulaic approaches as represented by the Medicare prognostic criteria. Disease-specific, serial measures of reserve capacity over the trajectory of fatal decline would also be instructive and might include immune function, neuroendocrinology and/or cardiovascular activity and other stress response variables. Disease-specific comparisons of psychosocial and spiritual, patient and caregiver issues over the course of fatal decline would also offer insight.
53 Chapter 4: Can the Predictive Performance of the Dementia Criterion for Medicare Hospice Eligibility Be Improved? Abstract Certain Medicare prognostic criteria validly predict short-term survival ( 6month) among Medicare hospice beneficiaries with dementia. Methodological difficulties exist in these nationally applied screening instruments that include high rates of false negative errors that restrict patient inclusiveness. Tests of the original Medicare dementia criteria individually and in all possible combinations revealed that the self-care skills criterion, when applied in isolation, yields improved prognostic performance over the original three criteria, including a better balance of false negative/false positive error rates. Functional impairment measures may offer improved prognosis in dementia because of their integrative rather than single organor body-system focus. Clinicians and healthcare planners should be aware of the potential usefulness of functional dependence as a prognostic indicator in end-stage dementia. Introduction Families and clinicians face difficult decisions in dementia care, particularly the initiation of palliative or hospice care in lieu of curative treatments. Hospice care may be an attractive health care option for family caregivers because an atypically comprehensive array of medical and psychosocial services is available for the care of
54 severely impaired and largely bedbound (99-101) demented patients. In addition to obvious service advantages offered, the growing proportion of patients with dementia among hospice populations can be demographically explained. Currently dementia is the fourth leading cause of death among older Americans (2). Whereas in 1995, 2% of all patients admitted to Medicare hospice were diagnosed with dementia; this proportion had climbed to 7 percent in 2001 (1). Health care professionals who treat demented individuals strongly endorse palliative goals in end-stage dementia care ( 102-104). However, because Medicarehospice eligibility requires physician-certified 6 months or less life expectancy, prognostic difficulty in dementia (105-106) poses significant barriers to hospice access (42 ) Furthermore, due to recent Centers for Medicare and Medicaid Services (CMS) regulatory innovation, patients with far-advanced multi-infarct dementias and those of the AlzheimerÂ’s type must now fulfill disease specific, clinically oriented Â“Medicare prognostic criteria for dementiaÂ” for hospice eligibility in addition to broader certification requirements (Table 14).
55 Table 14. Medicare prognostic criteria for dementia (Patient must criteria I, II and III) I. FUNCTIONAL ASSESSMENT STAGING Patient meets one FAST Stage 7 or beyond: A. 6 words Â– Speech ability limited to approx. half dozen words or fewer, in the course of an average day or in the course of an interview B. 1 word Â– Speech ability limited to the use of a single intelligible word in an average day or in the course of an intensive interview C. Unable to sit up D. Unable to smile E. Unable to hold head up II. KATZ INDEX OF ACTIVITIES OF DAYLY LIVING Patient has all of the 5 functional impairments listed: A. Unable to ambulate without assistance B. Unable to dress w/o assistance C. Unable to bathe without assistance D. Urinary and fecal incontinence, intermittent or constant E. No meaningful verbal communication, stereotypical phrases only, or ability to speak is limited to six or fewer intelligible words
56 III. MEDICAL COMPLICATIONS OF TERMINAL ILLNESS Patient has one of the medical complications listed within the past 12 months: A. Aspiration pneumonia B. Pyelonephritis or other upper urinary tract infection C. Septicemia D. Decubitus ulcers, multiple, stage 3-4 E. Fever, recurrent after antibiotics F. Inability to maintain sufficient fluid and calorie intake with a 10% weight loss during the previous six months or serum albumin <2.5 gm/dl. This investigation aimed to improve the predictive performance of the Medicare prognostic criteria for dementia through selective dropping and re-combination of the indices of which the policy is constructed. The specific objective was to reduce classificatory error rates and to achieve an improved false negative/false positive error rate balance. The three clinical indices (see Â“MethodsÂ” below) that make up the Medicare prognostic criteria for dementia were tested individually and in various combinations.
57 Methods According to the LMRP criteria for demen tia (Chapter 2, Table 1), patients should specifically fulfill all of the following three criteria to legitimize Medicare hospice certification: 1) Â“FAST (50-52)Â” meets one level of the Functional Assessment Staging Scale, Stage 7C; 2) Â“KATZ (107)Â”has five of five ADL impairments; 3) Â“MEDICALÂ” has one of six medical complications. Tests were conducted in which criteria were systematically dropped and re-examined in all possible combinations. Subsequent to each modification, risk estimates and measures of predictive validity were re-calculated. The strength of the association between fu lfillment of each set of criteria and the outcome variable ( 6-months survival) was assessed by simple logistic regression, yielding odds ratios (OR) and 95% confidence intervals. The sample was classified into 2 x 2 tables to calculate measures of predic tive validity. A previously described sampling fraction was applied (Chapter 2) a nd Statistical Applications Software (SAS)Â™ 8.2 (SAS Institute, Inc., Cary, NC) was used for all analyses The study was conducted at a Medicare-certified Florida hospice selected based on the availability of access to a large sample of medial records (n=187) of dementia patients admitted during a specified 18-month interval (1/1/97-6/30/98). Selection of the sample has been previously described (Chapter Two).
58 Results Demographic and clinical characteristics of the sample (n=187) are shown in Table 15. The sample was overwhelmingly Caucasian (96%), very old (mean age 86.3), and largely female (80.0%). The majority of sample patients received hospice care in nursing home settings (80.0%) as compared to home-based or other residential settings. The mean sample Karnofsky score (58), a global measure of disablement and dependency upon others to conduct daily life activities, was 25.6 (8.1). A score of 26 indicates Â“severely disabled status, with the possible need for hospital admission (58).Â”
59 Table 15. Characteristics of patients with dementia on first admission to hospice care Dementia (n=187) Mean ( SD) Age 86.3( 7.3) Karnofsky score 25.6( 8.1) Percent (%) Caucasian 96.0 Female 80.0 Nursing home-based 80.0 Bedbound 52.9 History of ER visits 13.9 Weight loss 78.1 Edema 4.8 Dyspnea 23.5 Altered speech 66.3 Decubitus, multiple, stage 3-4 30.0 Incontinent urine and stool 98.3 Agitation 20.3 Unresponsive 12.8 Pain 25.7 Anxiety 16.6 Depression 19.8
60 Predictive performance of differing combinations of the Medicare prognostic criteria Table 16 compares odds ratios for 6 month or less survival, false positive, and false negative rates for 7 models encompassing all possible combinations of the criteria. When criterion III was dropped, leaving criteria I and II, the strength of association between the independent variable (criteria) and dependent variable ( 6-month survival) increased, as evidenced by a higher odds ratio. Furthermore, false positive and false negative error rates were lower, a better balance of false positive/false negative errors was achieved; and a larger proportion of sample patients fulfilled the reduced set of two criteria. Because all patients met criterion I, the use of the functional criterion (II) in isolation yielded identical findings. Discussion Although the use of prognostic criteria for 6-month survival in dementia has been Centers for Medicare and Medicaid Services-sanctioned since 1998, the predictive validity of Medicare screening tools remain empirically unresolved. On the one hand, one early study showed little association between this predictive formulation and shortterm, 6-month survival (48), on the other, an independent research team reported that the FAST component of the Medicare prognostic criteria was significantly related to 6month survival times of hospice patients (49, 53). A 2003 study (108 ) evaluated both the predictive validity of the original Medicare prognostic criteria for dementia and related but novel long-term survival predictors. No significant association was found between the Medicare dementia criterion and short-term survival; however, advanced
61 Table 16. Assessment of the predictive validity of the Medicare Local Medical Review Policy for dementia; single and composite indices Odds ratio *False Â†False % patients Criteria included (95% CI) positive rate negative rate meeting criteria Â‡I, II and III 2.97 (1.52-5.82) 21.02 55.84 64.2 Â§I and II 5.12 (1.83-14.33) 24.38 37.78 86.6 II and III 2.05 (1.13-3.71) 22.63 62.50 59.4 #I and III 1.45 9 (0.78-2.69) 19.04 70.17 51.3 **I -100.0 Â†Â†II 5.12 (1.83-14.33) 24.38 37.80 86.6 Â‡Â‡`III 1.45 (.78-2.69) 25.78 66.44 67.9 *False positive rate Â– percent of those meeting criteria who were long-term survivors Â†False negative rate Â– percent of those not meeting the criteria who were short term survivors Â‡ Original Medicare criteria (I, II & III inclusive) Â§ FAST + Katz Katz + Medical Complications # FAST + Medical Complications **FAST Â†Â† KATZ Â‡Â‡ Medical Complications
62 age, and impaired nutritional and functional status were found to be independently associated with this outcome. The findings reported here demonstrate a significant association between the Medicare prognostic criterion for dementia and survival duration of 6 months. Furthermore, these findings demonstrate that improved predictive performance of this criterion may be achieved through dropping one of the criteria. A single Katz ADL criterion (107) used in lieu of the original Medicare composite criteria results in an improved risk estimate and a more acceptable balance between false negative and false positive error rates. Balanced error rates are important because these demonstrate equitable distribution of the costs and benefits of regulatory reform, in this case among Medicare payors, patients and providers. In sum, the use of a single functional, integrative measure of disease severity in lieu of more explicit and composite severity measures may yield prognostic improvements in the Medicare prognostic criteria, a finding of obvious policy relevance. Dementia prognosis, particularly prognosis in the early stages of dementia (109110), has been extensively studied and multiple factors have been identified as having a significant relationship with survival in this disease (111-116). The Medicare prognostic criteria are composites of specific clinical measures of internal physiologic function, such as laboratory values and vital signs, functional performance measures and signs of general physical debility that include a range of medical complications. As described by Stein and her group (117), measures of integrative function such as the Katz ADL index may support improved prognosis because they reflect the impact of illness on the whole person rather than single organ or body systems. Judging by study outcomes, measures
63 of integrative functioning add important information about the severity of end-stage dementia beyond that provided by internal physiologic measures. Of note is the fact that past and more current prognostic work strongly supports the relationship between functional status and mortality in general older adult populations (118), in hospitalized cohorts (119, 120) and in nursing home reside nts (121-124), who disproportionately tend to be patients with dementia. Alternately, findings may imply that time to death in endstage dementia is less a feature of individual health status and more of generalized physiologic reserve capacity (125). As previously referenced, the criteria that comprise FAST Stage 7 have been found associated with 6-month or less hospice survival (50-51). However, this criterion may not be suitable for end-of-life prognosis because dementia severity does not always progress in an ordinal fashion, as might be implied by the FAST structure (least severe to most severe indices). Furthermore, FAST severity indicators were developed exclusive of reference to the many non-AlzheimerÂ’s dementia sub-types among hospice populations. In addition, as demonstrated by our finding of 100 percent FAST criteria fulfillment, FAST stage 7 indices may not sufficiently reflect the levels of disease severity that are most prevalent among hospice populations. Co-morbidity, the third Medicare prognostic criterion for dementia, has been linked with increased dementia mortality (126, 127). Because of the severe motor impairment caused by dementia brain pathology, aspirations, decubitus ulcers, falls, incontinence and organ system infections are common occurrences among patient populations. Not surprisingly then, many patients with dementia die from secondary complications that most prominently feature pneumonia (128) rather than from the
64 assigned primary diagnosis (129). Given the body of prior research, the lack of association between the overall Medicare co-morbidity criterion and 6-month mortality reported here is surprising. It may be worthwhile to individually examine the predictive validity of the six conditions that comprise the co-morbidity index. Single vs. multiple study site design should be noted as a possible study limitation. The positive association reported here between predictor and outcome variables may be nationally atypical, reflective of a unique study site census, organizational structure and/or administrative approaches. While such a circumstance does not negate study results, multi-site replication would resolve the issue of the predictive utility of these criteria across nationally diverse hospice samples and organizations. Further Research An objective set of prognostic criteria would be advantageous to increase the confidence of families and physicians that the hospice care option is appropriate for patients with dementia. Although correlation between LMRP functional status measures and patient prognosis has been demonstrated, clos er examination of the data is required to define that point in time at which the discriminatory power of these measures diminishes. The question remains, are functional status measures useful for remote, 6-month prognoses, or merely as predictors of death within a few weeks or days timeframe? Further study of predictors of intermediate (2-3 months) versus long-term (6-months) mortality in dementia through survival analysis methods would be of value. Factors associated with Â“ultra-longÂ” patient survival (1 year or more) in end-stage dementia
65 would also be of considerable interest. Cost/benefit analyses would also be highly instructive to compare traditional, acute care costs of dementia patients judged hospice ineligible against dementia patients judged eligible with similar levels of disease severity treated in Medicare hospice settings. Finally, research on serial measures of reserve capacity in end-stage dementia have led to fascin ating insights (130) and if pursued might increase understanding of dementia time-to-death influences. Conclusion The composite Medicare prognostic criterion for dementia is a significant discriminator between 6-month/>6-month survival in end-stage disease, suggesting prognostic utility. However, the occurrence of false negative errors associated with these criteria persists and may be reduced by dropping one criterion, thus increasing the practical value as a screening tool for appropriate hospice enrollment.
66 Chapter Five: Medicare Local Medical Review Policies (LMRPs): Concepts and Consequences Abstract Due to concerns about United States medical costs, third-party and Medicare interest has increased in strategies to control health care usage. Local Medical Review Policies or Â“LMRPsÂ” represent a nationally rele vant example. Administrators apply these sets of clinical criteria, to justify Medicare claims payments or denials. LMRPs were nationally sanctioned and regionally implemented beginning in 1998, but remain controversial to this day on scientific (108, 37-38, Chapter 2) and social equitability grounds (131). The 1997 statement of a SUPPORT investigator proves prescient in an LMRP context, Â“Using statistical estimates of prognosis to designate a category of Â‘terminally illÂ’ patients for public policy purposes is unavoidably arbitrary, will often be contested, and will have differential effects upon those dying of different diagnoses (9).Â” Local Medical Review Policies: Â“Clinical Guidelines or Policies?Â” The Centers of Medicare and Medicaid Services have represented the LMRPs as akin to Â“clinical guidelinesÂ” that health care practitioners and hospice providers may flexibly interpret. This characterization does not appear to be accurate, however, from technical and applied perspectives. First, according to standard medical terminology, clinical policies apply to collections of patients and are designed to reduce clinician
67 subjectivity and to increase the uniformity of medical decision-making. Clinical guidelines are designed as clinical reference tools for use by clinicians as they formulate medical decisions in regard to individual patient judgments (132,133). Second, flexibility claims are not supported by the historical f acts of LMRP policy evolution. The National Hospice and Palliative Care Guidelines (36), the LMRP source document, suggest general hospice eligibility given the presence of three clinical conditions. These are the terminality and election requirements shown in Table 17, plus fulfillment of either the general physical debility or the disease-specific criteria. The LMRPs in contrast specify MHB eligibility if all four conditions shown in Table 17 are fulfilled. The Guidelines when appropriated for policy usage were thus altered to become more stringent. Moreover, a close examination of the disease-specific (as opposed to general) criteria exposes a cross-diagnostic differential in inclusiveness. As may be observed in Table 18, the numbers of criteria that must be fulfilled vary diagnostically (Table 18), denoting differential eligibility restriction across disease-specific screening instruments studied. LMRP Assumptive, Methodological, and Applied Limitations The Local Medical Review Policies are based on related assumptions that may or may not be valid: first, the existence of a discernable end of life phase in chronic, life threatening illnesses; second, the validity of cancer-based methods for non-cancer prognostication; and third, similar mortality curves as a function of similar non-cancer diagnoses. The second assumption is particularly relevant because, despite the exclusive LMRPs non-cancer focus, these criteria represent obvious extensions of cancer-based
68 Â“staging theoryÂ” for disease severity estimation. Cancer systems use tumor size and location (134), and more recently, performance status in lung cancer (135) to estimate Table 17. NHPCO Â“General Guidelines of Medicare hospice eligibilityÂ” The patientÂ’s condition is life-limiting The patient and/or family have elected palliative treatment goals The patient shows symptoms of severe physical debility : Patientor caregiver-reported decrements in patient health status over the months that precede MHB enrollment as documented by home health or hospice personnel. Qualifying symptoms may include multiple emergency room visits OR recent decline in functional status OR 10% unintentional weight loss over the prior six months. OR The patient shows signs of progression in disease severity Clinical or objective data obtained through serial physician assessment, or laboratory, radiologic or other studies. Table 18. Disease-specific comparison Local Medical Review Policies fulfillment requirements Disease Required Criteria by Number Heart 1(a & b) and 2 Pulmonary 1(a & b) and 2 Stroke 1 or 2 or 3 Dementia 1 or Fast Score 7c and 2 disease severity and time to death. The relationship between symptom severity and disease progression in non-cancer diseases ma y not be analogous, however (Chapter 3). Investigations including Chapter 2 of this work report little predictive relationship between the LMRP clinical indicators and 6-month mortality (108, 37-38).
69 Methodological Limitations A second LMRP concern is their assumptive as opposed to empirical origin (Chapter 2). Furthermore, although the LMRPs comprise several different scales, the validity and reliability of each of which has b een previously confirmed, the reliability of composite scale application has not been confirmed. When empirically assessed in the present investigation, the LMRPs appear to largely lack predictive validity. Applied Limitations Large-scale LMRP regulatory simulations such as this work call the objectivity, reliability and speed of application of these criteria into question. Reliability or consistency of results is a leading measure of an instrumentÂ’s quality. Only moderate inter-rater agreement (kappa=.06) was obtained between experienced and well-trained nurse assessors in applications of the heart disease LMRP. (In contrast, perfect agreement (1.0) was observed in the Kaplan correlation coefficients between these raters in audits of identical dementia and stroke LMPR medical records). A lesser rate of interrater reliability for the overall heart disease criteria tested (Chapter 2, Results section) indicates cross-diagnostic disparity in LMRP reliability. Sub-optimal reliability may result from the unavoidable subjectivity of certain LMRP criteria.. For example, one of the LMRP heart disease criteria, Â“the patient has the inability to carry on physical activity without discomfort ,Â” requires assessor reference to personal, internalized notions of pain and discomfort. Furthermore, in test appli cations, newly encountered clinical questions were often so perplexing that they could not be resolved without physicianÂ’s consultation.. Similar IRR comparison among Medicare auditors in field settings would
70 test the hypothesis that acceptable IRR on LMRP fulfillment varies with the intensiveness of training and the availability of ongoing medical consultation. Moreover, the pilot study conducted prior to commencement of Chapter 2 research demonstrated a 22to 35to 59-minute mean time variability among RNs who assessed LMRP fulfillment of identical patient records. Apparently, an average of about one-half hour is required for a careful and comprehensive LMRP assessment. A comparative study of time assessment among Medicare claims auditors in field settings would be instructive. Test applications conducted here identified additional LMRP logistical issues: Required 6-month background data were not reliably obtainable in hospice records Required laboratory data included as LMRP core and optional criteria were not reliably obtainable in hospice records, si nce they were not required prior to LMRP implementation. Certain functional status indices (bathing, dr essing) are inapplicable to mainly bed bound hospice patients; as evidenced by 100% fulfillment of certain such criteria (bathing, dressing) in this study. Such crite ria were thus not useful discriminators of short-tem from longer-term survival. The dichotomous LMRP format (fulfilled/failed to fulfill), does not allow for fine gradations of health status assessment. Expansion of disease severity grades would allow for more precise assessment. LMRPs: Background, Costs and Benefits The evolution of the Local Medical Review Policies into hospice eligibility standards was driven by two related regulatory concerns: skyrocketing hospice growth and inappropriate hospice utilization by non-terminally ill persons. Despite this, the
71 LMRPs are popularly framed as the regulator y sequel of fraud convictions obtained in Puerto Rican but not continental U.S. hospices (136). Wrongdoing in this highly publicized case involved hospice usage for essentially long-term, custodial as opposed to time-limited, palliative care. However objectionable the issue of Medicare fraud, the occurrence of longer-term hospice stays is relatively rare, and should not be overstated. At the time of 1998 LMRP initialization, slightly less than 15% of all hospice patients could be classified as Â“longterm survivors,Â” i.e., patients whose hospice survival duration exceeded 6-months duration (12). In that year, the long-stay population was balanced by almost an identical number of Â“short-term survivors,Â” i.e., patients who survived in hospice for one week or less (12). In 2001, however, the prevalence of long-term stay in hospice had dropped to 6% (1); but short-term stays had skyrocketed to 28 percent (39). These unprecedented shifts in hospice utilization rates have been more recently confirmed by a 2003 Centers for Disease Control study (7). The LMRPs may have profoundly affected hospice lengths of stay, dramatically altering the historical long-stay/short-stay equilibrium, and reducing the proportion of patients who survive in hospice for periods in excess of 6 months. As curbs to explosive hospice growth, the Local Medical Review Policies have additionally achieved the apparent goals of regulatory reform, but th rough channels less direct and perhaps not anticipated. On the one hand, the number of Medicare Hospice Benefit enrollees has continued to increase, more than doubling in the last decade, from 143,000 in 1992, to 360,000 in 1998 (39) to 885,000 in 2002, the most recent date for which data are currently available (1). On the other hand, in tandem with LMRP instigation, the 1974 to
72 1997 sharp annual growth rate in hospice provider organizations ceased in 1998, and remained level to 2003 (1). From 1999 to 2002, the numbers of organizational providers has essentially remained flat, at 1998 levels (1). While it may never be possible to link in itiation of the LMRPs and altered hospice utilization patterns, any resultant barriers to hospice access would be manifested by delayed patient referrals, biased certifications/re-certifications processes and inappropriate (premature) discharges. Precisely these trends are suggested by recent study results that show that patients discharged alive from hospice are more likely to be female, to have received hospice care for more than 60 days, and to have non-cancer diagnoses (7). Â“Brink of deathÂ” hospice admissions and/or discharges are contrary to 1982 federal mandates establishing hospice care as a feasible non-crisis/non-cure oriented health care option in the United States. The origin of hospice short-stay/long stay imbalance remains speculative, but may most reasonably be explained by altered pos t-LRMP decision-making behaviors that are systemic. Fear of professional and/or economic sanction, and the known unreliability of prognostic estimates coupled with policies that imply otherwise, can help to explain how and why the LMRPs have become so pervasively influential throughout the hospice enterprise, before, during and after the actual delivery of hospice care (Table 20). If such rationales were indeed valid, disparities in MHB eligibility might be observed across diagnoses. A multi-site hospice comparison of post-LMRP longterm/short-term survival trends by non-cancer diagnosis would shed light on this important question. The results of this study and one important other (36) do in fact show variability in numbers of patients who fulfill the criteria on a disease-specific basis.
73 Once again, the real-life consequence of a seemingly minor technicality in regulatory policy is hospice eligibility discrimination by diagnosis. This most worrisome LMRP consideration is clearly illustrated in Chapter 2, Table 4 that lists disparate rates of criteria fulfillment on a disease-specific basis. Additional hypothesized consequences of LMRP regulatory reform, some but not all testable, are listed in Table 19 and are graphically depicted in Table 20. Policy-based Alternatives to the Prognostic Criteria for Medicare Hospice Eligibility One alternative to present policy is a legal amendment to include 6 months or more Medicare hospice eligibility if the terminal illness runs its normal course. Given official recognition that a range of error naturally accrues to probability estimates, and that short-stay, long-stay rates in hospice are beneficially balanced, federally approved hospice utilization parameters could be more explicitly stated. For example, rather than current governmental scrutiny focused on individual long-term patient stays, organizational long-term stay utilization rates might be nationally monitored. Nonconcordat utilization rates would instigate cl oser regulatory scrutiny of individual hospice organizations. Another idea is a program whereby hospices might receive preauthorization from Medicare contractors for hospice care is cases in which prognosis is difficult. Suggestions for Systems-wide Research in Terminal Illness Health care systems research might include longitudinal outcome studies of patients who are questionably hospice eligible, and are therefore not admitted. Such
74 studies would examine length of survival, patterns of health care services utilization and trends of hospice readmission. A cost analysis of hospice ineligible admissions/eligible denials would also be of value. Additionally, a descriptive study of Â“ultra-long survivors,Â” i.e., patients who survive for one year or more post hospice admission, might prove administratively and clinically instructive.
75 Table 19. Potential outcomes of the Local Medical Review Policies for Medicare hospice benefit eligibility Patients hospice eligibility discrimination by diagnosis absent, delayed or prematurely discontinued hospice care discontinuity in model of care and locale diminution of patient quality of life adverse health outcomes Family caregivers increased caregiver burden/stress adverse caregiver health outcomes discontinuation of informal care Physicians jeopardized diagnostic/prognostic autonomy disrupted patient/physician clinical relationships Hospice Organizations resource shift from patient care to administrative compliance increased per-patient cost of care provision financial viability threatened U.S. Health Care System increased use of curative in lieu of palliative services increased resource allocation for end-of-life costs.
76 Table 20. Health-systems outcomes of the prognostic criteria for Medicare Hospice eligibility PRE-AND POST-ADMISSIONS OU TCOMES REIMBURSEMENT & POST-DISCHARGE OUTCOMES 1. Patients 2. Physicians 1. Patients 2. Physicians 3. Providers U.S. Health Care System 3. Providers Â• Admission Â• Altered decisionÂ• Premature Â• Compromised discrimination making processes discharge clinical by Dx. authority and Â• Late or avoided Â• Barriers to quality patient relations Referrals to hospice end-of-life care Â• Increased Â• National decline in Â• Financial threat to documentation/ hospice length of hospice viability reduced patient stay care Â• Increased use of Â• Reduced lengths acute care system of stay = increased severity, increased per patient expense
77 Conclusion Existent LMRP studies examine the capacity of LMRP clinical criteria to accurately categorize short-stay/long-stay hospice survival outcomes. The potential for disparity in Medicare Hospice Benefit access where groups with non-cancer diseases disproportionately encounter eligibility barriers is a troubling implication. A more comprehensive understanding might be achieved through systems-wide study of the costs and benefits of prognostically-based Local Medical Review Policies. The question remains Â“Is it possible to identify valid and useful predictors of 6month survival? What rate of classificatory error may be considered unacceptably high? Is there a viable alternative to LMRP/Medicare prognostic criteria governance of the Medicare Hospice Benefit? According to Joanne Lynn, MD, a well-known SUPPORT team scholar, reliable, disease-specific demarcation of severe from terminal illness may not represent an achievable scientific goal (9). Based on the findings reported within, it appears that prognostic science does not currently provide a reliable foundation upon which to establish exclusions for public health care benefits.
78 Dissertation Conclusion The tools of science, including quantitative, qualitative and analytical methods, have been applied to better understand the Medicare prognostic criteria, their validity, applied utility and patient, physician and provider impact. From the perspective of federal analysts, chart auditors need well-defined, time-efficient and nationally relevant standards to facilitate objective Medicare claims review. From a more global perspective, reliable markers of 6-month life expectancy would be undeniably valuable. A poorly designed policy, however, can result in mismanagement of thousands of patients and misallocation of millions of dollars (135). It is recognized that the limits of public health service are properly set by a society at any given time. However, if some deem the current bounds of public health care sub-optimal, substantive and articulated rationales for regulatory reform are required.
79 References 1. National Hospice and Palliative Care Organization NHPCO Facts and Figures January 2003. http://www.nhpco.org/files/public/Facts Figures Jan-03.pdf. Accessed September 25, 2003. 2. Minino AM, Smith BL. Deaths: Preliminary Data for 2000. National Vital Statistics Report. Hyattsville, MD: The National Center for Health Statistics 2001;Vol 49, No 12. 3. Robinson BE, Pham H. Cost-effectiveness of hospice care. Clin Geriatr Med. 1996;12: 417-428. 4. Centers for Disease Control and Prevention. Unrealized Prevention Opportunities: Reducing the Health and Economic Burden of Chronic Disease. Atlanta, GA: US Department of Health and Human Services, Centers for Disease Control and Prevention; 2000. 5. Knoll AM. Health insurance and palliative care. N. Engl J Med. 1998;338:767. 6. Social Security Act. (SSA)Â§1814(a)(7); CFRÂ§418.22, (1983). 7. National Center for Health Statistics. Characteristics of hospice care discharges and their length of service : United States, 2000. U.S. Department of Health and Human Services; Vital Health Stat 13. 2003 Aug;(154):1-36. ...Report No.: WA 900 V83h ser.13 no.154 2003. 8. Christakis NA, Lamont EB. Extent and determinants of error in doctors' prognoses in terminally ill patients: prospective cohort study. Brit Med J 2000;320:469-473. 9. Lynn J, Harrell F, Cohn F, Wagner, D, Connors, AF. Prognoses of seriously ill hospitalized patients on the days before death: implications for patient care and public policy. New Horizons 1997;5:56-61. 10. Pearlman RA. Variability in physician estimates of survival for acute respiratory failure in chronic obstructive pulmonary disease Chest. 1987;91:515-521. 11. Christakis NA. Timing of referral of terminally ill patients to an outpatient hospice. J Gen Intern Med. 1994; 9:314-320. 12. Christakis NA, Escarce JJ. Survival of Medicare patients after enrollment in hospice programs. N Engl J Med. 1996;335:172-178. 13. Edmondson B. The facts of death. Am Demogr. 1997;19:46-53.
80 14. Oldershaw PJ. An overview of survival studies in patients with heart failure. Postgrad Med J. 1992;68:606-610. 15. Kinzbrunner BM, Weinreb NJ, Merriman MP. Debility, unspecified: a terminal diagnosis Amer J Hosp Palliat Care. 1996;13:38-44. 16. Kohn RR. Cause of death in very old people. JAMA. 1982;20:2793-2797. 17. Morris JN, Suissa S, Sherwood S. Last days: a study of the quality of life of terminally ill cancer patients. J Chron Dis. 1986;39:47-62. 18. McGrath P. Caregivers' insights on th e dying trajectory in hematology oncology. Cancer Nurs. 2001. 24 (5):413-21. 19. Viitanen M, Erikisson S, Asplundi K, Wester PO, Winblad B. Determinants of long-term survival after stroke. Acta Medica Scandinavica. 1987;2221:349-356. 20. Kinzbrunner BM. The terminally ill patient. In: Abeloff MD, Armitage JO, Lichter AS, Niederhuber J, eds. Clinical Oncology New York: Churchill Livingstone, 1995;409. 21. Lewis SC, Dennis MS, OÂ’Rourke SJ, Sharpe M. Negative attitudes among short term stroke survivors predict worse long-term survival. Stroke. 2001;32:1640. 22. Wong DT. Knaus WA. Predicting outcome in critical care: the current status of the APACHE prognostic scoring system. Can Jl of Anaesth 1991;38:374-383. 23. Knaus WA, Zimmerman JE, Wagner DP, Wagner DP, Draper EA, Lawrence, DE. APACHE-Acute physiology and chronic health evaluation: a physiologically based classification system. Crit Care Med 1981;9:591-597. 24. Knaus WA, Wagner DP, Draper EA. The Apache III Prognostic System: risk prediction of hospital mortality for critically ill hospitalized adults. Chest. 1991;100:1619-1636. 25. Knaus WA, Harrell FE, Lynn et al. The SUPPORT prognostic model: objective estimates of survival for seriously ill hospitalized adults. Ann Intern Med. 1995;122:191-203. 26. Knaus WA, Wagner DP, Lynn J. Short-term mortality predictions for critically ill hospitalized adults: science and ethics. Science. 1994;254:389-394. 27. Hakim RB, Teno JM, Harrell FE Jr.et al. F actors associated with do-not-resuscitate orders: patients' preferences, prognoses, and physicians' judgments. SUPPORT
81 Investigators. Study to Understand Prognoses and Preferences for Outcomes and Risks of Treatment. Ann Intern Med. 1996;125:284-293. 28. Forster LE, Lynn J. Predicting life span for applicants to inpatient hospice. Arch Intern Med. 1988;148:2540-2543. 29. Parkes CM. Accuracy of predictions of survival in later stages of cancer. BMJ. 1972;1:29-31. 30. Lynn J, Harrell FE, Cohn F. Hamel M, Dawson, N. Wu AW. Defining the "terminally ill": insights from SUPPORT. Duquesne Law Rev. 1996;35:311-336. 31. Brody H. Lynn J. The physician's responsibility under the new Medicare reimbursement for hospice care. N Engl J Med. 1984;310:920-2. 32. Christakis NA. Predicting patient survival before and after hospice enrollment. Hosp J. 1998;13:71-87. 33. Gonnella JS, Hornbrook MC, Louis DZ. Staging of disease: a case-mix measurement. JAMA. 1984;251:637-644. 34. Shackney SE. The growth rate of cancers. In: Fries JE, Ehrlich GE, eds. Prognosis: Contemporary Outcomes of Disease Bowie, MD: The Charles Press Publishers; 1981,84-86. 35. Kinzbrunner BM. Non-malignant terminal diseases. Criteria for hospice admission. Hospice Update. 1993;3:3-6. 36. National Hospice Organization Standards and Accreditation Committee, Medical Guidelines Task Force. Medical Guidelines for Determining Prognosis in Selected Non-Cancer Diseases. 2nd ed. Arlington, Va: National Hospice Organization;1996. 37. National Hospice and Palliative Care Organization. Hospice Care and the Institutional Barriers to its Success: An Issues Paper Developed for NHPCOÂ’s Public Policy Committee Washington, D.C.:NHPCO. 2002. 38. Fox E, Landrum-McNif K, Zhong Z, Dawson NV, Wu AW, Lynn J. Evaluation of prognostic criteria for determining hospice eligibility in patients with advanced lung, heart or liver disease. JAMA. 1999;282:1638-1645. 39. United States General Accounting Office. Medicare: More beneficiaries use hospice but for fewer days of care GAO/HEHS-00-182 2000. September 2000.
82 40. Christakis NA. Timing of referral of terminally ill patients to an outpatient hospice. J Gen Intern Med. 1994;9:314-320. 41. Naik A, DeHaven MJ. Short stays in hospice. A review & update. Caring. 2001; 20:10-13. 42. Barriers to the Use of Hospice Care at the End of Life: Hearings before the Special Committee on Aging, 106th Cong, 2nd Sess. (April 23, 2000). 43. Baumrucker S. Is there an answer to the medical review policy nightmare? Am J Hosp Palliat Care. 1998;15:132-135. 44. Christakis NA, Iwashyna TJ. Attitudes and self-reported practice regarding prognosis in a national sample of internists. Arch Intern Med. 1998;158:23892395. 45. Lynn J, Teno JM, Harrell FE. Accurate prognostications of death: opportunities and challenges for clinicians. West J Med. 1995;163:250-257. 46. Claessens MT, Lynn J, Zhong Z et al. Dying with lung cancer or chronic obstructive pulmonary disease: insights from SUPPORT. J Am Geriatr Soc. 2000; 48:S146-153. 47. Medicare Regulations Program Integrity Manual, Chapter 13. Available at http://www.cms.gov/manuals/108.pim/plm83c.asp. Accessed 10-29-02. 48. Schonwetter RS. Soendker S. Perron V. Martin B. Robinson BE. Thal AE. Review of Medicare's proposed hospice eligibility criteria for select noncancer patients. Am J Hosp Palliat Care. 1998;May/June:155-158. 49. Luchins DJ, Hanrahan P, Murphy K. Criteria for enrolling dementia patients in hospice. J Am Geriatr Soc. 1997, 45:1054-1059. 50. Reisberg B. Dementia: A systematic approach to identifying reversible causes. Geriatrics. 1986; 41:30-46. 51. Reisberg B. Functional Assessment Staging (FAST). Psychopharmacol Bull. 1988; 24:653-659. 52. Sclan SG, Reisberg B. Functional Assessment Staging (FAST) in Alzheimer's disease: Reliability, validity and ordinality. Int Psychogeriatr. 1992;4:55-69. 53. Hanrahan P, Raymond M, McGowan E, Luchins DJ. Criteria for enrolling dementia patients in hospice: a replication. Am J Hosp Palliat Care. 1999;16:395400.
83 54. Palmetto RHHI. Available from: URL:: http://www.palmettogba.com. Accessed October 29, 2002. 55. Associated Hospital Services of Maine RHHI. Available from: URL: http://www.ahsmedicare.com. Accessed October 29, 2002. 56. Cahaba RHHI. Available from: URL: http:www.iamedicare.com. Accessed October 29, 2002. 57. United Government Services RHHI, formerly Blue Cross Blue Shield of California. Available from: URL: http://www.bcc-medicare.com. Accessed October 29, 2002. 58. Karnofsky DA, Abelmann WH, Craver LF, Burchenal JH. The use of nitrogen mustards in the palliative treatment of cancer. Cancer. 1948;1:634-656. 59. Hutchinson TA, Boyd NF, Feinstein AR. Scientific problems in clinical scales as demonstrated in the Karnofsky index of performance status. J Chron Dis. 1979;33:661-666. 60. Office of the Inspector General, Department of Health and Human Services. Review of Hospice Eligibility at the Hospice of the Florida Suncoast A-04-940211, August, 1996. 61. Lemeshow S, Teres D, Klar J. Mortality probability models (MPMII) based on an international cohort of intensive care unit patients. JAMA. 1993;270:2478-2486. 62. Teno J.M, Harrell F, Knaus WA et al. Prediction of survival of older hospitalized patients: the HELP survival model. J Am Geriatr Soc. 2000;48:S16-24. 63. Reuben DB, Mor V, Hiris J. Clinical symptoms and length of survival in patients with terminal cancer. Ann Intern Med. 1988;48:1586-1591. 64. Mortimer JA. Do psychosocial risk factors contribute to AlzheimerÂ’s disease? In A.S. Henderson & J.H. Henderson, eds. Etiology of Dementia of AlzheimerÂ’s Type. Chichester, New York, John Wiley & Sons, 1988. p. 165-78. 65. Krueger R, Casey MA. Focus groups: a practical guide for applied research. Thousand Oaks, New York, Sage Publications, 2000. 66. Auerbach, CF, Silverstein LB. Qualitative data: an introduction to coding and analysis. New York: New York University Press; 2003. p. 10-11. 67. Byock, IR. Dying Well. New York: Riverhead Books, 2001. p. 104.
84 68. Byock,IR. The nature of suffering and the nature of opportunity at the end of life. Clin Geriatr Med. 1996;12:237-252. 69. Anson J, Anson O. Death rests a while: holy day and Sabbath effects on Jewish mortality in Israel. Soc Sci Med. 2001;52:83-97. 70. Phillips DP, Smith DG. Postponement of death until symbolically meaningful occasions. JAMA. 1990;263:1947-1951. 71. Kane RL, Wales J, Bernstein L, Leibowitz A, Kaplan S. A randomized controlled trial of hospice care. Lancet. 1984; 8382:890-894. 72. Chochinov HM, Wilson KG, Enns M, Lander S. Depression, hopelessness, and suicidal ideation in the terminally ill. Psychosomatics. 1998;9:366-370. 73. Chochinov HM, Wilson KG, Enns M, et al. Desire for death in the terminally ill. Am J Psychiatr. 1995;152:1185-1191. 74. Linden M, Barnow S. The wish to die in very old persons near the end of life: a psychiatric problem? Results from the Berlin Aging Study. Inter Psychoger. 1997;9:291-307. 75. Tataryn D, Chochinov HM. Predicting the trajectory of will to live in terminally ill patients. Psychosomatics. 2002;43:370-377. 76. Smith DC, Maher MF. Achieving a healthy death: the dying personÂ’s attitudinal contributions. Hosp J. 1993;9:21-32. 77. Erikson E. Childhood and Society New York, New York, W.W. Norton & Co, 1950. 78. Oxman TE, Freeman DH, Manheimer ED. Lack of social participation or religious strength and comfort as risk factors for death after cardiac surgery in the elderly. Psychosom Med. 1995; 57:5-15. 79. Jarvis GK, Northcott HC. Religion and differences in morbidity and mortality. Soc Sci & Med .1987;2(7):813-824. 80. Levin JS. Religion and health: is there an association, is it valid, and is it causal? Soc Sci Med. 1994;38:1475-1482. 81. King DG. Religion and health relationships: a review. Journal of Religion & Health. 1990; 28:101-112.
85 82. Loenig HG, George K, Blazer DG, Pritichett JT, Meador KAG. The relationship between religion and anxiety in a sample of community-dwelling older adults. J Geriatr Psychiatry 1993;26:65-93. 83. Millison MB. A review of the research on spiritual care and hospice. Hosp J. 1995;10:3-18. 84. Hungelmann J, Kenkel-Rossi E, Klassen L, Stollenwerk RM. Spiritual well-being in older adults: harmonious interconnectedness. Journal of Religion & Health. 1985; 24:147-153. 85. Gibbs HW, Achterberg-Lawlis J. Spiritual values and death anxiety: implications for counseling with terminal cancer patients. J Counsl Psychol. 1978;25:563-569. 86. Levy S, Herberman R, Lippman M, DÂ’Angelo T. Correlation of stress factors with sustained depression of natural killer cell activity and predicted prognosis in patients with breast cancer. J Clin Oncol. 1987;53:348-353. 87. Byrd RC. Positive therapeutic effects of intercessory prayer in a coronary care unit population. South Med J. 1998;81:826-829. 88. Brummett BH, Barefoot JC, Siegler IC et al. Characteristics of socially isolated patients with coronary artery disease who are at elevated risk for mortality. Psychosom Med. 2001;63:267-272. 89. Schoenbach VJ, Kaplan BH, Fredman L, Kleinbaum DG. Social ties and mortality in Evans County, Georgia. Am J Epidemiol. 1986;123:577-591. 90. Seeman TE. Social ties and health: the benefits of social integration. Ann Epidemiol. 1996;6:442-451. 91. Berkman LF. Blumenthal J. Burg et al. Effects of treating depression and low perceived social support on clinical events after myocardial infarction: the Enhancing Recovery in Coronary Heart Disease Patients (ENRICHD). JAMA. 2003;289(23):3106-3116. 92. Davis BD, Cowley SA, Ryland RK. The effects of terminal illness on patients and their carers. J Adv Nur. 1996;23:512-520. 93. Karla wish JH, Casarett D, Klocinski J, Clark CM. The relationship between caregiverÂ’s global ratings of AlzheimerÂ’s disease, patientÂ’s quality of life, disease severity, and the caregiverÂ’s experience. J Am Geriatr Soc. 2000;49:1066-1070.
86 94. Schulz RS, Mendelson AB, Haley WE, et al End-of-life care and the effects of bereavement on family caregivers of persons with dementia. NEJM. 2003;349(20):1936-1942. 95. Haley WE, LaMonde LA, Han B, Narramou S, Schonwetter R. Family caregiving in hospice: effects on psychological and health functioning among spousal caregivers of hospice patients with lung cancer and dementia. Hosp J. 2001;15:118. 96. Greer DS, Mor V. An overview of the National Hospce Study findings. J Chron Dis. 1986;39:5-7. 97. Teno JM, Clarridge BR, Casey V. Family perspectives on end-of-life care at the last place of care. JAMA. 2004;291:94-98. 98. Jang Y, Haley WE, Small BJ Mortimer JA. The role of mastery and social resources in the associations between disability and depression in later life. Gerontologist. 2002;42:807-813. 99. Boyd CO, Vernon GM. Primary care of the older adult with end-stage AlzheimerÂ’s disease. Nurse Pract. 1998;23:63-69. 100. Brechling BG, Kuhn D. A specialized hospice for dementia patients and their families. Am J Hosp Palliat Care. 1986;6(3):27-30. 101. Collins C. Ogle K. Patterns of predeath service use by dementia patients with a family caregiver. J Am Geriatr Soc. 42(7):719-22. 102. Hanrahan P; Luchins DJ. Feasible criteria for enrolling end-stage dementia patients in home hospice care. Hosp J. 1995;10(3):47-54. 103. Luchins DJ, Hanrahan P. What is appropriate health care for end-stage dementia? J Am Geriatr Soc. 1995;41(1):25-30. 104. Volicer L, Volicer BJ, Hurley AC. Is hospice care appropriate for Alzheimer patients? Caring. 1993;12(11):50-55. 105. Brodaty H, Mcgilchrist C, Harris L, Peters KE. Time until institutionalization and death in patients with dementia. Arch Neurol. 1993;2:643-650. 106. Volicer BJ, Hurley A, Fabiszewski, KJ, Montgomery P, Volicer L. Predicting short-term survival for patients with advanced AlzheimerÂ’s disease. J Am Geriatr Soc. 1993;41:535-540.
87 107. Katz S, Ford AB, Moskowitz, RW,Jackson, BA, Jaffe,MA. Studies of illness in the aged, the index of ADL: a standardized measure of biological and psychosocial function. JAMA. 1963;185:914-919. 108. Schonwetter RS. Han B. Small BJ. Martin B. Tope K. Haley WE. Predictors of six-month survival among patients with dementia: an evaluation of hospice Medicare guidelines. Am J Hosp Palliat Care. 2003;20(2):105-113. 109. Brodaty H, McGilchrist C, Harris L, Peters KE. Time until institutionalization and death in patients with dementia. Arch Neurol. 1993;2:643-650. 110. Katzman R, Hill R, You ESH, et al. Predictors of mortality in clinically diagnosed dementia in a population survey of Shanghai, China. Arch Neurol. 1994;51(12):1220-1225. 111. Siebens H, Trupe E, Siebens A, Cook F, Anshen S, Hanauer R, Oster G Correlates and consequences of eating dependency in institutionalized elderly. J Am Geriatr Soc. 1986;34(3):192-198. 112. Aguero-Torres H, Fratiglioni L, Guo Z, Viitanen M, Winblad B. Prognostic factors in very old demented adults: a seven-year follow-up from a population-based survey in Stockholm. J Am Geriatri Soc. 1998;46:444-452. 113. Bowen JD, Malter AD, Sheppard L, et al. Predictors of mortality in patients diagnosed with probable AlzheimerÂ’s disease. Neurology. 1996;47:433-439. 114. Hier DB, Warach JD, Gorelick PB, Thomas J. Predictors of survival in clinically diagnosed AlzheimerÂ’s disease and multi-infarct dementia. Arch Neurol. 1998;46:1213-1216. 115. Stern Y, Tang MX, Albert MS, et al Predicting time to nursing home care and death in individuals with AlzheimerÂ’s disease. JAMA. S1997;277(24):1933-1934. 116. White H, Pieper C, Schmader K. The asso ciation of weight change in AlzheimerÂ’s disease with severity of disease and mortality: a longitudinal analysis. J Am Geriatr Soc. 1998;46(10):1223-1227. 117. Stein R, Gortmaker SL, Perrin EC, et al. Severity of illness: concepts and measurements. Lancet. 1987;2:1506-1509. 118. Clark LP, Dion DM, Barker WH. Taking to bed: rapid functional decline in an independently mobile older population living in an immediate-care facility. J Am Geriat. Soc. 1990;38:967-972.
88 119. Covinsky KE, Justice AC, Rosenthal GE Palmer RM, Landerfeld CS. Measuring prognosis and case mix in hospitalized elders Â– the importance of functional status. J Gen Intern Med. 1997;12(4):203-208. 120. Walter LC, Brand RJ, Counsell SR, et al. Development and validation of a prognostic index for 1-year mortality in older adults after hospitalization. JAMA. 2001;285(23):2987-2994. 121. Brody EM, Kleban MH, Lawton MP, Levy R, Waldow A Predictors of mortality in the mentally-impaired institutionalized aged. J Chronic Dis. 1972 Nov; 25(10):611-20. 122. Lichtenstein MJ, Federspiel CF, Schaffner, W. Factors associated with early demise in nursing home residents: a case control study. J Am Geriatr Soc. 1985;33:315-319. 123. Bianchetti A, Scuratti A, Zanetti O, et al. Predictors of mortality and institutionalization in AlzheimerÂ’s disease patients 1 year after discharge from an Alzheimer dementia unit. Dementia. 1995;6:108-112. 124. Herbert M, Parlato V, Lese GB, Dabaj A, Forette F, Boller F. Survival in institutionalized patients: influence of dementia and loss of functional capacities. Arch Neurol. 1995;52:469-476. 125. Snowdon DA, Greiner LH, Mortimer JA, Riley KP, Greiner PA, Markesbery WR. Brain infarction and the clinical expre ssion of Alzheimer disease. The Nun Study. JAMA. 1997;277(10):813-817. 126. Van Dijk PT.Dippel DW,Van Der Meulen JH, Habbema JD. Comorbidity and its effect on mortality in nursing home patients with dementia J Nerv Ment Dis. 1996;184(3). 127. Poses RM, McLish DK, Smith WR, Bekes C, Scott WE. Prediction of survival of critically ill patients by admission comorbidity. J Clin Epidemiol. 1996;49(7):743747. 128. Morrison RS, Siu AL. Survival in end-stage dementia following acute illness. JAMA. 2000;284(1):47-52. 129. Brauner DJ. Treating nondementia illnesses in patients with dementia. JAMA. 2000;283(4):3230-3235. 130. Neumann PJ, Araki SS, Arcelus A et al. Measuring AlzheimerÂ’s disease progression probabilities: estimates from CERAD. 2000. Neurology. 57(6):957-64.
89 131. The Wall Street Journal, Page B2, Dow Jones & Company. 08/18/2000. 132. Eddy DM. Clinical policies and quality clinical practice. N Eng J Med. 1982;6; 343-347. 133. Eddy DM. Designing a practice policy: standards, guidelines and options. JAMA. 1990; 263: 3077,3081,3044. 134. Shackney SE. The growth rate of cancers. In Fries, JE & Ehrlich, GE. Prognosis: Contemporary Outcomes of Disease Bowie, MD: The Charles Press Publishers;1981;84-86. 135. Schonwetter RS, Robinson BE, Ramirez G. Prognostic factors for survival in terminal lung cancer patients. J Gen Intern Med. 1994;9:366-371. 136. United States Office of Inspector General (OIG). Enhanced controls needed to assure validity of Medicare hospice enrollments. 1996, A-05-96-00023. 135.Eddy DM. Clinical decision making from theory to practice. three battles to win. JAMA. 1994;271: 520-526.
90 About the Author D. Helen Moore earned a BA in Social Work at Florida State University (1972) and an MA in Gerontology at the University of South Florida (1980). As a gerontologist serving for over 20 years in clinical, administrative and service planning roles, she has also authored a trade book on consumer-directed eldercare, 100 monthly aging issues columns for The St. Petersburg Times (1991-1998), 3 U.S. Administration on Aging monographs, numerous aging-related grant proposals and curricula. While an Aging Studies Doctoral Candidate at U.S. F, sh e was a research award recipient, 1997 copresident of the Aging Studies Student Associ ation, adjunct instructor in the U. S.F. Department of Gerontology, and a clinical evaluator in AlzheimerÂ’s clinical drug research. Research interests include AlzheimerÂ’s disease, functional impairment, end-oflife and caregiver issues and health-systems studies.