Provisions of the ACA have the potential to both increase and decrease physician and hospital revenues, and the overall net impact is not yet clear. On one hand, the ACA expanded health insurance coverage to millions of previously uninsured individuals, which is expected to increase health care utilization, and therefore increase physician and hospital revenues from the nonelderly patient population. On the other hand, the ACA enacted reforms to Medicare payment policy, which are likely to decrease heath care utilization by Medicare patients, and may decrease physician and hospital revenues from Medicare. In addition, there is significant uncertainty in the magnitudes of the ACA’s impact on health care utilization by either elderly or nonelderly patients due to uncertainty in underlying behavioral parameters and responses. For example, when demand for care exceeds supply, providers may respond by increasing prices until demand and supply are equal, physicians may respond by trying to increase the number of patients per day they are able to see, or physicians may do nothing and waiting times may increase, which will effectively decrease patient demand for services. In this talk, we will use a sensitivity analysis approach to describe conditions, defined as sets of behavioral responses and parameters, under which we would expect overall revenues to increase, or decrease. In addition, we will use an analysis of historical data to determine which behavioral responses and parameters best explain past trends, and would be most likely to realistically predict future revenues.
To estimate overall revenue impacts under a range of different behavioral parameters and responses, we use RAND’s Health Care Payment and Delivery Simulation Model (PADSIM). PADSIM uses a blended equilibrium concept, in which the quantity of services provided reflects a compromise between patient demand and the quantity that providers prefer to supply. Providers’ preferred supply is, in turn, determined by the generosity of provider payment policy. To reconcile supply and demand, PADSIM uses the concept of “congestion,” which includes non-price factors that reduce patient demand but increase (or leave unchanged) provider supply.
To determine what behavioral responses and parameters best explain historical data, we will use heuristic optimization techniques including genetic algorithms and particle swarm optimization to fit the PADSIM model to historical data, minimizing the sum of squared errors. We will additionally use bootstrapping analysis to estimate confidence intervals around the estimated parameters. Using the PADSIM model to estimate relevant behavioral mechanisms and parameters gives us the ability to test various hypotheses, such as whether the ACA is increasing or decreasing total health care spending, and generate confidence intervals around those hypothesis.
