Use of Medicare's Hospice Benefit by Nursing Facility Residents

03/01/2000

Susan C. Miller, Pedro Gozalo and Vincent Mor

Center for Gerontology and Health Care Research, Brown University


This report was prepared under contract #100-97-0010 between the U.S. Department of Health and Human Services (HHS), Office of Disability, Aging and Long-Term Care Policy (DALTCP) and the Urban Institute. For additional information about the study, you may visit the DALTCP home page at http://aspe.hhs.gov/daltcp/home.htm or contact the ASPE Project Officer, Jennie Harvell, at HHS/ASPE/DALTCP, Room 424E, H.H. Humphrey Building, 200 Independence Avenue, SW, Washington, DC 20201. Her e-mail address is: Jennie.Harvell@osaspe.dhhs.gov.


The goal of ASPE's Medicare Hospice Benefit study is to provide general information on the role of the Medicare hospice benefit and more specific information about how end of life care is provided to institutionalized beneficiaries. Six reports have been produced from this study:

Synthesis and Analysis of Medicare's Hospice Benefit: Executive Summary and Recommendations (report 1) briefly summarizes the methods used for each report and the findings and recommendations that emerged from each of the following reports under this study.

Important Questions for Hospice in the Next Century (report 2), synthesizes the literature related to the Medicare hospice benefit and summarizes discussions with key informants on the use of hospice in nursing homes.

Medicare's Hospice Benefit: Use and Expenditures (report 3), analyzes Medicare utilization and payments for hospice users in 1996.

Use of Medicare's Hospice Benefit by Nursing Facility Residents (report 4), examines differences in hospice utilization and expenditures as a function of when nursing facility residents started using hospice services (i.e., before or during a nursing home stay).

Outcomes and Utilization for Hospice and Non-Hospice Nursing Facility Decedents, (report 5) compares pain management and types of services provided to dying nursing home residents receiving hospice compared to other dying residents who did not receive hospice.

Hospice Benefits and Utilization in the Large Employer Market (report 6), reports on how hospice services are provided by 52 large employers and used by their employees, and identifies alternative approaches to designing and administering hospice benefits.

All of these reports are available on ASPE's Office of Disability, Aging and Long-Term Care Policy website (http://aspe.hhs.gov/daltcp/home.htm). Copies can be mailed out by contacting the Office of Disability, Aging and Long-Term Care Policy, Room 424E, H.H. Humphrey Building, 200 Independence Avenue, S.W., Washington, D.C., by fax at 202-401-7733, or by email at DALTCP2@osaspe.dhhs.gov.

TABLE OF CONTENTS

INTRODUCTION

METHODS

Data Sources

Population Studies

Measures

Methods of Analysis

RESULTS

Estimate of the Scope of Use of Medicare Hospice in Nursing Homes

Comparison of Pre, Overlap and Post Hospice Admissions

Description of "Post" Nursing Facility Hospice Patients

Utilization of Medicare Hospice Benefit and Related Utilization by "Post" Hospice Patients

DISCUSSION

Scope of Hospice in Nursing Homes

Different Types of Hospice in Nursing Facility Patients

Length of Hospice Stay

Concentration of Hospice Care in Selected Nursing Facilities

Hospice Admission Following SNF Discharge

Study Limitations

CONCLUSION

REFERENCES

APPENDIX A. Data and Variables Used for Hospice in Nursing Facility Analyses (separate file)

INTRODUCTION

With the passage of the Omnibus Budget Reconciliation Acts of 1985 and 1989, use of the Medicare Hospice benefit became more widely available to nursing facility residents. These adjustments to the Medicare Hospice eligibility criteria allowed for those residing in nursing facilities to receive the benefit under the same terms as applied to those living in a community setting. Since the enactment of OBRA '89, Medicare hospice beneficiaries residing in nursing facilities represent the most rapidly growing segment of hospice beneficiaries. Indeed, some have estimated that nursing facility residents constitute up to 35 percent of all hospice beneficiaries in some markets (DHHS, 1997). However, nearly 10 years after the passage of OBRA '89 little is known about the utilization of the hospice benefit by nursing facility residents.

This report characterizes those nursing facility residents who enroll in the Medicare hospice program including when they enroll, how long they receive the benefit and how much other services they use. A subsequent report in this study compares the utilization and outcomes experienced by hospice and non-hospice nursing facility decedents who have been matched on a variety of criteria.

METHODS

Data Sources

Several sources of data were combined in conducting the analyses presented in this report. These include computerized Minimum Data Set (MDS) information from nursing facility resident assessments, Health Care Financing Administration (HCFA) hospital, skilled nursing facility and hospice claims data, and organizational data on nursing home providers. The MDS data from the states included in our study included detailed prescription drug use data as well as detailed information on residents' functioning and clinical condition. On-line Survey Certification Automated Survey (OSCAR) nursing facility data and hospice provider of service data were used to characterize the hospices caring for patients and nursing facilities in which patients resided. A more detailed description of each of these sources of data and the types of elements drawn from them are included in Appendix A of this overall report.

Population Studied

We merged 1992 to 1996 MDS data in the 5 states with HCFA hospice claims data from 1991 to 1997 to identify nursing facility residents who had a hospice episode. Nursing facility admission was determined from the dates of admission on the MDS record in a patients' file and initiation of the Hospice Benefit was determined based upon the date of service field in the first Medicare hospice claim filed on behalf of the patient. A total of 11,964 residents with an MDS record had at least one hospice episode either before, overlapping with, or after nursing facility admission. This represents approximately 6 percent of all unduplicated nursing facility residents in the 5 states in the time period of 1992 to 1996. Based on the dates of admission and Hospice service, hospice patients were classified as:

  1. Initiating and stopping Hospice prior to nursing facility stay (pre);
  2. Hospice overlapped with the nursing facility stay (overlap);
  3. Hospice election followed nursing facility admission (post);

Ninety-five percent of the residents had only one hospice episode (N=11,376). For those residents with multiple episodes (N=588), some episodes were separated by more than 6 months. There were some hospices beneficiaries who represented a combination of the three types enumerated above. The decision rules used to assign those individuals who met multiple categories are summarized in detail in Table 1. Applying these rules, 1119 (9 percent) of nursing facility residents had "pre" hospice nursing facility care, 1546 (13 percent) had "overlap" hospice nursing facility care and 9299 (78 percent) had "post" hospice nursing facility care.

Measures

In conducting descriptive analyses of hospice beneficiaries with nursing facility stays, we used the MDS resident assessment record closest in time to the residents' hospice admission date as the basis for describing the characteristics of the individual. In those cases in which first hospice admission precedes, or overlaps with, the nursing facility episode, the admission MDS assessment was used.

In addition to standard demographic and clinical variables that are enumerated in Appendix A (Table A1), the variables used to describe hospice beneficiaries in nursing homes include:

  1. temporal category of hospice care (pre, overlap or post),
  2. length of hospice stay by length of stay categories,
  3. the principal diagnosis from first hospice claim,
  4. whether skilled nursing facility (SNF) episode occurred prior to hospice admission,
  5. distance between SNF episode and hospice admission, and
  6. Medicare inpatient, SNF and home health care expenditures 180 days prior to hospice admission.

Facility Variables. An earlier study of hospice in nursing homes that used HCFA's OSCAR data found that facility factors clearly differentiated those nursing homes with few or no residents receiving hospice care from those in which numerous residents received hospice (Petrisek & Mor, 1999). Consequentially, we constructed one contextual variable, hospice concentration. This represents a ratio of the total number of unduplicated hospice patients in a nursing facility in a one year period to the total number of unduplicated residents in that nursing facility during the same time period.

Conceptually, facility concentration should reflect both the length of the hospice-nursing facility relationship and the success of the relationship. It should be reflective of a nursing facility's greater accommodation to hospice care management practices and hospice philosophies. With this greater accommodation, the presence of policies, procedures and practices that facilitate and enhance the coordination of care with hospice are more likely. With this in mind, we examine how hospice patient characteristics and service utilization vary across nursing facilities with differing levels of hospice concentration.

Methods of Analysis

Univariate analyses were conducted using the values recorded on the MDS as well as the relevant utilization and expenditure measures. When stratifying by variables of interest, resident and facility level, percents and measures of central tendency (i.e., means, medians, and modes) are used to describe hospice patients and their service utilization.

RESULTS

Estimate of the Scope of Use of Medicare Hospice in Nursing Homes

Once nursing facility MDS assessment records had been matched to Medicare's enrollment files, we could be confident that all those with a Medicare Hospice record who also had an MDS record were Medicare Hospice beneficiaries who had been in a nursing facility. We identified a total of 5053 nursing facility patients (both "overlap" and "post" patients) who were actively enrolled in hospice in 1996. Obviously, these figures varied widely by state [Maine, 180; Mississippi 197; New York, 3,103; South Dakota, 217; Kansas, 1,356]. We inflated these numbers to adjust for the percentage of unduplicated patients in each state with an MDS in 1996 who we were unable to match to Medicare's enrollment file; an average of 13 percent of 1996 nursing facility patients could not be matched. The percent of patients with no match to the enrollment file varied by state [Maine, 7 percent; Mississippi, 26 percent; New York, 11 percent; South Dakota, 17 percent; Kansas, 15 percent]. After inflating estimates by the percent of unmatched patients per state, we compared these estimates to the state numbers of hospice enrollees served in 1996 (N=23,897) (See Table 8, in Medicare's Hospice Benefit: Use and Expenditures of this study). Overall, those hospice beneficiaries estimated to have resided in nursing facilities represent 24 percent of the total 1996 hospice enrollees in the five study states. However, estimates varied greatly by state with 46 percent and 48 percent of Medicare hospice beneficiaries estimated to be nursing facility residents in South Dakota and Kansas, but only 11 percent in Mississippi. Twenty percent of 1996 New York hospice beneficiaries are estimated to have resided in nursing facilities, and in Maine 27 percent. If the overall estimate of 24 percent could be applied to all states the connection between hospice and nursing facilities nationally would be considered substantial. Nonetheless, in the five study states our estimates reveal the provision of Medicare hospice care in nursing facilities to be an important model for hospice care provision.

Comparison of Pre, Overlap and Post Hospice Admissions

Table 2 summarizes the percent distribution of the three groups of beneficiaries from 1992 to 1996. As is evident, the "post" group has come to represent the predominant hospice nursing facility beneficiary. This is partially due to censoring (fewer pre-cases having died in later years) but predominantly due to a proportional increase in the number of "post" cases. Key characteristics of residents who had received Medicare hospice are compared below by the pre, overlap and post categories.

As is evident from Tables 3-5, hospice patients admitted after nursing facility admission more closely resemble the long-term care nursing facility population than do the other two hospice patient groups. "Post" patients are more often 85 years of age and older (48 percent versus 29 percent and 26 percent for "pre" and "overlap" patients), female (70 percent versus 56 percent and 58 percent), and widowed or divorced (68 percent versus 57 percent and 58 percent for "pre" and "overlap" patients) (Table 3). Additionally, "post" patients have cancer diagnoses less frequently recorded as the principal diagnosis than do "pre" or "overlap" patients (38 percent versus 62 percent and 72 percent respectively) (Table 4). No differences in racial/ethnic group representation across these 3 groups are observed (Table 3).

Patients with hospice admissions prior to, or overlapping with, their nursing facility admissions more closely resemble hospice patients cared for in the community. For "pre" and "overlap" patients, the most frequent terminal diagnosis, per principal diagnosis noted on the first hospice claim, is lung cancer while heart failure is the most frequent diagnosis for patients admitted to hospice after nursing facility admission. Together, Alzheimer's disease and dementia represent 9 percent of the principal diagnoses for "post" patients while these diagnoses only represent 2 percent of the principal diagnoses for "pre" patients and 1 percent for "overlap" patients (Table 4).

Using diagnoses from the MDS assessment and from inpatient claims, 59 percent of "post" patients had a diagnosis of cancer (not restricted to principal diagnosis) while 73 percent of "pre" and 77 percent of "overlap" patients had a diagnosis of cancer recorded (Table 5).

The average and median lengths of hospice stay are considerably influenced by the temporal sequence of hospice admission (Table 6). The "overlap" hospice patients, who begin their hospice episode in the community and then either die or are discharged from hospice while in a nursing facility, have the longest lengths of hospice stay, with an average of 166.9 days, a median of 92 days, and a mode of 26 days. The "pre" patients, who survive their hospice episode and are later admitted to a nursing facility, have an average length of stay approximately 50 percent longer than do the "post" patients (108.9 days versus 77.9 days) and a median length of stay 2 times as long (61 days versus 28 days) (Table 6).

Description of "Post" Nursing Facility Hospice Patients

Sociodemographic Characteristics

Descriptive statistics for sociodemographic characteristics of "post" patients, stratified by variables of interest, are shown in Tables 7-10. Hospice patients with dementia and no cancer are the oldest patients, with 63 percent being 85 years of age or older (Table 7). Patients with cancer and no dementia are the youngest patients, with only 33 percent being 85 years of age or older (Table 7). Hospice patients in nursing facilities with higher hospice concentrations are consistently older than are patients in facilities with lower hospice concentrations (Table 8). The age distribution is quite similar for hospice discharges and deaths, although patients who die are slightly older (Table 9). Age distributions vary by state, with New York and Kansas (states with the highest hospice volume) having the oldest hospice patients and Mississippi the youngest patients (Table 10).

Females represent almost 70 percent of patients who are admitted to hospice after nursing facility admission. A higher proportion of females (versus males) comprise those patients having dementia (with no cancer) and those patients with "other" diagnoses (80 percent and 73 percent respectively) while a smaller proportion of females comprise those patients having cancer without dementia and cancer with dementia (64 percent and 67 percent respectively) (Table 7). Also, higher proportions of female hospice patients are observed in nursing facilities with higher hospice concentrations than in those with lower concentrations (Table 8). Females represent 74 percent of patients discharged from hospice while they represent only 69 percent of patients who die on hospice (Table 9). Last, females comprise higher proportions of hospice patients in states where the volume of hospice nursing facility patients is greater (Table 10). This makes intuitive sense since nursing facilities having higher concentrations of hospice patients are more likely to be present in high volume states.

Only 6 percent of hospice patients electing after nursing facility admission are non-white. Minorities represent a greater proportion of hospice patients with cancer (approximately 6 percent) than they do hospice patients with dementia and no cancer (5 percent) and hospice patients with no dementia or cancer (4 percent) (Table 7). As a nursing facility's hospice concentration increases, the proportion of minority hospice patients decreases (Table 8). Additionally, a higher proportion of minority patients are discharged from hospice (8 percent) than die in hospice (5 percent) (Table 9). Last, as expected, the proportions of minority hospice patients across states vary by the underlying minority populations in each state, with Mississippi having the largest proportion of minority hospice patients (28 percent) and Maine and South Dakota the smallest (2 percent and 1 percent respectively) (Table 10)

Overall, 76 percent of hospice patients admitted after nursing facility admission are widowed, separated or divorced, or never married. When stratifying by the variables of interest, similar trends in the proportions of unmarried persons across variable categories are seen as are observed (and discussed above) for the proportions of female hospice patients (Table 7, Table 8 and Table 10). However, whether a patient is discharged from or dies in hospice varies little by marital status (Table 9).

Diagnoses

Overall, 59 percent of the "post" patients have a cancer diagnosis. As a facility's hospice concentration increases, the proportion of patients with cancer diagnoses decreases and the proportion with dementia, with or without cancer, increases (Table 11). Cancer patients (without dementia) are least frequently discharged from hospice (11 percent) while dementia patients (without cancer) are most frequently discharged from hospice (24 percent) (Table 12). New York and Kansas, the states with the greatest number of hospice nursing facility patients, also have the smallest proportion of hospice patients with cancer diagnoses (57 percent and 61 percent respectively), and the largest proportion of patients with dementia and no cancer (19 percent and 14 percent respectively) (Table 13). As shown in Table 14, the proportion of patients in each of the four diagnosis groups has changed little since 1993.

Clinical Characteristics

Descriptive statistics for limitations in activities of daily living (ADLs), cognitive performance and symptom assessment, stratified by variables of interest, are shown on Tables 15-18. It should be emphasized again that the MDS resident assessment closest to hospice admission was the source for these data. Hospice patients having greater limitations in ADL's and more cognitive impairment are those patients with dementia (with and without cancer) (Table 15). Also, in nursing facilities with greater hospice concentrations and in the states of New York and Kansas (states with greater hospice numbers) hospice patients have more limitations in ADLs and more cognitive impairment than do patients in low hospice concentration facilities and in the other study states (Table 16 and Table 18). Activity of daily living limitations and the level of cognitive impairment did not appear to differ when considering whether a hospice patient was discharged from or died on hospice (Table 17).

When considering the symptom prevalence for patients in the 4 diagnosis groups, a higher proportion of patients with cancer without dementia had pain and vomiting documented (41 percent and 7 percent respectively) and a higher proportion of patients with "other" diagnoses had shortness of breath (dyspnea) documented (23 percent) (Table 15). Patients in the diagnosis group of "other" also had a higher prevalence of persistent mood disturbance documented (16 percent) (Table 15). Patients residing in nursing facilities with less than 9 percent hospice concentration had a higher prevalence of pain and vomiting documented than did those patients in facilities with 9+ percent hospice concentration (Table 16). (As noted previously, facilities with lower hospice concentrations are more likely to have patients with cancer and no dementia.) (Table 11) Patients dying in hospice manifested more symptoms than did patients who were discharged alive from hospice (Table 17). Documented symptom prevalence differed by state, but we observe no consistent pattern. Hospice nursing facility patients in New York, 52 percent of who have dementia, had the lowest documented prevalence of pain, dyspnea and vomiting (Table 18).

Utilization of Medicare Hospice Benefit and Related Utilization by "Post" Hospice Patients

Lengths of Hospice Stay

The mean and median lengths of stay have decreased over time for hospice patients first using the benefit after nursing facility admission (Table 19). For example, in 1994 the mean length of stay was 93 days and the median length of stay 31 days, while in 1996 the mean was 65.3 days and the median 25 days. There is some time based censoring affecting these statistics (e.g., more recent hospice admissions do not have as much opportunity to remain in hospice for more than 210 days). However, for the most part these changes are the result of the changing length of stay distribution which increasingly includes stays of one week or less but includes fewer stays of over 210 days (Table 19). The strongest trend is the increase in the number of very short stays. Seven percent of hospice patients admitted after nursing facility admission are admitted for 2 days or less (data not shown), and almost one-fourth are admitted for 7 days or less (Table 19). As with the stays of 7 days or less (Table 19), the proportion of stays of 2 days or less has increased over time; it was 3 percent in 1992, 6 percent in 1993 and 8 percent in 1996 (data not shown). Compared to hospice patients with cancer and no dementia, hospice patients with dementia (with or without cancer) and with other non-cancer diagnoses have a greater percentage of stays of 7 days or less. In addition, patients with dementia (with or without cancer) also have a higher percentage of long stays than do patients with cancer diagnoses without dementia (Table 20).

Lengths of stay differ dramatically when considering the concentration of hospice patients within a nursing facility. Mean and median lengths of stay, as well as the proportion of patients with long lengths of stay, increase as a nursing facility's hospice concentration increases (Table 21). In nursing facilities with a 13+ percent hospice concentration the mean length of stay is 97.6 days and the median is 35 days, while in nursing facilities with a <2 percent hospice concentration the mean length of stay is 65.6 days and the median is 24 days (Table 22). When controlling for diagnosis group, the observed differences by hospice concentration are found to be largely attributable to patients with dementia, with and without cancer (data not shown). That is, the observed length of stay differences are due to the fact that more dementia patients are treated as hospice patients in facilities with many hospice patients than is the case in facilities in which hospice is less prevalent. For dementia patients, a greater nursing facility hospice concentration is consistently related to a higher proportion of lengths of stay being over 180 days (data not shown). In addition, while a consistent linear trend was not observed, nursing facilities having higher hospice concentrations tended to have a smaller proportion of dementia patients with lengths of stay of one week or less (data not shown).

Dramatic differences in length of stay statistics are also observed between hospice patients who die versus those who are discharged alive from hospice (Table 22). The mean length of stay for hospice discharges is almost 3 times as large as that for hospice deaths (175.7 days versus 60.8 days), and the median is 6½ times as large (129 versus 20) (Table 22).

Differences in length of stay statistics and distributions are also observed across states. Mean lengths of stay range from 76.6 days in Kansas to 84.3 days in Mississippi and median lengths of stay range from 25 days in South Dakota to 34 days in Maine (Table 23).

Few differences in length of stay distributions are observed across hospice organizational types (freestanding or hospital, home health agency of skilled nursing facility based hospices) (data not shown). However, freestanding hospice programs had a slightly lower percentage of patients with stays of one week or less (22 percent) than did hospital-based (26 percent), skilled nursing facility-based (28 percent) and home health agency based hospices (26 percent) (data not shown).

Hospice ownership in study states appears to influence lengths of stay. Nursing home patients cared for by for-profit hospice organizations have a mean length of stay of 84.7 days and a median of 33 days while those in non-profit organizations have a mean of 78.3 and a median of 27, and in government owned organizations a mean of 45.5 and a median of 18.5 days (Table 24). For-profit organizations care for a higher proportion of dementia patients (with or without cancer) and a lower proportion of cancer patients (without dementia) (data not shown). However, when stratifying by diagnosis groups, lengths of stay in for-profit organizations continue to be longer across all diagnosis groups except cancer and no dementia where the length of stay is shorter (data not shown). It is important to note that the vast majority (90 percent) of nursing facility hospice patients studied here are cared for by non-profit hospices; 7 percent and 2 percent respectively are cared for by for-profit and government owned hospices.

Hospice Expenditures

Table 25 shows the use of hospice care by nursing facility hospice patients, stratified by the four diagnosis groups. Ninety-four percent of hospice nursing facility patients use routine home care, while much smaller proportions use continuous care (3 percent), general hospice inpatient care (11 percent), inpatient respite care (1 percent), or physician visits (4 percent). Of note is that some of the use of respite care was during a hospice episode occurring prior to nursing facility admission and more than 6 months before the "post" hospice episode (see Table 1), and some of the use observed is probably attributed to faulty claims data. Because of the high use of routine home care versus the other levels of hospice care, total hospice expenditures primarily reflect routine home care expenditures and, therefore, length of stay (Table 26). The 26 percent of hospice patients with lengths of stay of one week or less have mean total hospice expenditures of $550 and median expenditures of $391 while the 9 percent of hospice patients with stays of over 210 days have mean expenditures of $36,421 and median expenditures of $32,250. Hospice expenditures overall average $7,848 per patient, with a median expenditure of $3,093 (Table 26).

Hospice nursing facility patients with diagnoses other than cancer and/or dementia have the lowest expenditures, with a mean of $6,970 and a median of $2,587 (Table 27). Of note, and as shown on Table 21, these patients, at 22 days, have the lowest median length of stay by diagnosis group. Compared to patients with dementia and/or cancer, patients with other diagnoses have more stays of 2 weeks or less (42 percent) and fewer stays of over 210 days (7.9 percent) (Table 20). Patients with dementia and no cancer have the longest mean and median lengths of stay (Table 20) and the highest expenditures (Table 27), with mean hospice expenditures of $9,175 and median expenditures of $4,069.

As with lengths of stay, expenditures differ dramatically by whether a hospice episode ends in death versus discharge. Mean expenditures for the patients discharged alive from hospice are $17,222 versus $6,214 for those who die on hospice, and median expenditures are $12,924 versus $2,383 (Table 28).

Skilled Nursing Facility Care and Hospice Admission

Seventy-three percent (6,827) of "post" hospice nursing facility patients has a skilled nursing facility (SNF) episode at some point prior to hospice admission. Nursing facilities with lower hospice concentrations had a slightly higher percentage of patients admitted to hospice following a SNF discharge. In facilities with a <5 percent hospice concentration, 76 to 77 percent of hospice patients had a SNF discharge while in facilities with a 5+ percent hospice concentration this percentage is approximately 70 percent (data not shown).

The median time between SNF discharge and hospice admission is 46 days, and the mean is 222 days. The mode number of days between SNF discharge and hospice admission is 0. Twenty-six percent (1,772) of hospice admissions that follow SNF discharge occur within 0 to 1 days of SNF discharge (Table 29). Of these admissions, 34 percent (637) follow a SNF episode of 100 days (the maximum number of reimbursable days for an individual SNF episode) (data not shown). A higher proportion of patients in government owned nursing facilities experience a hospice admission immediately following a SNF discharge (34 percent) than do patients in non-profit facilities (30 percent) or in for-profit nursing facilities (20 percent) (Table 29). While there is not substantial variation across diagnosis groups, hospice admission within 0 to 1 days of SNF discharge is experienced by a higher proportion of patients with cancer and no dementia (28 percent) and dementia and no cancer (30.8 percent) than by patients with cancer and dementia (24 percent) and patients with other diagnoses (24 percent) (Table 30).

The proportion of patients with a shorter time between SNF discharge and hospice admission is greater for patients residing in nursing facilities with a higher hospice concentration (Table 31). In facilities falling into the highest hospice concentration group (13+ percent), 43 percent of hospice admissions following SNF discharge occur within 0 to 1 days. Conversely, in facilities with the lowest hospice concentration (<2 percent), only 17 percent of the hospice admissions following SNF discharge occur within 0 to 1 days (Table 31). When controlling for diagnosis groups, the differences by hospice concentration remain, and generally at the same magnitude of difference across concentration groups, as shown in Table 31.

Medicare Expenditures in Six Months Prior to Hospice Admission

Overall, Medicare expenditures in the six months prior to hospice admission average (in 1996 dollars) $10,206, with a median of $6,114 (Table 32). Hospital inpatient expenditures contribute to 72 percent of the overall mean expenditures, and to 69 percent of median expenditures. The .4 percent of hospice patients cared for by skilled nursing facility based hospices incurred the highest mean ($20,825) and median ($19,583) Medicare expenditure in the six months prior to hospice admission. Patients cared for by home health agency based hospices (38 percent of total patients) incurred the lowest mean ($8,755) and median ($4,836) expenditures in the six months prior to hospice admission.

DISCUSSION

The data presented in this report represent the first comprehensive description of the population of patients using the Medicare hospice benefit while residing in nursing facilities. By matching nursing home MDS data obtained from computerized resident assessments with Medicare hospice claims data it was possible to describe the different types of populations of nursing home residents served by hospice and the types of services they receive. Several very important insights emerge from these analyses which raise important policy issues regarding the way in which hospice care is provided to persons living in nursing facilities. The paragraphs which follow examine each of the following issues; the scope of the hospice in nursing home phenomenon, the different pathways to hospice and nursing home care, the implications of the observed hospice length of stay distribution in the nursing home and its similarity to that observed for the program in general, and the limited use of any services other than routine home care, meaning that hospice costs are almost totally a function of length of hospice stay. Finally, we consider the policy implications of our finding that some nursing facilities have a much higher concentration of hospice beneficiaries than do others.

Scope of Hospice in Nursing Homes

In the five states studied we estimate that 24 percent of all 1996 hospice enrollees resided in nursing facilities while receiving Medicare hospice care. Nationally, however, we know that the concentration of hospice in nursing facilities is substantially higher in several other states such as Florida and California (Petrisek and Mor, 1999). Even in the five states we studied we found the range of 1996 Medicare hospice beneficiaries residing in nursing facilities to be between 11 percent and 48 percent. Whether this range and the 24 percent estimated overall percent reflect current national proportions is unknown.

Obviously, the only way to really know the scope of hospice care provided in the nursing facility setting is to match the newly available national repository of MDS data to the most recent files of Medicare hospice claims. Nonetheless, whether it is one-quarter, or even one-half of all Medicare hospice beneficiaries that are served in nursing facilities, it is clear that this setting is a major model in the provision of hospice care to Medicare beneficiaries.

Different Types of Hospice in Nursing Facility Patients

We found that the demographics, diagnoses and hospice utilization differ markedly between hospice patients completing a hospice episode prior to nursing facility admission ("pre"), patients having a hospice episode overlapping with facility admission ("overlap"), and those patients admitted to hospice after nursing facility admission ("post"). "Post" patients represent the vast majority of nursing facility Medicare beneficiaries (86 percent in 1996). In general, the characteristics and diagnoses of these patients more closely resemble long-term care nursing facility residents. Hospice lengths of stay are shortest for "post" patients and longer for "pre" and "overlap" patients. This makes intuitive sense as both "pre" and "overlap" patients have much smaller proportions of short stays than do "post" patients -- "pre" patients have survived a hospice episode in the community to later be admitted to a nursing facility and "overlap" patients have begun a hospice episode in the community to later be admitted to a nursing facility, presumably due to caregiver exhaustion. Both the "pre" and the "overlap" types of patients are "outliers" from the community based model of hospice care. In an earlier era many of these patients might have been hospitalized for extended periods, but this tends not to be an option in most US communities. In thinking of the whole policy issue of the provision of hospice care in the nursing facility setting, it is probably important to differentiate these two less prevalent classes of patients from those who initiated hospice care after already having been a nursing facility resident.

Diagnoses of "post" patients are associated with those that characterize the typical long stay nursing home resident. We found that females represent 70 percent of nursing facility hospice patients, a rate substantially higher than the 53 percent observed by Banaszak-Holl and Mor (1996) in skilled nursing facility-based hospices and greater than the 50 percent observed by the 1996 National Home and Hospice Care Survey (National Center for Health Statistics, 1998), and more than the 53 percent of females in hospice in the community as observed by Gage and colleagues in Medicare's Hospice Benefit: Use and Expenditures of this study. It is likely that these earlier studies were unable to differentiate among the different types of hospice patients receiving care in nursing facilities.

The proportion of African Americans (4 percent) and other minorities (1 percent) who enroll in Medicare hospice in nursing facilities is lower than that observed in the community (7 percent and 3 percent respectively) (See Medicare's Hospice Benefit: Use and Expenditures in this study). It is well known that the proportion of the minority elderly population residing in nursing facilities is lower than the proportion of the white population, and this may very well account for much of the observed difference in non-white hospice enrollment in nursing facilities. However, our data support the notion that there is differential election of hospice by race since non-whites represent 8 percent of the non-hospice comparative sample and only 5 percent of the hospice comparative sample (See Outcomes and Utilization for Hospice and Non-Hospice Nursing Facility Decedents in this study). Given that nursing facility staff are considered the family/primary caregiver for hospice nursing facility patients, the observation of lower non-white enrollment in nursing facilities, suggests that reasons other than, or in addition to, the lack of a primary caregiver in the community (Gordon, 1995) account for low minority hospice enrollment in the community. It must be noted that while non-white enrollment in the community is based upon on all U.S. states, our data is reflective of the 5 study states. Across these states we observe substantial variation in minority enrollment, primarily related to underlying population differences. For example, in Mississippi 28 percent of hospice nursing facility patients represent minority populations while in South Dakota, a state with a lower minority population, only 1 percent of hospice patients is non-white. Furthermore, the concentration of hospice patients varies substantially across facilities, as does the proportion of non-white residents. Perhaps those facilities caring for higher proportions of non-white residents are less likely to have established relationships with a hospice provider to serve their terminal patients.

Length of Hospice Stay

The 78 day average length of stay we observe for "post" hospice patients in nursing facilities is less than half of the 181 average length of stay observed by the Office of Inspector General (OIG) in its study of Hospice Patients in Nursing Homes (U.S. Department of Health and Human Services, 1997). (Comparisons of the median length of stay, which is less influenced by extreme values, cannot be made since the OIG did not report the median length of stay. Also, the OIG did not report on the percentage of stays over 210 days.) From our data, it is clear that the OIG's focus on prevalent hospice patients (i.e., those receiving hospice services in December 1995) resulted in its studying a disproportionate number of longer stay hospice patients since short stay patients are less likely to be observed in this short period of time. This statement is supported by analysis we did using our data and examining average length of stay for hospice patients in nursing homes in December 1995 (N=1103). We found that this prevalence sample of patients had an average length of hospice stay of 222 days (SD=224), almost 3 times as long as we know the average length of stay to be by using length of stay data from the entire population of patients (data not shown).

The length of stay distribution for hospice nursing facility patients is similar to the length of stay distributions for hospice patients in the community and to hospice length of stay distributions observed over time. In fact, the length of stay distribution observed here is similar to that observed in the National Hospice Study (NHS) (Mor et al., 1988). In the National Hospice Study (NHS), 53 percent of the demonstration patients in home-based hospice had stays of 35 days or less and 61 percent in hospital-based hospices had stays of 35 days or less. In our study, we find that 52 percent of the nursing facility hospice patients have stays of 30 days or less. Additionally, the proportion of NHS demonstration hospital-based patients experiencing a hospice stay of 7 days or less (26 percent) is very comparable to the percentage of hospice stays of 7 days or less we observe in nursing facilities (24 percent overall and 26 percent in 1996). Even though the NHS demonstration patients had a lower proportion of hospice stays of 210 days or longer (5 percent of patients in home care-based hospices and 4 percent of patients in hospital-based hospices) than the proportion we observe for nursing facility hospice patients (9 percent), the length of stay distributions are still quite similar. It should be noted that the comparability of these length of stay distributions exists in spite of the fact that the reimbursement systems under which the NHS demonstration occurred differed from the prospective, per diem model that is in use today. The similarities of hospice length of stay distributions over time and across hospice care settings speak to the difficulties of accurately selecting terminal patients for hospice. From our data, this selection appears no more or less easy in nursing facilities than it is in community hospice populations. However, in nursing facilities this selection is made more difficult because of the coordination between the hospice and nursing facility needed to accomplish a hospice admission. Additionally, contributing to the difficulty of an accurate prognosis in the nursing facility is the fact that larger percentages of dying patients in nursing homes have non-cancer diagnoses, for which accurate prognoses are more difficult (Christakis and Escarce, 1996; Luchins et al., 1997; Lynn et al., 1997a).

The difficulty in selecting terminal patients for hospice makes hospice programs highly subject to exogenous "shocks" relating to their selection processes. For example, the decrease in hospice median lengths of stay and the increase in the percentage of short stays in nursing facilities observed in our study may be a result of the 1995 OIG study and its questioning of hospice eligibility and lengths of stay (U.S. Department of Health and Human Services, 1997). As more caution in the selection process is exercised, short lengths of stay will increase and long lengths of stay will decrease. For example, in 1994 the median hospice length of stay was 31 days with 23 percent of the patients having stays of 7 days or less and 12 percent having stays of over 210 days. In 1996, after the OIG study and release of a related directive by the OIG, the median length of stay was 25 days with 26 percent of patients having stays of 7 days or less and 7 percent having stays of over 210 days.

For patients with dementia, with or without cancer, nursing facilities with higher hospice concentrations tend to have proportionately fewer short lengths of stay (of 7 days or less) and a greater proportion of long lengths of stay (of over 180 days) (data not shown). We speculate that these differences are a product of nursing facilities with higher concentrations having more practice in caring for hospice dementia patients. As such, these facilities may have better procedures and practices in place to help them to avoid "late admissions" to hospice.

The presence of a modal length of stay of 2 days in this population represents both a policy and clinical care concern. In light of clinical inertia and chronic quality of care problems that are endemic in nursing homes, aggressive approaches to symptom management are needed if hospice is to positively influence the quality of the patient's death. Analyses of patients with stays of 2 days or less show mean hospice expenditures to be only $249, and median expenditures to be only $188 (data not shown). Caring for these very short stay patients may be both demoralizing and costly for hospice programs. Enrolling patients, conducting the initial assessment and care plan and ordering the prescribed treatment regime entail substantial resource expenditures and all of these activities must occur regardless of how long the patient remains under the care of the hospice. It is doubtful whether the reimbursements from the relatively small number of long-stay patients compensate hospices for the losses incurred through the care of these late hospice admissions. However, there is no currently available cost data to examine this issue but this may be resolved in the future as hospice cost data are now being reported to HCFA.

Concentration of Hospice Care in Selected Nursing Facilities

In our sample of states and nursing facilities, we found that some facilities had a relatively high proportion of their decedents cared for under the Medicare hospice benefit, while many others had only a small proportion of their residents under hospice care. From a management and operational efficiency perspective, it would seem reasonable that those organizations that have established a modus operandi for working with a hospice would gain if more eligible residents elected the hospice benefit. In other areas of health services research, ranging from conducting open heart surgery to managing patients with complex HIV disease on an out-patient basis, "practice makes perfect" is a well established generality of medical care. Indeed, there are indications that those facilities with a higher concentration of hospice patients are somewhat less plagued with admissions of very short stay patients, although the modal length of stay even in these facilities is 7 days or less. Furthermore, these facilities are somewhat more likely to enroll traditional, demented, long stay patients onto the hospice benefit than are facilities that are less practiced. This is observed both in the diagnostic distributions as well as the fact that hospice enrollees in facilities with a higher concentration of hospice patients are more physically impaired in ADL functioning and in cognitive functioning.

Hospice Admission Following SNF Discharge

Of the 73 percent of hospice patients who experience a skilled nursing facility (SNF) discharge prior to hospice admission, 26 percent are admitted to hospice within 1 day of SNF discharge, suggesting that hospice care may have been appropriate earlier if it were available to patients in SNF beds. Of note, however, is that patients admitted to hospice within one day of SNF discharge (even when the SNF episode was 100 days) did not appear to die earlier than patients with longer times between SNF discharge and hospice admission (data not shown). As discussed in the literature review (See Important Questions for Hospice in the Next Century in this study), many have expressed concern that financial considerations, both by nursing facility administrators and patients/families, limit hospice access to dying patients who also qualify for SNF care. Patients cannot access both hospice and SNF care when the need for SNF care is related to the terminal condition (most commonly the case). Therefore, SNF care may be "encouraged" and chosen rather than hospice care because residence in a SNF bed results in higher payment to the nursing home (than payment by Medicaid for dual eligible residents) and less out-of-pocket payment by families of private pay patients (since room and board would not be covered if hospice is chosen).

The high percentage of hospice admissions immediately after SNF discharge may speak to delays in hospice referral because of financial concerns and/or it may reflect the presence of coordinated practices between hospices and nursing facilities to assure dying patients receive hospice as quickly as possibly after SNF discharge. Supporting this last notion is the finding that nursing facilities with the largest hospice concentrations (13+ percent), who should have the most advanced procedures and practices in place for coordinating care with hospice, have the highest percentage of hospice patients admitted within 1 day of SNF discharge (43 percent). Whether this behavior suggests maximizing reimbursement or minimizing family out of pocket expenditures by legally and legitimately concatenating reimbursable Medicare coverage across care settings is probably not knowable. Since SNF stays are inevitably preceded by a hospitalization, it may well be that this event begins to trigger a realization that the patient may be entering a final "terminal spiral". In the past, this process would have occurred in the hospital. With increasingly shorter stays, this drama is being played out across institutional boundaries between the hospital and the nursing facility. Given the very short lengths of stay, it is likely that the hospice is brought into the fray at the last minute. To really understand this phenomenon and the quality of care implications for the terminally ill, more detailed, records based studies will be required.

Study Limitations

Several limitations to this study are noted. First, although it is likely that hospice may have similar effects in states not included in this study, the results presented here are not necessarily generalizable to states other than Kansas, Maine, Mississippi, New York and South Dakota. We know from analyses of HCFA's nursing home level OSCAR data that the states included in our study are not necessarily those where hospice in nursing homes is most prevalent. Florida and Oklahoma, where this phenomenon is most prevalent, may represent very different approaches to the management of hospice care in nursing homes. While we did not see major differences as a function of the type of hospice that was serving the nursing home residents, the complex inter-relationship between states' Medicaid policies, Medicaid nursing home payment rates and even the level of penetration of Medicare managed care, might reasonably influence the behavior of hospitals, nursing facilities and hospices.

Second, although the MDS assessments used for this study are data rich, there is still substantial variability in the quality of MDS clinical data from facility to facility. Furthermore, we do not know the extent to which the MDS we used to characterize the residents around the time of hospice election reflects the input of the hospice assessment. Since the nursing home regulations that govern the conduct of the assessment on which the MDS is based do not explicitly take into consideration the input of care teams from outside the facility, it may be the case that the MDS does not reflect the input, and perhaps even the outcomes, of the hospice care staff and their interventions. This is an area that should be examined in considerable depth, in light of the scope of the hospice in nursing home phenomenon and the importance of integrating care planning between the regular facility nursing staff and the hospice staff.

CONCLUSION

In summary, we find that hospice in nursing home is a very prevalent phenomenon that appears to have many characteristics in common with both hospice in the community and the provision of regular long term care services to dying residents in the nursing home setting. As with most health services, the phenomenon is concentrated with some facilities having a relatively high proportion of their decedents served by hospice whereas others have only a few. To understand the costs and benefits of these arrangements requires a comparison of those between similar decedents who did and did not receive hospice care. This issue is addressed in the next report of this study.

REFERENCES

Banaszak-Holl, J., and Mor, V. Differences in patient demographics and expenditures among Medicare hospice providers. The Hospice Journal. 1996; 11(3):1-19.

Christakis, N. A., and Escarce, J. J. Survival of Medicare patients after enrollment in hospice programs. The New England Journal of Medicine/Special Article. 1996; 335:172-8.

Gage, B., and Dao, T. (2000). Medicine's Hospice Benefit: Use and Expenditures. In: Synthesis and Analysis of Medicine's Hospice Benefit. Project Report for Assistant Secretary for Planning and Evaluation, Washington, DC: U.S. Department for Health and Human Services.

Gordon, A. K. Deterrents to access and service for Blacks and Hispanics: The Medicare hospice benefit, healthcare utilization, and cultural barriers. The Hospice Journal. 1995; 19(2):65-83.

Health Care Financing Administration (HCFA). Minimum data se plus multistate nursing home case mix and quality demonstration training manual. Natick, MA: Eliot Press; 1991.

Luchins, D. J.; Hanrahan, P., and Murphy, K. Criteria for enrolling dementia patients in hospice. Journal of the American Geriatrics Society. 1997; 45:1054-1059.

Lynn, J.; Harrell, F. E. Jr.; Cohn F., et al. Prognoses of seriously ill hospitalized patients on the days before death: Implications for patient care and public policy. New Horizons. 1997a; 5:56-61.

Mor, V.; Greer, D., and Kastenbaum, R. (eds.). The hospice experiment. Baltimore, MD: John Hopkins University Press, 1988.

National Center for Health Statistics. An overview of home health and hospice care patients: 1996 National Home and Hospice Care Survey. (Advance Data from Vital and Health Statistics: No. 297). Hyattsville, MD: Haupt, B., 1998.

Petrisek, A. & Mor, V. (1998). Hospice in nursing homes: A facility level analysis of the distribution of hospice beneficiaries. The Gerontologist, 39, 279-290.U.S. Department of Health and Human Services. Hospice patients in nursing homes. (DHHS Publication No. OEI-05-95-00250): June Gibbs Brown, Inspector General, 1997.

U.S. Department of Health and Human Services. Hospice patients in nursing homes. (DHHS Publication No. OEI- 05-95-00250): June Gibbs Brown, Inspector General, 1997.

TABLE 1. Classification of Hospice Care Temporal Categories and Identification of Nursing Facility Hospice
Number of Hospice Episodes Temporal Location of Hospice Episode(s) in Relation to Nursing Facility Stay* 6-Month Information Applicable to Decision on Classification Assigned Temporal Category Nursing Facility Hospice Admission Date
1 Pre NA Pre Episode adm. date
1 Overlap NA Overlap Episode adm. date
1 Post NA Post Episode adm. date
2 2 Pre NA Pre 1st episode adm. date
2 2 Post NA Post 1st episode adm. date
2 1 Pre, 1 Overlap <6 months between episodes Overlap 1st episode adm. date
6+ months between episodes Overlap 2nd episode adm. date
2 1 Pre, 1 Post <6 months between episodes Overlap 1st episode adm. date
6+ months between episodes Post 2nd episode adm. date
2 1 Overlap, 1 Post NA Overlap 1st episode adm. date
3 3 Pre NA Pre 1st episode adm. date
3 3 Post NA Pre 1st episode adm. date
3 2 Pre, 1 Overlap <6 months bet. all episodes Overlap 1st episode adm. date
6+ months bet. 2 pre episodes but not bet. last pre and overlap episodes Overlap 2nd episode adm. date
6+ months bet. last pre and overlap episodes Overlap 3rd episode adm. date
3 2 Pre, 1 Post <6 months bet. all episodes Overlap 1st episode adm. date
6+ months bet. 2 pre episodes but not bet. last pre and post episodes Overlap 2nd episode adm. date
6+ months bet. last pre and overlap episodes Post 3rd episode adm. date
3 1 Pre, 1 Overlap, 1 Post <6 months bet. all episodes Overlap 1st episode adm. date
6+ months bet. pre and overlap episodes Overlap 2nd episode adm. date
3 1 Pre, 2 Post <6 months bet. pre and post episodes Overlap 1st episode adm. dat
6+ months bet. pre and 1st post episode Post 2nd episode adm. date
3 1 Overlap, 2 Post NA Overlap 1st episode adm. date
* Pre = Hospice episode occurred prior to nursing facility admission. Overlap = Hospice episode occurred prior to and after nursing facility admission. Post = Hospice episode occurred after nursing facility admission.
TABLE 2. Hospice Temporal Categories by Year of Hospice Administration, 1991-1996*
Hospice Temporary Category Admission Year
1991** 1992 1993 1994 1995 1996
Pre-Nursing Facility Admission 68(74.7%) 121(35.0%) 189(14.4%) 210(9.0%) 310(9.1%) 220(4.9%)
Overlapping with Nursing Facility Admission 23(25.3%) 144(41.6%) 237(18.1%) 325(14.0%) 412(12.1%) 405(9.1%)
Post Nursing Facility Admission 0(0%) 81(23.4%) 886(67.5%) 1,792(77.0%) 2,692(78.9%) 3,848(86.3%)
Total** 91(100%) 346(100%) 1,312(100%) 2,327(100%) 3,414(100%) 4,473(100%)
* 1 "Pre" beneficiary was admitted in 1990.** These beneficiaries had an MDS in 1992 but were admitted to hospice in 1991.
TABLE 3. Nursing Facility Resident Characteristics by Temporal Sequence of Hospice Admission, 1991-1997 Admissions*
  Timing of Hospice Admission
All Cases (100%)100% (N=11,964) Pre NHA (9.4%)100% (N=1,119) Overlap NHA (12.9%)100% (N=1,546) Post NHA (77.7%)100% (N=9,299)
Age**
Up to 65 370 (3.1%) 47 (4.2%) 72 (4.7%) 251 (2.7%)
65-74 1,942 (16.2%) 257 (23.0%) 401 (25.9%) 1,284 (13.8%)
75-84 4,499 (37.6%) 491 (43.9%) 678 (43.9% ) 3,330 (35.8%)
85 or older 5,153 (43.1%) 324 (29.0%) 395(25.6%) 4,434 (47.7%)
Gender**
Female 8,004 (66.9%) 629 (56.2%) 901 (58.3%) 6,474 (69.6%)
Male 3,960 (33.1%) 490 (43.8%) 645 (41.7%) 2,825 (30.4%)
Race/Ethnicity**
Native American 5 (.04%) 0 0 5 (.1%)
Asian 12 (.1%) 3 (.3%) 2 (.1%) 7 (.1%)
Black 487 (4.1%) 46 (4.1%) 53 (3.4%) 388 (4.2%)
Hispanic 19 (.2%) 1 (.1%) 5 (.3%) 13 (.1%)
White 11,300 (94.5%) 1,059 (94.6%) 1,460 (94.4%) 8,781 (94.4%)
Other 52 (.4%) 4 (.4%) 11 (.7%) 37 (.4%)
Unknown 89 (0.7%) 6 (.5%) 15 (1.0%) 68 (.7%)
Marital Status** (N=11632) (N=1053) (N=1459) (N=9210)
Married 3,063 (26.3%) 384 (36.5%) 505 (34.6%) 2,174 (23.8%)
Widowed/ Divorced 7,551 (63.1%) 597 (56.7%) 840 (57.6%) 6,114 (67.6%)
Separated 111 (1.0%) 8 (.8%) 16 (1.1%) 87 (1.0%)
Never Married 907 (7.8%) 64 (6.1%) 98 (6.1%) 745 (8.2%)
* One admission from 1990.** From HCFA denominator file.*** From MDS.
TABLE 4. Most Frequent Principal Hospice Diagnoses* and Total Percent with Cancer by Temporal Category of Hospice Admission**
Diagnosis Timing of Hospice Admission
TotalN=11,894 (100%) PreN=1,115 (100%) OverlapN=1,541 (100%) PostN=9,238 (100%)
Lung Cancer 1,107 (9.3%) 163 (14.6%) 296 (19.2%) 648 (7.0%)
Heart Failure 1,000 (8.4%) 80 (7.1%) 79 (5.1%) 841 (9.1%)
CVA 671 (5.6%) 25 (2.2%) 27 (1.8%) 619 (6.7%)
Breast Cancer 579 (4.9%) 67 (6.0%) 90 (5.8%) 422 (4.6%)
COPD 567 (4.8%) 47 (4.2%) 51 (3.3%) 469 (5.1%)
Alzheimer's Disease 441 (3.7%) 21 (1.9%) 16 (1.0%) 404 (4.4%)
Dementia 433 (3.6%) 9 (.1%) 4 (.3%) 420 (4.5%)
Rectal Cancer 353 (3.0%) 61 (5.5%) 60 (3.9%) 232 (2.5%)
Colon Cancer 350 (2.9%) 39 (3.5%) 67 (4.3%) 244 (2.6%)
Renal Failure 320 (2.7%) 19 (1.7%) 20 (1.3%) 281 (3.0%)
Failure to thrive 318 (2.7%) 10 (.1%) 5 (.3%) 303 (3.3%)
Parkinson's Disease 194 (1.6%) 8 (.7%) 10 (.6%) 176 (1.9%)
Unspecified Psychosis (298.9) 182 (1.5%) 2 (.2%) 6 (.04%) 174 (1.9%)
Metastasis (coded as principal diagnosis) 165 (1.4%) 18 (1.6%) 37 (2.4%) 111 (1.2%)
Chronic Ischemic Heart Disease 157 (1.3%) 8 (.7%) 18 (1.2%) 131 (1.4%)
Total with Cancer** 5,325 (44.5%) 694 (62.0%) 1,115 (72.1%) 3,516 (37.8%)
* Per principal diagnosis from first hospice claim.** Total number and percent with primary or metastatic cancer coded as principal diagnosis.
TABLE 5. Diagnosis Categories by Temporal Sequence of Hospice Admission, 1992-1996*
Diagnosis Group Timing of Hospice Admission
All CasesN=11,964 (100%) Pre NHAN=1,119 (100%) Overlap NHAN=1,546 (100%) Post NHAN=9,299 (100%)
Cancer with Dementia 3,916 (32.7%) 202 (18.1%) 210 (13.6%) 2,904 (31.2%)
Cancer without Dementia 4,496 (37.6%) 614 (54.8%) 982 (63.5%) 2,602 (28.0%)
Dementia, No Cancer 1,150 (9.6%) 90 (8.0%) 72 (4.2%) 1,546 (16.6%)
Other 2,402 (20.0%) 213 (19.0%) 282 (18.2%) 2,247 (24.2%)
* Using MDSs during nursing facility stay and inpatient claims 6 months prior to and during nursing facility episode.
TABLE 6. Hospice Length of Stay by Temporal Sequence of Hospice Admission, 1992-1996
Days in Hospice Timing of Hospice Admission
TotalN=11,964 (100%) Pre NHAN=1,119 (100%) OverlapN=1,546 (100%) PostN=9,299 (100%)
<8 2,397 (20.0%) 97 (8.7%) 41 (2.7%) 2,259 (24.3%)
8-14 1,408 (11.8%) 81 (7.2%) 87 (5.6%) 1,240 (13.3%)
15-30 1,723 (14.4%) 198 (17.7%) 178 (11.5%) 1,347 (14.5%)
31-90 2,857 (23.9%) 338 (30.2%) 461 (29.8%) 2,058 (22.12%)
91-180 1,756 (14.7%) 196 (17.5%) 330 (21.4%) 1,230 (13.2%)
181-210 417 (3.5%) 52 (4.7%) 70 (4.5%) 295 (3.2%)
>210 1,406 (11.8%) 157 (14.0%) 379 (24.5%) 870 (9.4%)
Mean (SD) 92.3 (139.9) 108.9 (133.21) 166.9 (206.8) 77.9 (121.6)
Median 37 61 92 28
Mode 2 7 26 2
TABLE 7. Nursing Facility Residents by Diagnosis GroupMedicare Hospice Beneficiaries Admitted After Nursing Facility Admission, 1992-1996
  All CasesN=9,229 (100%) Diagnosis Group
NeitherN=2,247 (100%) Dementia, No CancerN=1,546 (100%) Any Cancer, No DementiaN=2,602 (100%) Any Cancer with DementiaN=2,904 (100%)
Age*
Up to 65 251(2.7%) 105(4.7%) 29(1.9%) 82(3.2%) 35(1.2%)
65-74 1,284(13.8%) 325(14.5%) 106(6.9%) 587(22.6%) 266(9.2%)
75-84 3,330(35.8%) 779(34.7%) 445(28.8%) 1,076(41.4%) 1,030(35.5%)
85 or older 4,434(47.7%) 1,038(46.2%) 966(62.5%) 857(32.9%) 1,573(54.2%)
Gender*
Female 6,474(69.6%) 1,642(73.1%) 1,236(80.0%) 1,657(63.7%) 1,939(66.8%)
Male 2,825(30.4%) 605(26.9%) 310(20.1%) 945(36.3%) 965(33.2%)
Race/Ethnicity*
Native American 5(0.1%) 1(0.0%) 0(0.0%) 2(0.1%) 2(0.1%)
Asian 7(0.1%) 1(0.0%) 5(0.3%) 1(0.0%) 0(0.0%)
Black 388(4.2%) 62(2.8%) 48(3.1%) 138(5.3%) 140(4.8%)
Hispanic 13(0.1%) 7(0.3%) 3(0.2%) 2(0.1%) 1(0.0%)
White 8,781(94.4%) 2,152(95.8%) 1,467(94.9%) 2,435(93.6%) 2,727(93.9%)
Other 37(0.4%) 11(0.5%) 3(0.2%) 11(0.4%) 12(0.4%)
Unknown 68(0.7%) 13(0.6%) 20(1.3%) 13(0.5%) 22(0.8%)
Marital Status** N=9,120(100%) N=2,177(100%) N=1,517(100%) N=2,564(100%) N2,862(100%)
Never Married 745(8.2%) 168(7.7%) 137(9.0%) 220(8.6%) 220(7.7%)
Married 2,174(23.8%) 465(21.4%) 265(17.5%) 680(26.5%) 764(26.7%)
Widowed 5,660(62.1%) 1,393(64.0%) 1,053(69.4%) 1,477(57.6%) 1,737(60.7%)
Separated 87(1.0%) 22(1.0%) 12(0.8%) 29(1.1%) 24(0.8%)
Divorced 454(5.0%) 129(5.9%) 50(3.3%) 158(6.2%) 117(4.1%)
* From HCFA denominator file.** From MDS.
TABLE 8. Resident Characteristics by Hospice in Nursing Facility Concentration CategoriesMedicare Hospice Beneficiaries Admitted After Nursing Facility Admission, 1992-1996
  All CasesN=9,229(100%) Concentration of Hospice in Nursing Homes
<2%N=2,052(100%) 2 - <5%N=2,827(100%) 5 - <9%N=2,186(100%) 9 - <13%N=1,190(100%) 13+%N=1,044(100%)
Age*
Up to 65 251(2.7%) 77(3.8%) 97(3.4%) 52(2.4%) 17(1.4%) 8(0.8%)
65-74 1,284(13.8%) 353(17.2%) 481(17.0%) 272(12.4%) 111(9.3%) 67(6.4%)
75-84 3,330(35.8%) 851(41.5%) 1,006(35.6%) 761(34.8%) 386(32.4%) 326(31.2%)
85 or older 4,434(47.7%) 771(37.6%) 1,243(44.0%) 1,101(50.4%) 676(56.8%) 643(61.6%)
Gender*
Female 6,474(69.6%) 1,332(64.9%) 1,891(66.9%) 1,555(71.1%) 891(74.9%) 805(77.1%)
Male 2,825(30.4%) 720(35.1%) 936(33.1%) 631(28.9%) 299(25.1%) 239(22.9%)
Race/Ethnicity*
Native American 5(0.1%) 1(0.1%) 2(0.1%) 2(0.1%) 0(0.0%) 0(0.0%)
Asian 7(0.1%) 1(0.1%) 3(0.1%) 1(0.1%) 2(0.2%) 0(0.0%)
Black 388(4.2%) 114(5.6%) 139(4.9%) 72(3.3%) 33(2.8%) 30(2.9%)
Hispanic 13(0.1%) 5(0.2%) 5(0.2%) 2(0.1%) 0(0.0%) 1(0.1%)
White 8,781(94.4%) 1,905(92.8%) 2,642(93.5%) 2,092(95.7%) 1,145(96.2%) 997(95.6%)
Other 37(0.4%) 13(0.6%) 16(0.6%) 5(0.2%) 1(0.1%) 2(0.2%)
Unknown 68(0.7%) 13(0.6%) 20(0.7%) 12(0.6%) 9(0.8%) 14(1.3%)
Marital Status** N=9,120(100%) N=1,994(100%) N=2,776(100%) N=2,145(100%) N=1,175(100%) N=1,030(100%)
Never Married 745(8.2%) 169(8.5%) 229(8.3%) 160(7.5%) 96(8.2%) 91(8.8%)
Married 2,174(23.8%) 574(28.8%) 716(25.8%) 461(21.5%) 235(20.0%) 188(18.3%)
Widowed 5,660(62.1%) 1,127(56.5%) 1,641(59.1%) 1,388(64.7%) 799(68.0%) 705(68.5%)
Separated 87(1.0%) 24(1.2%) 33(1.2%) 21(0.1%) 4(0.3%) 5(0.5%)
Divorced 454(5.0%) 100(5.0%) 157(5.7%) 115(5.4%) 41(3.5%) 41(4.0%)
* From HCFA denominator file.** From MDS.
TABLE 9. Nursing Facility Resident Characteristics by Hospice Discharge/DeathMedicare Hospice Beneficiaries Admitted After Nursing Facility Admission, 1992-1996
  Type of Hospice Discharge
All CasesN=9,299 (100%) Discharged from HospiceN=1,380 (100%) Died7,919 (100%)
Age*
Up to 65 251 (2.7%) 63 (4.6%) 188 (2.4%)
65-74 1,284 (13.8%) 214 (15.5%) 1,070 (13.5%)
75-84 3,330 (35.8%) 471 (34.1%) 2,859 (36.1%)
85 or older 4,434 (47.7%) 632 (45.8%) 3,802 (48.0%)
Gender*
Female 6,474 (69.6%) 1,106 (73.6%) 5,458 (68.9%)
Male 2,825 (30.4%) 364 (26.4%) 2,461 (31.1%)
Race/Ethnicity*
Native American 5 (0.1%) 1 (0.1%) 4 (0.1%)
Asian 7 (0.1%) 1 (0.1%) 6 (0.1%)
Black 388 (4.2%) 83 (6.0%) 305 (3.9%)
Hispanic 13 (0.1%) 2 (0.1%) 11 (0.1%)
White 8,781 (94.4%) 1,271 (92.1%) 7,510 (94.8%)
Unknown 68 (0.7%) 14 (1.0%) 54 (0.7%)
Other 37 (0.4%) 8 (0.6%) 29 (0.4%)
Marital Status** N=9,120 (100%) N=1,365 (100%) N=7,755 (100%)
Never Married 745 (8.2%) 110 (8.1%) 635 (8.2%)
Married 2,174 (23,8%) 322 (23.6%) 1,852 (23.9%)
Widowed 5,660 (62.1%) 833 (61.0%) 4,827 (62.2%)
Separated 87 (1.0%) 15 (1.1%) 72 (0.9%)
Divorced 454 (5.0%) 85 (6.2%) 369 (4.8%)
* From HCFA denominator file.** From MDS.
TABLE 10. Resident Characteristics by State of Nursing Facility ResidenceMedicare Hospice Beneficiaries Admitted After Nursing Facility Admission, 1992-1996
  TotalN=9,299(100%) State of Nursing Facility Residence
New YorkN=5,914(100%) MaineN=244(100%) KansasN=2,376(100%) MississippiN=347(100%) South DakotaN=418(100%)
Age*
Up to 65 251(2.7%) 156(2.6%) 8(3.3%) 58(2.4%) 17(5.0%) 12(2.9%)
65-74 1,284(13.8%) 707(12.0%) 54(22.1%) 358(15.1%) 86(24.8%) 79(18.9%)
75-84 3,330(35.8%) 2,02734.3%) 98(40.2%) 905(38.1%) 137(39.5%) 163(39.0%)
85 or older 4,434(47.7%) 3,024(51.1%) 84(34.4%) 1,055(44.4%) 107(30.8%) 164(39.2%)
Gender*
Female 6,474(69.6%) 4,305(72.8%) 147(60.3%) 1,550(65.2%) 211(60.8%) 261(62.4%)
Male 2,825(30.4%) 1,609(27.2%) 97(39.8%) 826(34.8%) 136(39.2%) 157(37.6%)
Race/Ethnicity*
Native American 5(0.1%) 3(0.1%) 0(0.0%) 0(0.0%) 0(0.0%) 2(0.5%)
Asian 7(0.1%) 7(0.1%) 0(0.0%) 0(0.0%) 0(0.0%) 0(0.0%)
Black 388(4.2%) 214(3.6%) 1(0.4%) 78(3.3%) 95(27.4%) 0(0.0%)
Hispanic 13(0.1%) 10(0.2%) 0(0.0%) 3(0.1%) 0(0.0%) 0(0.0%)
White 8,781(94.4%) 5,613(95.0%) 240(98.4%) 2,267(95.4%) 249(71.8%) 412(98.6%)
Other 37(0.4%) 22(0.4%) 1(0.4%) 11(0.5%) 0(0.0%) 3(0.7%)
Unknown 68(0.7%) 45(0.8%) 2(0.8%) 17(0.7%) 3(0.9%) 1(0.2%)
Marital Status** N=9,120(100%) N=5,772(100%) N=243(100%) N=2,342(100%) N=345(100%) N=418(100%)
Never Married 745(8.2%) 570(9.9%) 14(5.8%) 121(5.2%) 13(3.8%) 27(6.5%)
Married 2,174(23.8%) 1,146(19.9%) 84(34.6%) 691(29.5%) 117(33.9%) 136(32.5%)
Widowed 5,660(62.1%) 3,744(64.9%) 124(51.0%) 1,385(59.1%) 180(52.2%) 227(54.3%)
Separated 87(1.0%) 69(1.2%) 0(0.0%) 8(0.3%) 8(2.3%) 2(0.5%)
Divorced 454(5.0%) 243(4.2%) 21(8.6%) 137(5.9%) 27(7.8%) 26(6.2%)
* From HCFA denominator file.** From MDS.
TABLE 11. Diagnosis Categories by Hospice in Nursing Facility Concentration CategoriesMedicare Hospice Beneficiaries Admitted After Nursing Facility Admission, 1992-1996
Diagnosis Group All CasesN=9,229(100%) Concentration of Hospice in Nursing Homes
<2%N=2,052(100%) 2 - <5%N=2,827(100%) 5 - <9%N=2,186(100%) 9 - <13%N=1,190(100%) 13+%N=1,044(100%)
Cancer with Dementia 2,904(31.2%) 570(27.8%) 832(29.4%) 707(32.3%) 440(37.0%) 355(34.0%)
Cancer without Dementia 2,602(28.0%) 801(309%) 1,037(36.7%) 467(21.4%) 172(14.5%) 125(12.0%)
Dementia, No Cancer 1,546(16.6%) 186(9.1%) 294(10.4%) 443(20.3%) 297(25.0%) 326(31.2%)
Other 2,247(24.2%) 495(24.1%) 664(23.5%) 569(26.0%) 281(23.6%) 238(22.8%)
TABLE 12. Hospice Discharge/Death by Diagnosis GroupMedicare Hospice Beneficiaries Admitted After Nursing Facility Admission, 1992-1996
Diagnosis Group (100%) Type of Hospice Discharge
All CasesN=9,299 (100%) Discharged from HospiceN=1,380 (14.8%) Died7,919 (85.1%)
Cancer with Dementia 2,904 (100%) 398 (13.7%) 2,506 (86.3%)
Cancer without Dementia 2,602 (100%) 283 (10.9%) 2,319 (89.1%)
Dementia, no Cancer 1,546 (100%) 374 (24.2%) 1,172 (75.8%)
Other 2,247 (100%) 325 (14.5%) 1,922 (85.5%)
TABLE 13. State of Nursing Facility Residence by Diagnosis GroupMedicare Hospice Beneficiaries Admitted After Nursing Facility Admission, 1992-1996
Diagnosis Group All CasesN=9,299(100%) State of Nursing Facility Residence
New YorkN=5,914(100%) MaineN=244(100%) KansasN=2,376(100%) MississippiN=347(100%) South DakotaN=418(100%)
Cancer with Dementia 2,904(31.2%) 1,939(32.8%) 52(21.3%) 729(30.7%) 105(30.3%) 79(18.9%)
Cancer without Dementia 2,602(28.0%) 1,416(23.9%) 104(42.6%) 724(30.5%) 145(41.8%) 213(51.0%)
Dementia, No Cancer 1,546(16.6%) 1,142(19.3%) 17(7.0%) 332(14.0%) 25(7.2%) 30(7.2%)
Other 2,247(24.2%) 1,417(24.0%) 71(29.1%) 591(24.9%) 72(20.8%) 96(23.0%)
TABLE 14. Year of Hospice Admission by Diagnosis GroupsMedicare Hospice Beneficiaries Admitted After Nursing Facility Admission, 1992-1996
Diagnosis Group (100%) Year of Hospice Admission
All CasesN=9,299(100%) 1992N=81(100%) 1993N=886(100%) 1994N=1,792(100%) 1995N=2,692(100%) 1996N=3,848(100%)
Cancer with Dementia 2,904(31.2%) 17(21.0%) 230(26.0%) 546(30.5%) 862(32.0%) 1,249(32.5%)
Cancer without Dementia 2,602(28.0%) 34(42.0%) 246(27.8%) 500(27.9%) 748(27.8%) 1,074(27.9%)
Dementia, No Cancer 1,546(16.6%) 4(5.0%) 145(16.4%) 297(16.6%) 483(17.9%) 617(16.0%)
Other 2,247(24.2%) 26(32.1%) 265(30.0%) 449(25.1%) 599(23.3%) 908(23.6%)
TABLE 15. Clinical Status at Hospice Admission by Diagnosis CategoriesMedicare Hospice Beneficiaries Admitted After Nursing Facility Admission, 1992-1996
  All CasesN=9,194(100%) Diagnosis Group
OtherN=2,216(100%) Dementia, No CancerN=1,540(100%) Any Cancer, No DementiaN=2,559(100%) Any Cancer with DementiaN=2,879(100%)
ADL*
Minimal Oversight 179(2.0%) 53(2.4%) 8(0.5%) 86(3.4%) 32(1.1%)
Extensive Oversight 479(5.2%) 122(5.5%) 32(2.1%) 246(9.6%) 79(2.7%)
Limited Assistance 1,308(14.2%) 334(15.1%) 104(6.8%) 576(22.5%) 294(10.2%)
Extensive Assistance 2,100(22.8%) 553(25.0%) 222(14.4%) 779(30.4%) 546(19.0%)
Dependent 2,508(27.3%) 645(29.1%) 421(27.3%) 593(23.2%) 849(29.5%)
Highly Dependent 2,620(28.5%) 509(23.0%) 753(48.9%) 279(10.9%) 1,079(37.5%)
CPS**
Intact 1,716(18.7%) 583(26.4%) 36(2.4%) 973(37.8%) 124(4.3%)
Borderline Intact 1,288(14.0%) 454(20.5%) 54(3.5%) 582(22.6%) 198(6.9%)
Mild Impairment 1,163(12.7%) 336(15.2%) 123(8.0%) 405(15.7%) 229(10.4%)
Moderate Impairment 1,925(21.0%) 380(17.2%) 390(25.5%) 372(14.5%) 783(27.3%)
Moderately Severe Impairment 838(9.1%) 157(7.1%) 202(13.2%) 111(4.3%) 368(12.8%)
Severe Impairment 638(7.0%) 77(3.5%) 172(11.3%) 46(1.8%) 343(12.0%)
Very Severe Impairment 1,610(17.5%) 223(10.1%) 552(36.1%) 84(3.3%) 751(26.2%)
Symptoms
Pain 2,277(25.0%) 519(23.9%) 203(13.3%) 1,031(40.6%) 524(18.3%)
Shortness of Breath 1,346(14.8%) 511(23.4%) 123(8.0%) 436(17.2%) 276(9.6%)
Nausea/Vomiting 495(5.4%) 117(5.4%) 59(3.9%) 189(7.4%) 130(4.5%)
Persistent Mood Disturbance 1,282(14.0%) 359(16.3%) 209(13.7%) 323(12.6%) 391(13.6%)
* Activities of Daily Living.** Cognitive Performance Scale.
TABLE 16. Clinical Status by Hospice in Nursing Facility Concentration CategoriesMedicare Hospice Beneficiaries Admitted After Nursing Facility Admission, 1992-1996
  All CasesN=9,194(100%) Concentration of Hospice in Nursing Homes
<2%N=2,032(100%) 2 - <5%N=2,777(100%) 5 - <9%N=2,160(100%) 9 - <13%N=1,187(100%) 13+%N=1,038(100%)
ADL*
Minimal Oversight 179(2.0%) 58(2.9%) 69(2.5%) 29(1.3%) 11(0.9%) 12(1.2%)
Extensive Oversight 479(5.2%) 168(8.3%) 171(6.2%) 76(3.5%) 27(2.3%) 37(3.6%)
Limited Assistance 1,308(14.2%) 390(19.2%) 422(15.2%) 266(12.3%) 115(9.7%) 115(11.1%)
ExtensiveAssistance 2,100(22.8%) 543(26.7%) 675(24.3%) 434(20.1%) 254(21.4%) 194(18.7%)
Dependent 2,508(27.3%) 480(23.6%) 694(25.0%) 640(29.6%) 340(28.6%) 354(34.1%)
Highly Dependent 2,620(28.5%) 393(19.3%) 746(26.9%) 715(33.1%) 440(37.1%) 326(31.4%)
CPS**
Intact 1,716(18.7%) 590(29.0%) 575(20.7%) 328(15.2%) 120(10.2%) 103(10.0%)
Borderline Intact 1,288(14.0%) 371(18.2%) 441(15.8%) 262(12.2%) 129(11.0%) 85(8.3%)
Mild Impairment 1,163(12.7%) 276(13.6%) 391(14.0%) 250(11.6%) 112(9.5%) 134(13.0%)
Moderate Impairment 1,925(21.0%) 335(16.5%) 578(20.8%) 465(21.6%) 283(24.0%) 264(25.6%)
Moderately Severe Impairment 838(9.1%) 122(6.0%) 222(8.0%) 197(9.2%) 168(14.3%) 129(12.5%)
Severe Impairment 638(7.0%) 119(5.9%) 149(5.4%) 169(7.9%) 91(7.7%) 110(10.7%)
Very Severe Impairment 1,610(17.5%) 221(10.9%) 428(15.4%) 481(22.4%) 275(23.3%) 205(19.9%)
Symptoms
Pain 2,270(25.0%) 504(25.1%) 846(30.7%) 495(23.1%) 232(19.9%) 200(19.4%)
Shortness of Breath 1,346(14.8%) 270(13.5%) 468(16.9%) 334(15.6%) 157(13.4%) 117(11.3%)
Nausea/Vomiting 495(5.4%) 101(5.0%) 189(6.8%) 111(5.2%) 51(4.4%) 43(4.2%)
Persistent Mood Disturbance 1,282(14.0%) 246(12.1%) 406(14.7%) 335(15.6%) 156(13.2%) 139(13.4%)
* Activities of Daily Living.** Cognitive Performance Scale.
TABLE 17. Clinical Status by Hospice Discharge/Death*Medicare Hospice Beneficiaries Admitted After Nursing Facility Admission, 1992-1996
  Type of Hospice Discharge
All CasesN=9,299(100%) Discharged from HospiceN=1,380(100%) Died7,919(100%)
ADL**
Minimal Oversight 179 (2.0%) 24 (1.8%) 155 (2.0%)
Extensive Oversight 479 (5.2%) 65 (4.8%) 414 (5.3%)
Limited Assistance 1,308 (14.2%) 175 (12.9%) 1,133 (14.5%)
Extensive Assistance 2,100 (22.8%) 289 (21.2%) 1,811 (23.1%)
Dependent 2,508 (27.3%) 353 (25.9%) 2,155 (27.5%)
Highly Dependent 2,620 (28.5%) 455 (33.4%) 2,165 (27.6%)
CPS**
Intact 1,716 (18.7%) 223 (16.4%) 1,493 (19.1%)
Borderline Intact 1,288 (14.0%) 166 (12.2%) 1,122 (14.4%)
Mild Impairment 1,163 (12.7%) 160 (11.8%) 1,003 (12.8%)
Moderate Impairment 1,925 (21.0%) 289 (21.2%) 1,636 (20.9%)
Moderately Severe Impairment 838 (9.1%) 125 (9.2%) 713 (9.1%)
Severe Impairment 638 (7.0%) 104 (7.6%) 534 (6.8%)
Very Severe Impairment 1,610 (17.5%) 294 (21.6%) 1,316 (16.8%)
Symptoms
Pain 2,277 (25.0%) 296 (21.8%) 1,981 (25.6%)
Shortness of Breath 1,346 (14.8%) 170 (12.5%) 1,176 (15.2%)
Nausea/Vomiting 495 (5.4%) 48 (3.5%) 447 (5.8%)
Persistent Mood Disturbance 1,282 (14.0%) 180 (13.2%) 1,102 (14.1%)
* Numbers may not add up to total because missing values.** Activities of Daily Living.*** Cognitive Performance Scale.
TABLE 18. Clinical Status by State of Nursing Facility ResidenceMedicare Hospice Beneficiaries Admitted After Nursing Facility Admission, 1992-1996
  TotalN=9,194(100%) State of Nursing Facility Residence
New YorkN=5,909(100%) MaineN=237(100%) KansasN=2,302(100%) MississippiN=331(100%) South DakotaN=415(100%)
ADL*
Minimal Oversight 179(2.0%) 81(1.4%) 8(3.4%) 67(2.9%) 9(2.7%) 14(3.4%)
Extensive Oversight 479(5.2%) 247(4.2%) 21(8.9%) 153(6.7%) 22(6.7%) 36(8.7%)
Limited Assistance 1,308(14.2%) 708(12.0%) 48(20.3%) 384(16.7%) 72(21.8%) 96(23.1%)
Extensive Assistance 2,100(22.8%) 1,252(21.2%) 70(29.5%) 541(23.5%) 108(32.6%) 129(31.1%)
Dependent 2,508(27.3%) 1,722(29.1%) 45(19.0%) 605(26.3%) 60(18.1%) 76(18.3%)
Highly Dependent 2,620(28.5%) 1,899(32.1%) 45(19.0%) 552(24.0%) 60(18.1%) 64(15.4%)
CPS**
Intact 1,716(18.7%) 1,034(17.6%) 66(27.1%) 374(16.2%) 129(37.6%) 113(27.3%)
Borderline Intact 1,288(14.0%) 781(13.3%) 42(17.2%) 333(14.4%) 45(13.1%) 87(21.0%)
Mild Impairment 1,163(12.7%) 718(12.2%) 35(14.3%) 324(14.0%) 26(7.6%) 60(14.5%)
Moderate Impairment 1,925(21.0%) 1,217(20.7%) 48(19.7%) 526(22.8%) 51(14.9%) 83(20.1%)
Moderately Severe Impairment 838(9.1%) 596(10.2%) 13(5.3%) 191(8.3%) 21(6.1%) 17(4.1%)
Severe Impairment 638(7.0%) 362(6.2%) 13(5.3%) 219(9.5%) 24(7.0%) 20(4.8%)
Very Severe Impairment 1,610(17.5%) 1,161(19.8%) 27(11.1%) 341(14.8%) 47(13.7%) 34(8.2%)
Symptoms
Pain 2,277(25.0%) 1,099(18.6%) 100(41.0%) 787(35.8%) 99(29.8%) 192(45.9%)
Shortness of Breath 1,346(14.8%) 665(11.2%) 72(29.5%) 449(20.4%) 64(18.6%) 96(23.0%)
Vomiting 495(5.4%) 301(5.1%) 20(8.2%) 118(5.4%) 16(4.7%) 40(9.6%)
Persistent Mood Disturbance 1,282(14.0%) 730(12.4%) 58(24.2%) 417(18.0%) 16(4.9%) 61(15.0%)
* Activities of Daily Living.** Cognitive Performance Scale.
TABLE 19. Length of Stay by Year of Hospice AdmissionMedicare Hospice Beneficiaries Admitted After Nursing Facility Admission, 1992-1996
Days in Hospice Year of Hospice Admission
TotalN=9,299(100%) 1992N=81(100%) 1993N=886(100%) 1994N=1,792(100%) 1995N=2,692(100%) 1996N=3,848(100%)
<8 2,259(24.3%) 14(17.3%) 194(21.9%) 404(22.5%) 667(24.8%) 980(25.5%)
8-14 1,240(13.3%) 12(14.8%) 119(13.4%) 223(12.4%) 362(13.5%) 524(13.6%)
15-30 1,347(14.5%) 11(13.6%) 138(15.6%) 262(14.6%) 385(14.3%) 551(14.3%)
31-90 2,058(22.1%) 17(21.0%) 197(22.2%) 377(21.0%) 561(20.8%) 906(23.5%)
91-180 1,230(13.2%) 9(11.1%) 98(11.1%) 245(13.7%) 362(13.5%) 515(13.4%)
181-210 295(3.2%) 2(2.5%) 30(3.4%) 58(3.2%) 90(3.3%) 115(3.0%)
>210 870(9.4%) 16(19.8%) 110(12.4%) 223(12.4%) 265(9.8%) 256(6.7%)
Mean(SD) 77.9(121.6) 142.2(247.8) 97.7(170.6) 93(146.1) 76.8(112.4) 65.3(92.3)
Median 28 41 29 31 27 25
Mode 2 3 3 3 2 2
TABLE 20. Length of Stay in Hospice by Diagnosis GroupMedicare Hospice Beneficiaries Admitted After Nursing Home Admission, 1992-1996
Days in Hospice TotalN=9,299(100%) Diagnosis Group
Cancer, No DementiaN=2,602(100%) Cancer, With DementiaN=2,904(100%) Dementia, No CancerN=1,546(100%) OtherN=2,247(100%)
<8 2,259(24.3%) 483(18.6%) 750(25.8%) 386(25.0%) 640(28.5%)
8-14 1,240(13.3%) 354(13.6%) 416(14.3%) 172(11.1%) 298(13.3%)
15-30 1,347(14.5%) 443(17.0%) 408(14.1%) 166(10.7%) 330(14.7%)
31-90 2,058(22.1%) 696(26.8%) 587(20.2%) 320(20.7%) 455(20.3%)
91-180 1,230(13.2%) 331(12.8%) 372(12.8%) 249(16.1%) 278(12.4%)
181-210 295(3.2%) 66(2.5%) 86(3.0%) 74(4.8%) 69(3.1%)
>210 870(9.4%) 229(8.8%) 285(9.8%) 179(11.6%) 177(7.9%)
Mean(SD) 77.9(121.6) 76.0(118.6) 78.2(125.5) 93.0(130.8) 68.3(112.1)
Median 28 32 25 39 22
Mode 2 2 2 3 2
TABLE 21. Lenth of Stay in Hospice by Nursing Facility Hospice ConcentrationMedicare Hospice Beneficiaries Admitted After Nursing Facility Admission, 1992-1996
Days From First Claim to Last Claim TotalN=9,229(100%) Hospice Concentration
<2%N=2,052(100%) 2 - <5%N=2,827(100%) 5 - <9%N=2,186(100%) 9 - <13%N=1,190(100%) 13+%N=1,044(100%)
<8 2,259(24.3%) 497(24.2%) 667(23.6%) 532(24.2%) 300(25.2%) 263(25.2%)
8-14 1,240(13.3%) 296(14.4%) 365(12.9%) 298(13.6%) 155(13.0%) 126(12.1%)
15-30 1,347(14.5%) 338(16.5%) 452(16.0%) 284(13.0%) 149(12.5%) 124(11.9%)
31-90 2,058(22.1%) 490(23.9%) 659(23.3%) 471(21.6%) 240(20.2%) 198(19.0%)
91-180 1,230(13.2%) 231(11.3%) 362(12.8%) 302(13.8%) 177(14.9%) 158(15.3%)
181-210 295(3.3%) 59(2.9%) 72(2.6%) 76(3.5%) 46(3.9%) 42(4.0%)
>210 870(9.4%) 141(6.9%) 250(8.8%) 223(10.2%) 123(10.3%) 133(12.7%)
Mean(SD) 77.9(121.6) 65.6(106.8) 74.8(118.0) 77.0(116.0) 82.7(127.5) 97.6(145.5)
Median 28 24 28 27 29 35
Mode 2 2 2 3 3 2
TABLE 22. Length of Stay in Hospice For Nursing Facility Residents by Hospice Discharge/DeathMedicare Hospice Beneficiaries Admitted After Nursing Facility Admission, 1992-1996
Days in Hospice Type of Hospice Discharge
All CasesN=9,299(100%) Discharged from HospiceN=1,380(100%) Died7,919(100%)
<8 2,259 (24.3%) 81 (5.9%) 2,178 (27.5%)
8-14 1,240 (13.3%) 49 (3.6%) 1,191 (15.0%)
15-30 1,347 (14.5%) 91 (6.6%) 1,256 (15.9%)
31-90 2,058 (22.1%) 361 (26.2%) 1,697 (21.4%)
91-180 1,230 (13.2%) 371 (26.9%) 859 (10.9%)
181-210 295 (3.2%) 107 (7.8%) 188 (2.4%)
>210 870 (9.4%) 320 (23.2%) 550 (7.0%)
Mean (SD) 77.9 (121.6) 175.7 (172.9) 60.8 (101.0)
Median 28 129 20
Mode 2 90 2
TABLE 23. Length of Stay in Hospice by State of Nursing Facility ResidenceMedicare Hospice Beneficiaries Admitted After Nursing Facility Admission, 1992-1996
Days in Hospice TotalN=9,299(100%) State of Nursing Facility Residence
New YorkN=5,914(100%) MississippiN=347(100%) KansasN=2,376(100%) MaineN=244(100%) South DakotaN=418(100%)
<8 2,259(24.3%) 1,478(25.0%) 68(19.6%) 576(24.2%) 33(13.5%) 104(24.9%)
8-14 1,240(13.3%) 782(13.2%) 45(13.0%) 316(13.3%) 38(15.6%) 59(14.1%)
15-30 1,347(14.5%) 830(14.0%) 51(14.7%) 350(22.8%) 49(20.1%) 67(16.0%)
31-90 2,058(22.1%) 1,256(21.2%) 101(29.1%) 542(14.7%) 57(23.4%) 102(24.4%)
91-180 1,230(13.2%) 794(13.4%) 37(10.7%) 319(13.4%) 31(12.7%) 49(11.7%)
181-210 295(3.2%) 208(3.5%) 8(2.3%) 67(2.8%) 5(2.1%) 7(1.7%)
>210 870(9.4%) 566(9.6%) 37(10.7%) 206(8.7%) 31(12.7%) 30(7.2%)
Mean(SD) 77.9(121.6) 77.9(118.6) 84.3(121.8) 76.6(122.2) 81.6(125.7) 77.7(152.9)
Median 28 28 33 27 34 25
Mode 2 2 4 3 2 2
TABLE 24. Length of Stay in Hospice by Hospice Ownership StatusMedicare Hospice Beneficiaries Admitted After Nursing Facility Admission, 1992-1996
Days in Hospice TotalN=9,268 (100%) Type of Ownership
GovernmentN=214 (100%) Non-ProfitN=8,379 (100%) For-ProfitN=675 (100%)
<8   56 (26.2%) 2,048 (24.4%) 148 (21.9%)
8-14   39 (18.2%) 1,119 (13.4%) 79 (11.7%)
15-30   34 (15.9%) 1,209 (14.4%) 98 (14.5%)
31-90   57 (26.6%) 1,839 (22.0%) 154 (22.8%)
91-180   19 (8.9%) 1,099 (13.1%) 106 (15.7%)
181-210   2 (.9%) 273 (3.3%) 19 (2.8%)
>210   7 (3.3%) 792 (9.5%) 71 (10.5%)
Mean (SD) 77.9 (121.6) 45.1 (72.3) 78.3 (122.5) 84.7 (122.7)
Median 28 18.5 27 33
Mode 2 4 2 3
TABLE 25. Proportion of Patients Using Hospice Levels of Care by Diagnosis GroupHospice Patients Admitted After Nursing Home Admission, 1992-1996
Diagnosis Group Level of Care
Routine Home Care Continuous Care General Inpatient Respite Inpatient Physician Visits
Cancer with dementia 92.9% 2.0% 11.5% .6% 3.4%
Cancer without dementia 95.5% 4.2% 13.1% 1.6% 6.5%
Dementia, no cancer 96.3% 1.2% 5.7% .3% 3.2%
Other 92.4% 3.2% 11.3% .8% 3.4%
All Patients 94.1% 2.8% 11.0% .8% 4.2%
TABLE 26. Average and Median Hospice Expenditures (in 1996 dollars) by Days in HospiceMedicare Hospice Beneficiaries Admitted After Nursing Facility Admission, 1992-1996
Days in Hospice Total Routine Continuous Care Inpatient General Inpatient Respite Physician Visits
Mean Median Mean Median Mean Median Mean Median Mean Median Mean Median
<8(N=2,259) 550 391 316 211 14 0 221 0 0 0 3 0
8-14(N=1,240) 1,460 1,052 896 950 41 0 523 0 1 0 5 0
15-30(N=1,347) 2,643 2,152 1,942 1,954 76 0 637 0 2 0 9 0
31-90(N=2,058) 6,055 5,535 5,397 5,135 70 0 601 0 7 0 9 0
91-180(N=1,230) 13,370 12,915 12,822 12,567 50 0 531 0 7 0 14 0
181-210(N=295) 19,559 18,894 18,720 18,585 178 0 679 0 5 0 12 0
>210(N=870) 36,421 32,250 35,987 31,898 171 0 392 0 19 0 22 0
All Patients(N=9,229) 7,848 3,093 7,329 2,534 64 0 477 0 5 0 9 0
TABLE 27. Average and Median Hospice Expenditures (in whole dollars) by Diagnosis GroupMedicare Hospice Beneficiaries Admitted After Nursing Facility Admission, 1992-1996
Diagnosis Group Total Routine Continuous Care Inpatient General Inpatient Respite Physician Visits
Mean Median Mean Median Mean Median Mean Median Mean Median Mean Median
Cancer, with Dementia(N=2904) 7,945 2,916 7,382 2,271 48 0 535 0 4 0 8 0
Cancer, no Dementia(N=2602) 7,708 3,437 7,019 2,813 109 0 588 0 8 0 15 0
Dementia, no Cancer(N=1546) 9,175 4,069 8,967 3,659 15 0 232 0 1 0 4 0
Other(N=2247) 6,970 2,587 6,491 2,015 63 0 443 0 4 0 6 0
All Patients(N=9299) 7,848 3,093 7,329 2,534 64 0 477 0 5 0 9 0
TABLE 28. Average and Median Hospice Expenditures (in 1996 dollars) by Hospice Death versus DischargeMedicare Hospice Beneficiaries Admitted After Nursing Facility Admission, 1992-1996
Type of Discharge Total Routine Continuous Care Inpatient General Inpatient Respite Physician Visits
Mean Median Mean Median Mean Median Mean Median Mean Median Mean Median
Died on Hospice(N=7,919) 6,214 2,383 5,713 1,865 60 0 457 0 4 0 9 0
Discharged from Hospice Prior to Death(N=1,380) 17,222 12,924 16,600 11,841 87 0 591 0 11 0 10 0
All Patients(N=9,229) 7,848 3,093 7,329 2,534 64 0 477 0 5 0 9 0
TABLE 29. Distance Between SNF Discharge and Hospice Admission by Nursing Home Ownership
Days Bet. SNF & Hospice TotalN=6,827 (100%) Ownership Type
For ProfitN=2,881 (100%) Non-ProfitN=3,006 (100%) GovernmentN=684 (100%)
0-1 1,772 (26.0%) 565 (19.6%) 892 (29.7%) 232 (33.9%)
2-7 442 (6.5%) 213 (7.4%) 153 (5.1%) 51 (7.5%)
8-31 911 (13.3%) 421 (14.6%) 349 (11.6%) 115 (16.8%)
>31 Days 3,702 (54.2%) 1,682 (58.4%) 1,612 (53.6%) 286 (41.8%)
TABLE 30. Distance Between SNF Discharge and Hospice Admission by Diagnosis GroupMedicare Hospice Beneficiaries Admitted After Nursing Facility Admission, 1992-1996
Days Bet. SNF and Hospice TotalN=6,827(100%) Diagnosis Group
Cancer, No DementiaN=2,049(100%) Cancer with DementiaN=2,198(100%) Dementia, No CancerN=923(100%) OtherN=1,657(100%)
0-1 1,772(26.0%) 566(27.6%) 519(23.6%) 284(30.8%) 403(24.3%)
2-7 442(6.5%) 159(7.8%) 136(6.2%) 45(4.9%) 102(6.2%)
8-31 911(13.3%) 276(13.5%) 328(14.9%) 76(8.2%) 231(13.9%)
>31 Days 3,702(54.2%) 1,048(51.2%) 1,215(55.3%) 518(56.1%) 921(55.6%)
TABLE 31. Distance Between SNF Discharge and Hospice Admission by Hospice ConcentrationMedicare Hospice Beneficiaries Admitted After Nursing Facility Admission, 1992-1996
Days Bet. SNF and Hospice TotalN=6,827(100%) Hospice Concentration
<2%N=1,596(100%) 2 - <5%N=2,152(100%) 5 - <9%N=1,519(100%) 9 - <13%N=822(100%) 13+%N=738(100%)
0-1 1,772(26.0%) 286(17.9%) 578(26.9%) 373(24.6%) 221(26.9%) 314(42.6%)
2-7 442(6.5%) 101(6.3%) 158(7.3%) 96(6.3%) 54(6.6%) 33(4.5%)
8-31 911(13.3%) 254(15.9%) 307(14.3%) 203(13.4%) 84(10.2%) 63(8.5%)
>31 Days 3,702(54.2%) 955(59.8%) 1,109(51.5%) 847(55.8%) 463(56.3%) 328(44.4%)
TABLE 32. Average and Median Medicare Expenditures (in 1996 Dollars) in 180 Days Prior to Medicare Hospice Election by Hospice Provider TypeMedicare Hospice Beneficiaries Admitted After Nursing Facility Admission, 1992-1996
Provider Type Total Medicare Expenditure Category
Inpatient Skill Nursing Home Health
% with Expenditures Mean Median % with Expenditures Mean Median % with Expenditures Mean Median % with Expenditures Mean Median
Hospital BasedN=2,540(27.3%) 72.2% 11,296 6,666 68.1% 8,327 4,590 41.0% 2,400 0 21.6% 570 0
Skilled Nursing Facility BasedN=36(.4%) 77.8% 20,825 19,583 72.2% 15,968 13,564 52.8% 3,043 1,993 41.7% 1,814 0
Home Health Agency BasedN=3,482 (37.6%) 61.9% 8,755 4,836 57.4% 6,422 3,541 36.9% 1,708 0 20.4% 624 0
FreestandingN=3,210 (34.7%) 69.1% 10,715 6,906 64.5% 7,341 4,446 42.5% 2,651 0 22.0% 723 0
All Hospice PatientsN=9,268 (100%)* 67.3% 10,206 6,114 62,8% 7,311 4,205 40.0% 2,236 0 21.4% 659 0
* There was no hospice provider match for 31 hospice patients.

Appendix A.

Data and Variables Used for Hospice in Nursing Facility Analyses

March 2000


This appendix was used for both Report 4 (Use of Medicare's Hospice Benefit by Nursing Facility Residents) and Report 5 ( Outcomes and Utilization for Hospice and Non-Hospice Nursing Facility Decedents).


 

DATA SOURCES

Resident Assessment Instrument (the Minimum Data Set or MDS)

The Omnibus Budget Reconciliation Act of 1987 (OBRA '87) contained the most far-reaching revisions to the standards, inspection process and enforcement system in nursing facilities since the passage of Medicare and Medicaid in 1965 (Hawes, 1998). A major feature of this legislation was the introduction of a uniform, comprehensive resident assessment instrument (the MDS) to guide the clinical care planning process in order to systematically document residents' needs. The MDS is not only used to systematically assess the resident and to generate a comprehensive care plan to document clinical progress as that plan is implemented, but it is used by regulators to focus on resident outcomes and by facilities to improve their performance. In the time period studied here documentation of the resident assessments were required: at admission (by 15th day), quarterly (by 90th day), and annually (by 365th day). Reassessments were required when a resident was readmitted after hospital admissions and when significant change occurred. Resident assessments were to have been completed on all nursing facility residents cared for in facilities receiving any Medicare or Medicaid payment. These nursing facilities represent 96 percent of the facilities in the United States.

Topics covered in the MDS include cognitive function, communication/hearing problems, physical functioning, continence, psychosocial well-being, mood state, activity and recreation, disease diagnoses, health conditions/symptoms, nutritional status, oral/dental status, skin condition, special treatments, and medication use. A number of studies (Morris et al., 1994; Frederiksen et al., 1996; Hartmaier et al., 1994 & 1995; Phillips et al., 1993; Mor et al., 1994) demonstrate that researchers and clinicians using the MDS can achieve high levels of inter-rater reliability. Using an earlier version of the data set used in this study, Gambassi and colleagues found reasonably high levels of validity and good internal consistency comparing diagnoses on the MDS with HCFA claims and medical conditions with patterns of use of specific drugs (Gambassi et al, 1988). While the accuracy of the MDS data have been questioned (Berlowitz et al., 1997; Kramer et al., 199 ), and considerable anecdotal evidence reveals that some facilities have not taken the time to train their staff properly in its use, it is unlikely that misclassification errors in recording of information will be differential with respect to the outcomes of interest.

For the comparative study hospice and non-hospice patients had to have had at least 2 MDS assessments performed. The need for the presence of 2 MDS assessments for our comparative analyses was originally recognized after preliminary analysis documented the presence of ascertainment bias on selected symptoms. Specifically, hospice residents were significantly more likely to have pain and dyspnea recorded than were non-hospice decedents. For example, controlling for other patient factors, residents with a dementia diagnoses were 3 times as likely to have pain recorded than were non-hospice residents. Therefore, to more correctly represent the hospice influence on the presence and management of symptoms (rather than merely the increased likelihood of hospice to assess symptoms) we felt that it was necessary to control for the status of symptoms at the time of the penultimate MDS.

Health Care Financing Administration (HCFA) Claims Data

The HCFA claims data were merged to the MDS file using the Health Insurance Claim number of Medicare beneficiaries. To ensure confidentiality, these identifiers were replaced with unique identifiers using the claim number as a seed. Two files (beneficiary information and claims data) comprise the HCFA data. The beneficiary file (Denominator file) contains gender, date of birth, and survival status (verified date of death). The claims data used include all Medicare Part A claims including hospital, skilled nursing facility, hospice, and home health agency claims. We achieved a match rate of MDS data to HCFA beneficiary data of approximately 85 percent using HCFA data from 1991through 1997.

Drug Data

As part of the resident assessment, nursing facility staffs code up to eighteen drugs taken within the seven days preceding the assessment. Nursing home staffs code each drug according to the National Drug Coding (NDC) system. Field tests of the MDS showed that 100 percent of the medication use items were reliable with the average reliability being 0.73 (Hawes et al., 1995b). While prescription drug products must use an NDC, most non-prescription drugs are also primarily referenced by the NDC. NDCs are unique 11-digit codes that identify discrete drug products. The first five digits refer to the manufacturer. The next four digits correspond to the drug product. The last two digits indicate the packaging. As pharmaceutical companies merge, new products are introduced, and drugs are no longer active, changes in NDCs occur continuously. Consequently, the NDCs are commercially-oriented and do not contain any mechanism to group drugs according to ingredients or categories of ingredients. Therefore, linking NDCs to specific descriptive information is critical to enable research. This NDC matching entails several steps. To match the NDC codes, we used a historical reference archive for drug products that listed all NDCs ever attributed and eventually discontinued between 1991 and 1996. For scientific drug research, we translated NDC codes into a hierarchical therapeutic classification scheme as recommended by WHO (Pahor et al., 1994). NDC codes were merged to useable therapeutic class and sub-class information using the Master Drug Data Base (MediSpan™) (1995). MediSpan™ contains complete records for prescriptions common in retail pharmacy as well as unit-dose and injectables used by hospitals and external facilities. MediSpan™ now includes over 100,000 generic drug products, products from regional manufacturers, and information on over 90,000 inactive drugs. The hierarchical identifier, the Generic Product Identifier (GPI) contained in MediSpan™, is a 14-character field consisting of seven subsets, each providing increasingly more specific information about the drug. (See below example.) While MediSpan™ incorporates the American Hospital Formulary Service (AHFS) (1994), a classification system based on the pharmacological uses of drugs, MediSpan™ also groups drugs with comparable compounds in the same therapeutic class and allows the same drug to be classified into multiple therapeutic classes.

Medi-span™ Classification System - Example of an Antidepressant
GPI Coding Example
58- Drug group Antidepressants
58-20- Drug class Tricyclic agents
58-20-00- Drug sub-class --
58-20-00-60 Drug name Nortiptyline
58-20-00-60-10 Drug name extension Hydochloride
58-20-00-60-10-01 Dosage form 10mg

A recent study analyzed the MDS drug codes with respect to: (1) completeness; (2) internal consistency; and (3) external validity (Gambassi et al., 1998). Investigators found the overall match rate between the NDC and the MediSpan™ greater than 90 percent with only 5.4 percent of the original NDC codes contained in the MDS data from the states being studied in this project to be incomplete or incorrect. Gender-specific medications had a high concordance with gender (>90 percent). For example, all residents taking tamoxifen were women; all residents taking goserelin were men; and 92 percent of estrogen users were women. Cross-linkages between drugs and MDS condition variables revealed adequate to high rates of concordance (range: 51 percent (gout) - 100 percent (rheumatoid arthritis)). High rates of concordance were reported when cross-checking levo-dopa with Parkinson's disease (88.9 percent); hypoglycemic agents with diabetes mellitus (93.2 percent); and sore care products with pressure sores (83.7 percent). These data show that the MDS drug data are consistent and reliable (Gambassi et al., 1998).

Hospice Provider of Service File

The Provider of Service (POS) File is compiled and managed by HCFA to determine the capacity of Medicare/Medicaid institutional providers to render acceptable care. This file contains information on program characteristics, collected by State surveyors under Federal guidelines. Since the periodic inspection of hospices is not mandated, surveys are conducted according to state priorities and resources, and, as a consequence, hospice and nursing facility data does not match across time. We used the hospice provider number on the HCFA hospice claim to link the hospice provider information to MDS and claims data. For our descriptive analyses, we used 1995 hospice provider information. For our analyses we were most interested in hospice provider type (freestanding or home care, hospital or nursing facility based), and ownership.

 

VARIABLES

Table of Variables

Table A1 lists all variables studied in both the descriptive and comparative hospice in nursing facility analyses. Variable measurement and data source are shown. For selected variables, the text below provides more information on their documentation and/or their reliability and validity.

Pain

Pain was not a major focus of the MDS version used in these states during the time period in question, and, as such, there are several limitations to the measure used in these analyses. A major limitation is that the level of pain intensity is not recorded; the newer MDS 2.0 requires documentation of pain intensity.

The MDS pain data used in this study is based on assessment by nursing personnel which is supposed to be performed according to instructions provided in the MDS manual (HCFA, 1991). Nursing home personnel are supposed to evaluate signs and symptoms of pain, but since pain is a subjective experience, they are instructed to record whatever the residents said it was. Residents were to be asked whether they had experienced any pain in the last seven days. Furthermore, residents were to be asked to describe the pain and how often it was manifest. To elicit complete and satisfactory answers, the assessors were instructed to ask neutral and non-directive questions. Questions such as: "What do you mean?" "Tell me what you have in mind." "Tell me more about that." "Please be more specific." "Give me an example". Moreover, the assessors were instructed to validate their understanding of what the resident was really saying. Statements like "I think I hear you saying that ….." or "Let's see if I understood you correctly. You said ….. Is it right?" were suggested in the MDS Instruction Manual.

For MDS assessment purposes, pain refers to any type of physical pain or discomfort in any part of the body experienced on a daily basis. If the assessor had difficulty discriminating the frequency, the instructions were to code as daily. Pain could have been localized or more generalized. It could have been acute or chronic, continuous or intermittent, occurring at rest or with movement. Pain recording could have depended exclusively on the observation of signs of pain. According to the MDS Manual, these include moaning, crying, and other vocalizations; wincing or frowning and other facial expressions; or body posture such as guarding/protecting an area of the body, or lying very still. In these cases, the assessors were instructed to ask the nurse assistants and therapists who might have been working with the resident, whether he/she had complaints or signs of pain during their shifts. In some residents, those who have dementia and cannot verbalize the pain experience, the assessor was instructed to look for particular behaviors such as calling out for help, pained facial expressions, refusing to eat, or striking out at a nurse assistant who was trying to move them or touch a body part.

Dyspnea

Shortness of breath (dyspnea) is recorded on the MDS if the problem is present in the "last 7 days" prior to completion of the MDS assessment. The degree or frequency of dyspnea is not indicated, only its presence. We compare the presence or absence of dyspnea in hospice versus non-hospice decedents stratifying for its presence on the penultimate MDS assessment.

Persistent Mood Disturbance

Persistent mood disturbance is defined as "persistent sad or anxious mood that has existed over the last 7 days and was not easily altered by attempt to "cheer up" the resident." (MDS+ Reference Manual, 1993) For MDS assessment purposes, a sad or anxious mood is a distressed mood that is characterized by explicit verbal or gestural expressions of feeling depressed or anxious (or a synonym such as feeling sad, miserable, blue, hopeless, empty, or tearful). Assessors are instructed to draw upon their own interactions with the residents as well as to statements of direct-care staff, social workers, and licensed personnel who may have evaluated the resident in this area. Suggested cues are: Does the resident cry or look dejected (unhappy) when no one is talking with him or her? When you talk with the resident, does he or she sound hopeless, fearful, sad, anxious? Does the resident appear withdrawn, apathetic, without emotion? (MDS+ Reference Manual, 1993)

Cognitive Performance

The MDS includes seven direct measures of cognition: short and long term memory, recall or orientation items (season, location or room, staff names/faces, orientation to nursing home), and decision-making ability. Good reliability (0.7) of these items has been reported (Hawes et al, 1995b). The cognitive performance scale (CPS) used in this study is a categorical measure of cognition using these MDS items and several items which indirectly evaluate cognitive function (i.e. comatose state, total dependent eating) (Morris et al., 1994). Based on two standard cognitive assessment tools, the Mini-Mental State Examination (Folstein et al., 1975) and the Test for Severe Impairment (Albert & Cohen, 1992), the CPS has excellent reliability with estimates published in the range of 0.66 - 0.88 (159). Using the MMSE as the gold standard, the CPS has high sensitivity (>90 percent) and specificity (>85 percent), yielding high diagnostic accuracy, regardless of patient education level (Hartmaier et al., 1995). Furthermore, the CPS has excellent reproducibility (Kappa >0.76) (155).

Activities of Daily Living

The reliability of the ADL scores range from 0.87-0.92 (Hawes et al., 1995b) and is highly correlated (0.89) with the Physical Signs and Symptoms Scale (Lawton & Brody, 1969). Furthermore, a recent study found these measures useful in pharmacoepidemiologic studies (Bernabei et al., 1998).

TABLE A1. Comparative Descriptive and Analytic Analyses--Variables, Measures, and Data Sources Outcomes
Type Variable Empirical Measure Source Data
Patient Acute Care Hospitalization and Average Hospital Days
-- in 30 days prior to death
-- in 90 days prior to death
-- in 6 months prior to death

Hospice decedents in hospice for total time period studied (and their matched controls) who have acute care hospitalization/ hospice decedents in hospice for total time period studied (and their matched controls).

Days in hospital in time period for hospice decedents in hospice for total time period studied (and their matched controls)/ hospice decedents in hospice for total time period studied (and their matched controls).

Medicare Part A Hospital Claims
Pain Management (regular treatment) Decedents in pain and receiving WHO level analgesia at least twice a day, or, for level III drugs, having a drug patch. Resident Assessment and Drug Information
Persistent Mood Disturbance (treatment of) Decedents with persistent mood disturbance and receiving antiananxiety or antidepressant medication in the 5 to 7 days prior to MDS assessment data. Resident Assessment
Independent Variables
Type Variable Empirical Measure Source Dataset
Patient Hospice enrollment Resident elected hospice prior to nursing facility admission. Medicare hospice claims and Resident assessment
Covariates and Variables for Descriptive Comparisons
Type Variable Empirical Measure Source Dataset
Facility-
  Nursing
  Facility

Facility-
  Hospice

Hospice Concentraction Unduplicated nursing facility residents on Medicare hospice in given year/ unduplicated nursing facility residents in a given year. Medicare hospice claims and Resident Assessment
For-profit ownership

Government ownership

Organizational type

A for-profit organization controls and operates the hospice.
A government entity controls and operates the hospice.
Hospice is freestanding or under administrative control of a hospital, home health agency or nursing home.
Hospice Provider of Service File
    "

    "

Level Variable Empirical Measure Source Dataset
Patient State of residence

Year of death
Race/ Ethnicity

State location of nursing facility in which resident resides
Year in which resident died
White/ Afr.Am./ Latino/ Native Am./ Asian
Resident Assessment
HCFA Denominator File
    "
Gender Female/ Male     "
Age Years     "
Marital Status Married/ Widowed or Divorced/ Separated/ Never married - most recent status Resident Assessment
Activities of daily living ADL score - see text     "
Cognitive performance CPS score - see text     "
Body Mass Index Weight (kg)/ height2 (m2) Resident Assessment
Diagnosis Diagnostic categories of (1) Cancer without dementia (2) Cancer with dementia (3) Alzheimer's disease/ dementia, (4) Other Resident Assissments, Inpatient Claims 6 months before NF adm. & during stay
Individual selected diagnoses for multivariate analysis. Resident Assessment
Pain

Dyspnea

Persistent Mood Disturbance

Vomiting

Complains or shows evidence of pain daily or almost daily (in 7 days prior to assessment).
Difficulty breathing occurring at rest, with activity, in response to illness or anxiety, or when lying flat.
Persistent sad or anxious mood that has existed over the last 7 days and was not easily attended by attempt to "cheer up" the resident.
Vomiting in 7 days prior to assessment.
Resident Assessment

    "

    "

Analgesic consumed

IM medications
IV medications

Tube feedings

Restraints

Therapies

Analgesic level received by WHO I, II and III levels, and daily frequency for multivariate analysis.
Any drug given intramuscularly.
Any drug or biological (e.g., contrast material give by IV push or drip) in the 7 days prior to assessment.
Presence of any tube that can deliver food/ nutritional substances, other directly into the gastrointestinal system.
Any use of truck or limb restraints or chair that prevents rising in 7 days prior to assessment.
Any speech, occupational or physical therapy in 7 days prior to assessment.
Resident Assessment and Drug Information

Resident Assessment

    "

    "

    "

Advance directives--
Do not resuscitate
Do not hospitalize
Feeding restrictions
Medication restrictions
Other treatment restrictions
Documentation of preference must also be present in resident's healthcare record.     "
Average Expenditures--
Hospice
  Routine home care
  Continuous home care
  Respite inpatient
  General inpatient
  Physician visits
Medicare Part A
  Acute care hospital
  Skilled nursing
  Home health care    
Total expenditures per category divided by decedents (hospice, non-hospice and total).
-- for 30 days (stratified by hospice lengths of stay of: <30 days and 30+ days)
-- for last 6 months (stratified by hospice lengths of stay of: <30 days, 30-59 days, 60-119 days and 120+ days)
Non-hospice decedents are placed in the length of stay category of the hospice case to which they are matched.
Medicare claims.

 

REFERENCES

Albert, M. and Cohen, C. The Test for Severe Impairment: An instrument for the assessment of patients with severe cognitive dysfunction. J Am Geriatr Soc. 1992 May; 40(5):449-53.

American Hospital Formulary Service (AHFS)Drug Information. Bethesda, MD: American Society of Hospital Pharmacists; 1994.

Berlowitz, D. R.; Brandeis, G. H., and Moskowitz, M. A. Using administrative databases to evaluate long-term care. J Am Geriatr Soc. 1997 May; 45(5):618-23.

Bernabei, R.; Gambassi, G.; Lapane, K.; Landi, F.; Gatsonis, C.; Dunlop, R.; Lipsitz, L.; Steel, K., and Mor, V. Management of pain in elderly patients with cancer. SAGE Study Group. Systematic Assessment of Geriatric Drug Use via Epidemiology [see comments]. JAMA. 1998 Jun 17; 279(23):1877-82.

Folstein, M. F.; Folstein, S. E., and McHugh, P. R. "Mini-mental state". A practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res. 1975 Nov; 12(3):189-98.

Frederiksen, K.; Tariot, P., and De Jonghe, E. Minimum Data Set Plus (MDS+) scores compared with scores from five rating scales. J Am Geriatr Soc. 1996 Mar; 44(3):305-9.

Gambassi, G.; Landi, F.; Peng, L.; Brostrup-Jensen, C.; Calore, K.; Hiris, J.; Lipsitz, L.; Mor, V., and Bernabei, R. on behalf of the SAGE study group. Validity of diagnostic and drug data in standardized nursing home resident assessment: Potential for geriatric pharmacoepidemiology. Medical Care. 1998; 36(2):167-179.

Hartmaier, S. L.; Sloane, P. D.; Guess, H. A., and Koch, G. G. The MDS Cognition Scale: A valid instrument for identifying and staging nursing home residents with dementia using the minimum data set [see comments]. J Am Geriatr Soc. 1994 Nov; 42(11):1173-9.

Hartmaier, S. L.; Sloane, P. D.; Guess, H. A.; Koch, G. G.; Mitchell, C. M., and Phillips, C. D. Validation of the Minimum Data Set Cognitive Performance Scale: Agreement with the Mini-Mental State Examination. J Gerontol A Biol Sci Med Sci. 1995 Mar; 50(2):M128-33.

Hawes, C. Regulation and the politics of long-term care. Generations Journal of the American Society on Aging. 1998; 21(4):5-9.

Hawes, C.; Morris, J. N.; Phillips, C. D.; Mor, V.; Fries, B. E., and Nonemaker, S. Reliability estimates for the Minimum Data Set for nursing home resident assessment and care screening (MDS). Gerontologist. 1995b Apr; 35(2):172-8.

Health Care Financing Administration (HCFA). Minimum data se plus multistate nursing home case mix and quality demonstration training manual. Natick, MA: Eliot Press; 1991.

Lawton, M. and Brody, E. Assessment of older people: Self-maintaining and instrumental activities of daily living. The Gerontologist. 1969; 9179-186.

Mor, V.; Allen, S., and Malin, M. The psychosocial impact of cancer on older versus younger patients and their families. Cancer. 1994 Oct 1; 74(7 Suppl):2118-27.

Morris, J. N.; Fries, B. E.; Mehr, D. R.; Hawes, C.; Phillips, C.; Mor, V., and Lipsitz, L. A. MDS Cognitive Performance Scale. J Gerontol. 1994 Jul; 49(4):M174-82.

Pahor, M.; Chrischilles, E. A.; Guralnik, J. M.; Brown, S. L.; Wallace, R. B., and Carbonin, P. Drug data coding and analysis in epidemiologic studies. Eur J Epidemiol. 1994; 10(4):405-411.

Phillips, C. D.; Chu, C. W.; Morris, J. N., and Hawes, C. Effects of cognitive impairment on the reliability of geriatric assessments in nursing homes. J Am Geriatr Soc. 1993 Feb; 41(2):136-42.

 

APPENDIX A. DATA AND VARIABLES USED FOR HOSPICE IN NURSING FACILITY ANALYSES

DATA SOURCES

Resident Assessment Instrument (the Minimum Data Set or MDS)

The Omnibus Budget Reconciliation Act of 1987 (OBRA '87) contained the most far-reaching revisions to the standards, inspection process and enforcement system in nursing facilities since the passage of Medicare and Medicaid in 1965 (Hawes, 1998). A major feature of this legislation was the introduction of a uniform, comprehensive resident assessment instrument (the MDS) to guide the clinical care planning process in order to systematically document residents' needs. The MDS is not only used to systematically assess the resident and to generate a comprehensive care plan to document clinical progress as that plan is implemented, but it is used by regulators to focus on resident outcomes and by facilities to improve their performance. In the time period studied here documentation of the resident assessments were required: at admission (by 15th day), quarterly (by 90th day), and annually (by 365th day). Reassessments were required when a resident was readmitted after hospital admissions and when significant change occurred. Resident assessments were to have been completed on all nursing facility residents cared for in facilities receiving any Medicare or Medicaid payment. These nursing facilities represent 96 percent of the facilities in the United States.

Topics covered in the MDS include cognitive function, communication/hearing problems, physical functioning, continence, psychosocial well-being, mood state, activity and recreation, disease diagnoses, health conditions/symptoms, nutritional status, oral/dental status, skin condition, special treatments, and medication use. A number of studies (Morris et al., 1994; Frederiksen et al., 1996; Hartmaier et al., 1994 & 1995; Phillips et al., 1993; Mor et al., 1994) demonstrate that researchers and clinicians using the MDS can achieve high levels of inter-rater reliability. Using an earlier version of the data set used in this study, Gambassi and colleagues found reasonably high levels of validity and good internal consistency comparing diagnoses on the MDS with HCFA claims and medical conditions with patterns of use of specific drugs (Gambassi et al, 1988). While the accuracy of the MDS data have been questioned (Berlowitz et al., 1997; Kramer et al., 199 ), and considerable anecdotal evidence reveals that some facilities have not taken the time to train their staff properly in its use, it is unlikely that misclassification errors in recording of information will be differential with respect to the outcomes of interest.

For the comparative study hospice and non-hospice patients had to have had at least 2 MDS assessments performed. The need for the presence of 2 MDS assessments for our comparative analyses was originally recognized after preliminary analysis documented the presence of ascertainment bias on selected symptoms. Specifically, hospice residents were significantly more likely to have pain and dyspnea recorded than were non-hospice decedents. For example, controlling for other patient factors, residents with a dementia diagnoses were 3 times as likely to have pain recorded than were non-hospice residents. Therefore, to more correctly represent the hospice influence on the presence and management of symptoms (rather than merely the increased likelihood of hospice to assess symptoms) we felt that it was necessary to control for the status of symptoms at the time of the penultimate MDS.

Health Care Financing Administration (HCFA) Claims Data

The HCFA claims data were merged to the MDS file using the Health Insurance Claim number of Medicare beneficiaries. To ensure confidentiality, these identifiers were replaced with unique identifiers using the claim number as a seed. Two files (beneficiary information and claims data) comprise the HCFA data. The beneficiary file (Denominator file) contains gender, date of birth, and survival status (verified date of death). The claims data used include all Medicare Part A claims including hospital, skilled nursing facility, hospice, and home health agency claims. We achieved a match rate of MDS data to HCFA beneficiary data of approximately 85 percent using HCFA data from 1991through 1997.

Drug Data

As part of the resident assessment, nursing facility staffs code up to eighteen drugs taken within the seven days preceding the assessment. Nursing home staffs code each drug according to the National Drug Coding (NDC) system. Field tests of the MDS showed that 100 percent of the medication use items were reliable with the average reliability being 0.73 (Hawes et al., 1995b). While prescription drug products must use an NDC, most non-prescription drugs are also primarily referenced by the NDC. NDCs are unique 11-digit codes that identify discrete drug products. The first five digits refer to the manufacturer. The next four digits correspond to the drug product. The last two digits indicate the packaging. As pharmaceutical companies merge, new products are introduced, and drugs are no longer active, changes in NDCs occur continuously. Consequently, the NDCs are commercially-oriented and do not contain any mechanism to group drugs according to ingredients or categories of ingredients. Therefore, linking NDCs to specific descriptive information is critical to enable research. This NDC matching entails several steps. To match the NDC codes, we used a historical reference archive for drug products that listed all NDCs ever attributed and eventually discontinued between 1991 and 1996. For scientific drug research, we translated NDC codes into a hierarchical therapeutic classification scheme as recommended by WHO (Pahor et al., 1994). NDC codes were merged to useable therapeutic class and sub-class information using the Master Drug Data Base (MediSpan™) (1995). MediSpan™ contains complete records for prescriptions common in retail pharmacy as well as unit-dose and injectables used by hospitals and external facilities. MediSpan™ now includes over 100,000 generic drug products, products from regional manufacturers, and information on over 90,000 inactive drugs. The hierarchical identifier, the Generic Product Identifier (GPI) contained in MediSpan™, is a 14-character field consisting of seven subsets, each providing increasingly more specific information about the drug. (See below example.) While MediSpan™ incorporates the American Hospital Formulary Service (AHFS) (1994), a classification system based on the pharmacological uses of drugs, MediSpan™ also groups drugs with comparable compounds in the same therapeutic class and allows the same drug to be classified into multiple therapeutic classes.

Medi-span™ Classification System - Example of an Antidepressant
GPI Coding Example
58- Drug group Antidepressants
58-20- Drug class Tricyclic agents
58-20-00- Drug sub-class --
58-20-00-60 Drug name Nortiptyline
58-20-00-60-10 Drug name extension Hydochloride
58-20-00-60-10-01 Dosage form 10mg

A recent study analyzed the MDS drug codes with respect to: (1) completeness; (2) internal consistency; and (3) external validity (Gambassi et al., 1998). Investigators found the overall match rate between the NDC and the MediSpan™ greater than 90 percent with only 5.4 percent of the original NDC codes contained in the MDS data from the states being studied in this project to be incomplete or incorrect. Gender-specific medications had a high concordance with gender (>90 percent). For example, all residents taking tamoxifen were women; all residents taking goserelin were men; and 92 percent of estrogen users were women. Cross-linkages between drugs and MDS condition variables revealed adequate to high rates of concordance (range: 51 percent (gout) - 100 percent (rheumatoid arthritis)). High rates of concordance were reported when cross-checking levo-dopa with Parkinson's disease (88.9 percent); hypoglycemic agents with diabetes mellitus (93.2 percent); and sore care products with pressure sores (83.7 percent). These data show that the MDS drug data are consistent and reliable (Gambassi et al., 1998).

Hospice Provider of Service File

The Provider of Service (POS) File is compiled and managed by HCFA to determine the capacity of Medicare/Medicaid institutional providers to render acceptable care. This file contains information on program characteristics, collected by State surveyors under Federal guidelines. Since the periodic inspection of hospices is not mandated, surveys are conducted according to state priorities and resources, and, as a consequence, hospice and nursing facility data does not match across time. We used the hospice provider number on the HCFA hospice claim to link the hospice provider information to MDS and claims data. For our descriptive analyses, we used 1995 hospice provider information. For our analyses we were most interested in hospice provider type (freestanding or home care, hospital or nursing facility based), and ownership.

 

VARIABLES

Table of Variables

Table A1 lists all variables studied in both the descriptive and comparative hospice in nursing facility analyses. Variable measurement and data source are shown. For selected variables, the text below provides more information on their documentation and/or their reliability and validity.

Pain

Pain was not a major focus of the MDS version used in these states during the time period in question, and, as such, there are several limitations to the measure used in these analyses. A major limitation is that the level of pain intensity is not recorded; the newer MDS 2.0 requires documentation of pain intensity.

The MDS pain data used in this study is based on assessment by nursing personnel which is supposed to be performed according to instructions provided in the MDS manual (HCFA, 1991). Nursing home personnel are supposed to evaluate signs and symptoms of pain, but since pain is a subjective experience, they are instructed to record whatever the residents said it was. Residents were to be asked whether they had experienced any pain in the last seven days. Furthermore, residents were to be asked to describe the pain and how often it was manifest. To elicit complete and satisfactory answers, the assessors were instructed to ask neutral and non-directive questions. Questions such as: "What do you mean?" "Tell me what you have in mind." "Tell me more about that." "Please be more specific." "Give me an example". Moreover, the assessors were instructed to validate their understanding of what the resident was really saying. Statements like "I think I hear you saying that ….." or "Let's see if I understood you correctly. You said ….. Is it right?" were suggested in the MDS Instruction Manual.

For MDS assessment purposes, pain refers to any type of physical pain or discomfort in any part of the body experienced on a daily basis. If the assessor had difficulty discriminating the frequency, the instructions were to code as daily. Pain could have been localized or more generalized. It could have been acute or chronic, continuous or intermittent, occurring at rest or with movement. Pain recording could have depended exclusively on the observation of signs of pain. According to the MDS Manual, these include moaning, crying, and other vocalizations; wincing or frowning and other facial expressions; or body posture such as guarding/protecting an area of the body, or lying very still. In these cases, the assessors were instructed to ask the nurse assistants and therapists who might have been working with the resident, whether he/she had complaints or signs of pain during their shifts. In some residents, those who have dementia and cannot verbalize the pain experience, the assessor was instructed to look for particular behaviors such as calling out for help, pained facial expressions, refusing to eat, or striking out at a nurse assistant who was trying to move them or touch a body part.

Dyspnea

Shortness of breath (dyspnea) is recorded on the MDS if the problem is present in the "last 7 days" prior to completion of the MDS assessment. The degree or frequency of dyspnea is not indicated, only its presence. We compare the presence or absence of dyspnea in hospice versus non-hospice decedents stratifying for its presence on the penultimate MDS assessment.

Persistent Mood Disturbance

Persistent mood disturbance is defined as "persistent sad or anxious mood that has existed over the last 7 days and was not easily altered by attempt to "cheer up" the resident." (MDS+ Reference Manual, 1993) For MDS assessment purposes, a sad or anxious mood is a distressed mood that is characterized by explicit verbal or gestural expressions of feeling depressed or anxious (or a synonym such as feeling sad, miserable, blue, hopeless, empty, or tearful). Assessors are instructed to draw upon their own interactions with the residents as well as to statements of direct-care staff, social workers, and licensed personnel who may have evaluated the resident in this area. Suggested cues are: Does the resident cry or look dejected (unhappy) when no one is talking with him or her? When you talk with the resident, does he or she sound hopeless, fearful, sad, anxious? Does the resident appear withdrawn, apathetic, without emotion? (MDS+ Reference Manual, 1993)

Cognitive Performance

The MDS includes seven direct measures of cognition: short and long term memory, recall or orientation items (season, location or room, staff names/faces, orientation to nursing home), and decision-making ability. Good reliability (0.7) of these items has been reported (Hawes et al, 1995b). The cognitive performance scale (CPS) used in this study is a categorical measure of cognition using these MDS items and several items which indirectly evaluate cognitive function (i.e. comatose state, total dependent eating) (Morris et al., 1994). Based on two standard cognitive assessment tools, the Mini-Mental State Examination (Folstein et al., 1975) and the Test for Severe Impairment (Albert & Cohen, 1992), the CPS has excellent reliability with estimates published in the range of 0.66 - 0.88 (159). Using the MMSE as the gold standard, the CPS has high sensitivity (>90 percent) and specificity (>85 percent), yielding high diagnostic accuracy, regardless of patient education level (Hartmaier et al., 1995). Furthermore, the CPS has excellent reproducibility (Kappa >0.76) (155).

Activities of Daily Living

The reliability of the ADL scores range from 0.87-0.92 (Hawes et al., 1995b) and is highly correlated (0.89) with the Physical Signs and Symptoms Scale (Lawton & Brody, 1969). Furthermore, a recent study found these measures useful in pharmacoepidemiologic studies (Bernabei et al., 1998).

TABLE A1. Comparative Descriptive and Analytic Analyses--Variables, Measures, and Data Sources Outcomes
Type Variable Empirical Measure Source Data
Patient Acute Care Hospitalization and Average Hospital Days
-- in 30 days prior to death
-- in 90 days prior to death
-- in 6 months prior to death

Hospice decedents in hospice for total time period studied (and their matched controls) who have acute care hospitalization/ hospice decedents in hospice for total time period studied (and their matched controls).

Days in hospital in time period for hospice decedents in hospice for total time period studied (and their matched controls)/ hospice decedents in hospice for total time period studied (and their matched controls).

Medicare Part A Hospital Claims
Pain Management (regular treatment) Decedents in pain and receiving WHO level analgesia at least twice a day, or, for level III drugs, having a drug patch. Resident Assessment and Drug Information
Persistent Mood Disturbance (treatment of) Decedents with persistent mood disturbance and receiving antiananxiety or antidepressant medication in the 5 to 7 days prior to MDS assessment data. Resident Assessment
Independent Variables
Type Variable Empirical Measure Source Dataset
Patient Hospice enrollment Resident elected hospice prior to nursing facility admission. Medicare hospice claims and Resident assessment
Covariates and Variables for Descriptive Comparisons
Type Variable Empirical Measure Source Dataset
Facility-
  Nursing
  Facility

Facility-
  Hospice

Hospice Concentraction Unduplicated nursing facility residents on Medicare hospice in given year/ unduplicated nursing facility residents in a given year. Medicare hospice claims and Resident Assessment
For-profit ownership

Government ownership

Organizational type

A for-profit organization controls and operates the hospice.
A government entity controls and operates the hospice.
Hospice is freestanding or under administrative control of a hospital, home health agency or nursing home.
Hospice Provider of Service File
    "

    "

Level Variable Empirical Measure Source Dataset
Patient State of residence

Year of death
Race/ Ethnicity

State location of nursing facility in which resident resides
Year in which resident died
White/ Afr.Am./ Latino/ Native Am./ Asian
Resident Assessment
HCFA Denominator File
    "
Gender Female/ Male     "
Age Years     "
Marital Status Married/ Widowed or Divorced/ Separated/ Never married - most recent status Resident Assessment
Activities of daily living ADL score - see text     "
Cognitive performance CPS score - see text     "
Body Mass Index Weight (kg)/ height2 (m2) Resident Assessment
Diagnosis Diagnostic categories of (1) Cancer without dementia (2) Cancer with dementia (3) Alzheimer's disease/ dementia, (4) Other Resident Assissments, Inpatient Claims 6 months before NF adm. & during stay
Individual selected diagnoses for multivariate analysis. Resident Assessment
Pain

Dyspnea

Persistent Mood Disturbance

Vomiting

Complains or shows evidence of pain daily or almost daily (in 7 days prior to assessment).
Difficulty breathing occurring at rest, with activity, in response to illness or anxiety, or when lying flat.
Persistent sad or anxious mood that has existed over the last 7 days and was not easily attended by attempt to "cheer up" the resident.
Vomiting in 7 days prior to assessment.
Resident Assessment

    "

    "

Analgesic consumed

IM medications
IV medications

Tube feedings

Restraints

Therapies

Analgesic level received by WHO I, II and III levels, and daily frequency for multivariate analysis.
Any drug given intramuscularly.
Any drug or biological (e.g., contrast material give by IV push or drip) in the 7 days prior to assessment.
Presence of any tube that can deliver food/ nutritional substances, other directly into the gastrointestinal system.
Any use of truck or limb restraints or chair that prevents rising in 7 days prior to assessment.
Any speech, occupational or physical therapy in 7 days prior to assessment.
Resident Assessment and Drug Information

Resident Assessment

    "

    "

    "

Advance directives--
Do not resuscitate
Do not hospitalize
Feeding restrictions
Medication restrictions
Other treatment restrictions
Documentation of preference must also be present in resident's healthcare record.     "
Average Expenditures--
Hospice
  Routine home care
  Continuous home care
  Respite inpatient
  General inpatient
  Physician visits
Medicare Part A
  Acute care hospital
  Skilled nursing
  Home health care    
Total expenditures per category divided by decedents (hospice, non-hospice and total).
-- for 30 days (stratified by hospice lengths of stay of: <30 days and 30+ days)
-- for last 6 months (stratified by hospice lengths of stay of: <30 days, 30-59 days, 60-119 days and 120+ days)
Non-hospice decedents are placed in the length of stay category of the hospice case to which they are matched.
Medicare claims.

 

REFERENCES

Albert, M. and Cohen, C. The Test for Severe Impairment: An instrument for the assessment of patients with severe cognitive dysfunction. J Am Geriatr Soc. 1992 May; 40(5):449-53.

American Hospital Formulary Service (AHFS)Drug Information. Bethesda, MD: American Society of Hospital Pharmacists; 1994.

Berlowitz, D. R.; Brandeis, G. H., and Moskowitz, M. A. Using administrative databases to evaluate long-term care. J Am Geriatr Soc. 1997 May; 45(5):618-23.

Bernabei, R.; Gambassi, G.; Lapane, K.; Landi, F.; Gatsonis, C.; Dunlop, R.; Lipsitz, L.; Steel, K., and Mor, V. Management of pain in elderly patients with cancer. SAGE Study Group. Systematic Assessment of Geriatric Drug Use via Epidemiology [see comments]. JAMA. 1998 Jun 17; 279(23):1877-82.

Folstein, M. F.; Folstein, S. E., and McHugh, P. R. "Mini-mental state". A practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res. 1975 Nov; 12(3):189-98.

Frederiksen, K.; Tariot, P., and De Jonghe, E. Minimum Data Set Plus (MDS+) scores compared with scores from five rating scales. J Am Geriatr Soc. 1996 Mar; 44(3):305-9.

Gambassi, G.; Landi, F.; Peng, L.; Brostrup-Jensen, C.; Calore, K.; Hiris, J.; Lipsitz, L.; Mor, V., and Bernabei, R. on behalf of the SAGE study group. Validity of diagnostic and drug data in standardized nursing home resident assessment: Potential for geriatric pharmacoepidemiology. Medical Care. 1998; 36(2):167-179.

Hartmaier, S. L.; Sloane, P. D.; Guess, H. A., and Koch, G. G. The MDS Cognition Scale: A valid instrument for identifying and staging nursing home residents with dementia using the minimum data set [see comments]. J Am Geriatr Soc. 1994 Nov; 42(11):1173-9.

Hartmaier, S. L.; Sloane, P. D.; Guess, H. A.; Koch, G. G.; Mitchell, C. M., and Phillips, C. D. Validation of the Minimum Data Set Cognitive Performance Scale: Agreement with the Mini-Mental State Examination. J Gerontol A Biol Sci Med Sci. 1995 Mar; 50(2):M128-33.

Hawes, C. Regulation and the politics of long-term care. Generations Journal of the American Society on Aging. 1998; 21(4):5-9.

Hawes, C.; Morris, J. N.; Phillips, C. D.; Mor, V.; Fries, B. E., and Nonemaker, S. Reliability estimates for the Minimum Data Set for nursing home resident assessment and care screening (MDS). Gerontologist. 1995b Apr; 35(2):172-8.

Health Care Financing Administration (HCFA). Minimum data se plus multistate nursing home case mix and quality demonstration training manual. Natick, MA: Eliot Press; 1991.

Lawton, M. and Brody, E. Assessment of older people: Self-maintaining and instrumental activities of daily living. The Gerontologist. 1969; 9179-186.

Mor, V.; Allen, S., and Malin, M. The psychosocial impact of cancer on older versus younger patients and their families. Cancer. 1994 Oct 1; 74(7 Suppl):2118-27.

Morris, J. N.; Fries, B. E.; Mehr, D. R.; Hawes, C.; Phillips, C.; Mor, V., and Lipsitz, L. A. MDS Cognitive Performance Scale. J Gerontol. 1994 Jul; 49(4):M174-82.

Pahor, M.; Chrischilles, E. A.; Guralnik, J. M.; Brown, S. L.; Wallace, R. B., and Carbonin, P. Drug data coding and analysis in epidemiologic studies. Eur J Epidemiol. 1994; 10(4):405-411.

Phillips, C. D.; Chu, C. W.; Morris, J. N., and Hawes, C. Effects of cognitive impairment on the reliability of geriatric assessments in nursing homes. J Am Geriatr Soc. 1993 Feb; 41(2):136-42.