Use of medicare services before and after introduction of the prospective payment system

Health Services Research,  August, 1993  by Kenneth G. Manton,  Max A. Woodbury,  James C. Vertrees,  Eric Stallard

• E-mail
• Print
• Link

Medicare's prospective payment system (PPS) for reimbursing inpatient hospital stays, phased in between October 1, 1983 and September 30, 1984, sought to improve efficiency by paying a fixed amount for treating cases in specific diagnosis-related groups (DRGs). Because the risk that costs would exceed DRG payments could have motivated hospitals to discharge patients sooner than medically warranted, physician and hospital reimbursements were separated. Consequently, hospital profit is higher the shorter the stay, but the physician's fee is lower. The physician's and hospital's risk of malpractice liability reduces incentives for premature discharge. To compensate for shorter stays, hospitals may discharge patients to skilled nursing facilities (SNFs) or home health agencies (HHAs) for post-acute care -- changes intended in PPS.

Concern about the effect of PPS on quality of care required monitoring to determine if hospital and physician incentives were balanced. Hospital length of stay (LOS) decreased and post-acute care use increased (DesHarnais, Kobrinski, Chesney, et al. 1987; Iezzoni 1987; Manton and Liu 1990) with few adverse effects. Sager et al. (1989) suggest that some patients may have been discharged early to SNFs unequipped to provide intensive care -- a finding not confirmed in other analyses (e.g., Manton, Vertrees, and Wrigley 1990).

PPS also changed service use. Comparison of services used before, and after, PPS requires controlling for (a) health and (b) substitution of different durations and types of service (e.g., substituting home health visits for hospital days). We formed case-mix groups using "Grade of Membership" (GoM) procedures applied to reports of chronic (90 days +) impairment of activities of daily living (ADL) (Katz and Akpom 1976), instrumental activities of daily living (IADL) (Lawton and Brody 1969), physical performance, and medical conditions from the 1982 and 1984 National Long Term Care Surveys (NLTCS). In addition to describing a group's profile of health traits, GoM generates scores representing the closeness with which an individual's traits match each profile. These scores are used to control for health in constructing life tables describing the use of services over time.

To analyze both the duration and substitution of services, life tables for each health group were estimated for each type of service. Service use life tables for individuals are calculated by weighting life tables for the K groups by an individual's scores (which add up to one). Average scores (which also sum to one) can be used to weight tables to represent the service use in a population with a specific case-mix distribution. Life tables for two populations, standardized to a common case-mix distribution, can be used to compare service differences. GoM has been used to analyze patients using nursing homes (Manton, Vertrees, and Woodbury 1990), home health (Manton and Hausner 1987), hospitals (Vertrees and Manton 1986), and long-term care (LTC) demonstrations |e.g., the HIO (Vertrees, Manton, and Mitchell 1989); and the "2176" program (Vertrees, Manton, and Adler 1989)~.

By using nationally representative surveys (the 1982, 1984 NLTCS), we can examine the health, long-term care use, and out-of-pocket costs of persons not using acute care. Other studies show that PPS did not increase hospital mortality or short-term readmission. We analyzed change in the rate and volume of use of Medicare home health services by each health group overall, and specific to marital status and age. The NLTCS thus provides health data on impaired persons who may have needed, but did not receive, Medicare services, that is, people with chronic health problems who were not admitted to a hospital post-PPS, but who might have been admitted pre-PPS.

DATA

National Long-Term Care Surveys

The NLTCS represent the entire elderly (65 +) Medicare-eligible population in both 1982 and 1984. The health data from the surveys (which use the same instruments, field staffs, and survey methods) are not available in Medicare records. In conducting the 1982 NLTCS, since the prevalence of chronic disability for persons age 65 + was not known (i.e., disability lasting, or expected to last 90 + days, due to health or age on at least one ADL or IADL), 55,000 persons were sampled from Medicare files. Initially, 15,000 persons were screened. It was determined that by screening 35,000 + cases, over 6,000 chronically disabled community residents could be identified. Of the 35,789 screened, 781 were dead (or had moved) by April 1, 1982; 489 died between April 1, 1982 and the screen. Screens were completed (80 percent by telephone and 20 percent in person) with 34,012 survivors (98.5 percent) identifying 6,393 community residents with chronic disability, 25,627 without chronic disability, and 1,992 in institutions. Of the 6,393 total community residents, 67 died and 58 were institutionalized before the detailed interview; 6,268 community residents were contacted, of whom 6,088 responded. The 97.1 percent interview and 98.5 percent screen rates yielded a two-stage response of 95.6 percent, excluding deaths. Reports of deaths were used to estimate mortality, an outcome of importance.