Articles Changes in Case Mix and Outcomes of Readmissions to Nursing Homes Between 1980 and 1984 Mary Ann Lewis, Barbara Leake, Virginia Clark, and Margaret Leal-Sotelo This study compares the personal characteristics, measures offunctional status! case mix, and immediate discharge outcomes of two cohorts of nursing home patients (1980 and 1984). All of these patients had a prior history of nursing home care and all were readmitted to skilkd nursing facilities from hospitals. In 1984, readmissions were more disabkd, more debilitated, and significantly less likely to return home. They were almost twice as likely to be rehospitalized within 30 days ofdischargefrom the hospital (26. 7 percent versus 48. 9 percent). Analyses showedpoorer health status and an increased proportion ofnursing home deaths and rehospitalizations in the 1984 group to be a function of time (1984 versus 1980) rather than- of insurance coverage (Medicare versus other). Nursing home readmissions appear to be quite sensitive to cost-containment efforts, and they may require additional hospital days to stabilize their conditions in an effort to reduce the rate of hospital readmissions and the overall cost of care.
This article was supported in part by a grant (HS 04946) from the National Center for Health Services Research and by computer support from the School of Nursing at the University of California, Los Angeles. Address correspondence and requests for reprints to Mary Ann Lewis, Dr.Ph., R.N., Adjunct Professor of Nursing and Medicine, University of California, Los Angeles, School of Nursing, 10833 Le Conte Avenue, Los Angeles, CA 90024-6917. Barbara Leake, Ph.D. is Senior Statistician in the Division of General Internal Medicine (GIM) and Health Services Research (HSR), Department of Medicine, School of Medicine; Virginia Clark, Ph.D. is Professor of Biostatistics, School of Public Health; and Margaret Leal-Sotelo, M.S.W. is a Research Assistant in the Division of GIM/ HSR, Department of Medicine, School of Medicine, all at the University of California, Los Angeles.
HSR: Health Services Research 24:6 (February 1990)
The advent of the Medicare diagnosis related groups (DRGs)-based prospective payment system (PPS) for financing the care of the elderly has provoked considerable debate and concern about the effect of federal cost-containment policies on the quality of care provided to elderly patients (Stern and Epstein 1985; Williams 1986). There are studies in progress to address allegations of reduced quality of care in short-term general hospitals related to cost-containment efforts stimulated by the prospective payment system (Eggers 1987). Others have noted the potential effect on the long-term care sector, particularly in the transfer of patients whose care needs may exceed the resources available in such facilities (Lewis, Kane, Cretin, et al. 1985; Meiners and Coffey 1985; Shaughnessy, Kramer, Schlenker, et al. 1985). Two recent studies provide some evidence of changes in the patterns of patient care following the PPS. In Wisconsin, 12 months following the implementation of the PPS, the rate of hospitalization among the institutionalized elderly Medicaid population increased 72 percent. Their length of hospital stay decreased 26 percent and, while fewer of these patients died in the hospitals, the proportion dying in nursing homes increased by 26.2 percent (Sager, Leventhal, and Easterling 1987). In Indiana, a study of elderly patients (N = 70) with an initial hip fracture showed a reduction in average length of stay from 16.6 days in 1981 to 10.3 days in 1985. Patients received fewer physical therapy sessions (9.7 versus 4.9) and were more likely to be discharged to a skilled nursing facility in 1985 (48 percent versus 21 percent). The proportion of these patients who remained in that facility for six months increased over this time period from 13 to 39 percent. We previously reported the admission characteristics, initial discharge outcomes (Lewis, Kane, Cretin, et al. 1985), and natural history of nursing home patients admitted to skilled nursing facilities for the first time (Lewis, Cretin, and Kane 1985). We also demonstrated that there were no significant changes over time in case mix or outcomes of first admissions, that is, persons admitted to nursing homes for the first time (1980, 1982/1983 and 1984) (Lewis, Leake, LealSotelo, et al. 1987). However, persons readmitted to nursing homes following an acute care hospitalization differ from first admissions in that they are sicker and more debilitated. Any changes in health care policies that promote early discharge of patients whose conditions are more fragile are likely to affect those with prior needs for hospitalization - a proxy for prior health status. Any adverse effects resulting from changes in hospital practices should be evident in a period immediately following hospitalization, with either increased death rates or readmissions to hospitals within a short period of time.
Readmissions to Nursing Homes
In fact, Eggers (1987) suggests that an assessment of the effect of the PPS needs to be done within the context of rates of readmission to hospitals and postadmission mortality rates. However, because of the multiplicity of factors related to postreadmission mortality rates, a simple association cannot be construed as inferring causality. In this article, we present data from 1980 and 1984 contrasting patients with a prior history of nursing home care (readmissions) who were transferred from hospitals to nursing homes, in terms of personal characteristics, functional status, and discharge outcomes. In the analyses, we test the hypothesis that changes in hospital practices and patterns of care between 1980 and 1984 would be associated with significant changes in the case mix (severity) and outcomes of care among patients readmitted to nursing homes in 1984 compared to readmissions in 1980.
METHODS The study population includes a 1980 discharge cohort (N = 289) and a 1984 (July-September) admission cohort (N = 334) of nursing home readmissions who came from hospitals. The 1980 sample includes patients from 24 nursing homes that were stratified by bed size and randomly selected from 47 nursing facilities located in the San Bernardino/Riverside/Ontario Standard Metropolitan Statistical Area (SMSA) in Southern California. All 47 homes were classified as nonhospital-based skilled nursing facilities. For the 1980 sample, discharge records were selected within each nursing home on a random basis proportional to that facility's contribution to the total number of discharges during 1980 from all 24 nursing homes. The 1984 sample includes patients in 45 of the same 47 facilities from which the 1980 sample was drawn. The 1984 sample is derived from admissions to all 45 nursing homes during the 90-day period July 1-September 30, 1984. We have nursing home discharge status on 90 percent of this 1984 admission cohort. Only one nursing home of the 25 chosen at random refused to participate in the 1980 study; 2 of 47 declined in 1984. The 1984 admission cohort was part of a pilot study to assess the effect of hospital cost-containment policies on patients admitted to nursing homes. An admission versus a discharge sample was chosen due to the constraints of obtaining an adequate discharge sample within the pilot study period. In the 1980 discharge cohort, 62.9 percent of the patients were admitted in 1980; 22.4 percent in 1979; 7.2 percent in 1978; and 7.5 percent in 1977 or earlier.
HSR: Health Services Research 24:6 (February 1990)
The patient descriptor variables abstracted from the admission notes of nursing home medical records for both cohorts included the physician's hospital discharge summary and the hospital's nursing discharge plan. Trained nurses abstracted the records; 10 percent were abstracted a second time to assure a reliability of over 95 percent. Personal characteristics included age, sex, marital status, and location prior to admission, that is, hospital, nursing home, or home in the community. History of prior admissions to nursing homes was recorded to identify first admissions and readmissions. Because of interest in changes of case mix and increased coverage by Medicare (Lewis, Leake, Leal-Sotelo, et al. 1987), payment source was classified as Medicare (Medicare supplemented by other payments) versus other (Medicaid, self-payment, and/or private insurance). Source of payment was also recorded at the time of discharge to check the validity of the insurance classification of admissions claimed to be eligible for Medicare. Social support was measured primarily by the presence or absence of visitors during the first month of residence in the nursing homes. Functional-status variables included mental status (oriented, confused, comatose), bladder and bowel continence, mobility, and ability to perform self-care and activities of daily living (ADL). The ADL score at admission recorded on the functional status assessment form was obtained by summing the amount of help (1 = can do alone; 2 = can do with help; 3 = must be done for) required in six areas (eating, bathing, dressing, grooming, toileting, and transferring). A score of 6 indicated independent performance, and a score of 18 indicated total dependence. A seven-point ordinal scale of mobility was used that ranged from 1 = walks alone to 7 = bed-confined. The medical diagnoses were abstracted as recorded. Discharge outcomes were classified as: (1) returned home to the community, (2) transferred to another nursing home, (3) died in the nursing home, (4) transferred to the hospital and died, or (5) transferred to the hospital and survived. Patients in both cohorts were followed for a period of two years after entry into the study. ANALYTICAL TECHNIQUES
Chi-square analyses were conducted to examine the relationship of the sociodemographic descriptors, functional status variables, diagnoses, and discharge outcomes of patients in each of the two cohorts. Loglinear analyses were also performed to simultaneously examine relationships among (1) the two time periods; (2) insurance status (Medicare versus other); and (3) patient variables of interest, for example, dis-
Readmissions to Nursing Homes
charge outcomes and discrete case-mix characteristics such as being comatose or disoriented. Insurance status was differentiated into patients supported by Medicare, with or without supplementation from other sources, compared to those who received no Medicare benefits. Since exact length of stay (LOS) information could not be determined for 1984 patients who remained in the nursing home two years after admission, their LOS was set to exactly two years. To achieve comparability between cohorts, LOS for patients in the 1980 cohort who were discharged more than two years after admission was also set to two years. We chose to truncate these data, and establish equal periods of follow-up, rather than to eliminate those persons with LOS greater than two years, in order not to lose or discard data on the twoyear period of observation. This recoding did not affect calculation of median values for LOS since relatively few patients were involved. Differences in LOS distributions between the cohorts were assessed using the Kruskal-Wallis test. Finally, logistic regression was used to examine differences between the two cohorts with regard to major discharge outcomes while controlling for case mix (Dixon 1988). For example, in one logistic-regression analysis, patients discharged home to the community were contrasted with those who died in the nursing home or were discharged elsewhere. These analyses were performed on a total of 524 patients as a result of missing data. Because multiple tests of significance were performed in this study, only differences significant at the p < .01 level are presented.
RESULTS For the two cohorts, Table 1 presents the median length of stay in the nursing home, sociodemographic and functional-status characteristics, common diagnoses, and discharge outcomes for all readmissions. As shown, the proportion of Medicare patients increased from 19.4 percent for readmissions in 1980 to 52.6 percent for readmissions in 1984. Median length of stay in the nursing home also increased dramatically between 1980 and 1984 almost doubling from approximately two to four months. In Table 1, a comparison of readmissions from 1980 to those of 1984 showed significant differences in 12 of the 23 variables. In addition to the increase in the proportion covered by Medicare and the proportion at least 75 years of age, 1984 readmissions were more disoriented, bed-confined, and/or incontinent, and more had catheters. Readmissions in 1984 were more likely to have a diagnosis of demen-
HSR: Health Services Research 24:6 (February 1990)
Table 1: Characteristics and Outcomes of All Nursing Home Readmissions Transferred from Hospitals to Nursing Homes in 1980 and 1984 Number of patients Median LOSt
1980 Readmissions 1984 Readmissions* 289 334 68 118.5 Percentages
Personal Characteristics Male