Yesterday, I reviewed one paper claiming PPS along with competition reduced cost, but these cost reductions occurred mostly for the most expensive (read most sick) patients. Continuing the PPS theme, today we will look at a 1995 article in Econometrica in which David Cutler attempts to measure “The Incidence of Adverse Medical Outcomes under Prospective Payment.” Under PPS, hospitals no longer received any revenue from preforming the marginal service, or equivalently hospitals bear the full cost of marginal treatments. Also, average cost figures for each procedure changed depending on how the DRG was valued. The price a hospital received for a procedure is:
The DRG value of a certain diagnosis ‘d‘ is given by variable ‘WGT‘ which is adjusted by a hospital specific factor ‘P_h‘ which reflects the local wage conditions. Other adjustment factors are the indirect medical education costs (‘IME‘) and a federal subsidy to hospitals which provide a disproportionate share of their services to the poor (‘DSH’ – see Baicker and Staigler post). Cutler hypothesizes that the effect hospitals bearing the cost of marginal treatment will be increased mortality. Also, any decrease in average price for a treatment should result in increased mortality. Cutler also investigates re-admission rates as well.
Using 1981-1988 data from Medicare and Social Security records, first preforms a difference in difference estimate using Massachusetts (which adopted a PPS in 1986) as a control for the federal PPS (which was adopted in 1984). He finds that mortality increased after the PPS was passed and readmission probability also increased. To refine the analysis, Cutler then uses a maximum likelihood estimation on a logistic hazard function to estimate the impact of PPS into 2 components: 1) the effect of eliminating revenue for the marginal treatment and 2) the impact of any changes in average cost for the procedure.
The impact of the elimination of marginal reimbursement led to a 25% decline in mortality, and a mild increase in the readmission probability. On the other hand, a one standard deviation decrease in average payment for a procedure led to an increase in mortality of 0.5%. The average price decrease also led to a lower re-admission rate.
How can we explain these findings? The elimination of marginal reimbursement may have decreased the use of unnecessary procedures. A more likely explanation is that of DRG creep where doctors code mildly sick patients as having severe diseases in order to increase their compensation. The result of DRG creep is that the patient pool in each DRG group is healthier after PPS than before it simply due to these classification changes and not due to any improvements in the medical care received. Since hospitals do not get paid for marginal services, they may wish to readmit their patients in order to double-bill Medicare.
The finding that an increase in average cost of a procedure reduced mortality is intuitive since paying hospitals more money to take care of certain illness will likely lead to better care for those illnesses. Readmissions are less likely as well, since increasing the profitability of a procedure would reduce the probability of a hospital needing to use double billing.
The true value of this paper is performing clever econometric specifications using MLE estimators. A discrete set of mass points [as suggested by Heckman and Singer (1984)] is used to account for unobserved heterogeneity for each individual. As Cutler admits, it is difficult to judge if many of the health impacts of PPS are due to true changes in the level of medical care or simply due to accounting changes such as DRG creep, so one should interpret these findings with caution.
Cutler (1995), “The Incidence of Adverse Medical Outcomes under Prospective Payment,” Econometrica, Vol. 63, No. 1. (Jan., 1995), pp. 29-50.