ity is observed between 1982 and 1983, this is attributed to patient characteristics, data reporting, and unexplained factors.1 In any case, these data do not bear directly on whether changes in Medicare ESRD affect mortality.
The committee is aware of only two studies that attempt to assess this question directly. Both focused on the effects of implementation of the 1983 ESRD composite rate, which significantly reduced reimbursement to many facilities. The first, by Held and his colleagues at the Urban Institute (1987),2 noted a general increase in reported mortality, suggested that the increasing age of the patient population might account for much of this change, and observed that changing patient selection criteria might account for some increase in the proportion of sicker patients. Specifically, with respect to the effect of the composite rate on mortality, the study reported that ''evidence of a correlation between changes in mortality and the extent of composite rate changes was not found, i.e., mortality did not rise more where the composite rate changes were larger'' (Held et al., 1987) The authors noted, however, that further analysis was needed to draw firm conclusions.
This 1987 Urban Institute study was done under pressure of time. The present IOM study subcontracted with the Urban Institute to perform further analysis and to update its earlier study. More follow-up data were evaluated, and an improved research design was used. This new report (Held et al., 1990b) uses patient data from two nonconcurrent, prevalent cohorts of patients who were in the system on January 1, 1982, and January 1, 1984. It uses price data for the 1982 and the 1984–87 reimbursement rates adjusted for area wage differences to capture the real resource costs across areas, an adjustment not undertaken in the earlier study.
Two types of analyses ("models") are presented. The price-level model analyzes whether mortality is associated with variations in price levels among facilities at a given time. It analyzes whether mortality is higher at facilities receiving lower standardized prices (payments) during a specific year. Data from 1982 and from 1984 were analyzed in the new report. The first-difference model uses each facility as its own control by comparing the mortality rates in each facility at two different times. It examines whether the mortality at a facility changed when the price it received changed.
Analyses from the price-level model suggest that there is an inverse correlation between patient mortality and the standardized price that dialysis facilities received. These data suggest that higher standardized prices tend to be associated with lower mortality rates. Results from the first-difference model, however, are less definitive. None of the correlation coefficients for the five primary disease groups (diabetes, hypertension, glomerulonephritis, other diseases, unknown) in the first-difference model is statistically significant. All five coefficients, however, consistently point in the same direction, i.e., that a larger decrease in standardized payment to a given