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

EFFECTS OF PRIOR SERVICE USE

We examined the dependency of current service use on prior service use. Ideally, case mix is based on health. However, if health data are limited, then service use, after controlling for case mix, may be correlated over time. To test this, we redefined the four groups by adding four variables representing Medicare services used in the 12 months prior to the index episode. This was in order to increase the number of variables in the GoM analysis from 56 to 60. This addition of variables did not improve the ability to predict service use or mortality. Differences in 1982-1984 between mortality rates for the four groups, with and without prior service use, were small compared to differences either between 1982 and 1984 or across groups (e.g., for TABULAR DATA OMITTED hospital stays in 1984, the frail had five times the mortality of the acute medical problems group). Thus, prior service use did not add information to that in the case-mix groups.

SUMMARY AND CONCLUSIONS

Pre- and post-PPS service use was analyzed. Hospital LOS and admission rates (and numbers) declined after PPS was put in place. Changes in the timing and place of death showed little evidence of adverse effects. Changes in services used by the chronically disabled elderly suggested the importance of targeting quality assurance activities. LOS and HHA use increased for certain groups (those with acute medical problems, for example). Hospital and HHA changes were reasonable for each health group. Changes in post-acute care showed substitution of HHA for hospital days post-PPS. Although HHA service was used to supplement hospital care for unmarried people and for people over 80, controlling for case mix, hospitals did recognize the informal care available to married people. Younger people were found to be less likely to die in the hospital post-PPS, while the likelihood increased for people over age 80. Thus, post-acute care use after PPS did not decline in quality as measured by (a) differences in service use by health groups with different needs, and (b) mortality. By supplementing acute care with post-acute care -- especially for unmarried persons, PPS controlled costs (Russell and Manning 1989) without adversely affecting quality of care.

There is concern that DRGs have not adequately reflected severity of illness, although PPS day and cost outliers, and "teaching hospital" adjustments have done so implicitly. Proposals have been made to increase the number of DRGs so that severity of illness can be better represented. The original set of 474 DRGs is complex. Elaborated systems, which would be even more difficult to manage, would impose heavier burdens on medical and administrative data systems making case review and auditing more difficult. The results cited here suggest that such complexity is not necessary because PPS has been flexible enough to control costs while responding appropriately to medically complex cases.

Instead of increasing the number of DRGs, one could use scores for individuals on a small number of health dimensions to determine reimbursement. This has advantages. First, DRGs are based on historical service patterns. As technology continues to evolve, services continue to change, as does the use of services by different patient types. The health variables used to define groups in a "scored" system, however, do not change. Second, case mix formed from historical data may perpetuate allocation problems. For example, in RUGS-II, because persons with dementia and behavior problems had received few services, their reimbursement was low (Manton, Vertrees, and Woodbury 1990). Clinical evaluation suggested that they needed more resources than they had historically received. Third, assigning individuals to "categories" means that persons merely fulfilling group criteria have higher rates of misclassification. Scores calculated from multiple measures are less affected by measurement error than are category assignments. Thus, the use of scores, which focuses attention on care outcomes for persons with specific health problems, meets the original goal of PPS while it preserves the fiscal stability of small providers.