Hospital responses to price changes: Evidence from the 2007 switch to MS-DRGs

This paper analyzes hospital responses to price shocks using the natural experiment generated by a 2007 change in the diagnosis-related group (DRG) system that Medicare uses to pay hospitals. The new system introduced both positive and negative price shocks with anticipated heterogeneity in effects across different types of hospitals. I use this policy change to study the comprehensive range of potential hospital responses (including upcoding, cost-shifting, and changes in patient care and admission volume). I also measure direct and spillover effects of the price changes and assess impacts over time and across different types of hospitals and markets.

Effective October 1, 2007, Medicare introduced the “severity-adjusted” MS-DRG patient classification system, the most significant change to DRGs since the system was originally implemented in 1983. The primary goal of the new system was to improve how patient severity is captured, allowing for better matching of payment levels to variations in severity and reduced disparity of payment for and treatment of cases within a DRG category. Industry estimates suggest that some hospitals may see total Medicare payment reductions of as much as 30% and others could see increases of up to and over 100%, with smaller, rural hospitals facing the largest payment reductions.

I use data from the Healthcare Cost and Utilization Project (HCUP)’s State Inpatient Discharge (SID) and Medicare Cost Reports to analyze the effects of switch to MS-DRGs in two ways. First, I use the pre-MS-DRG “DRG Grouper” to group patients who are admitted in the post-MS-DRG period into the DRG that they would have been entered as if they were admitted pre-policy change. In the absence of coding and other behavioral changes, this will give a whole-hospital picture of what the hospital might have looked like in the absence of the MS-DRG switch. I use this pseudo-counterfactual to compare treatment practices and intensity, as well as costs and patient outcomes under the DRG and MS-DRG systems. Second, I follow existing literature in using Medicare hospital cost report data to instrument for changes in Medicare prices and revenues at the hospital-level using price changes that were exogenously or “mechanically” imposed by the switch to MS-DRGs. The resulting predicted changes in prices and revenues are then used to evaluate cost-shifting as well as changes in hospital volume, treatment intensity, and patient outcomes. Outcome measures include number and types of procedures, length of stay, discharge status, and readmission rates.

SID data are linked to data from the American Hospital Association and Area Resource File to explore variations in hospital responses to price shocks by hospital characteristics, including size, scope, location, non-profit status, ownership (e.g., physician-owned), system affiliation, Medicare “bite,” payer mix, patient characteristics, and local market structure. In addition, I look at whether price changes alter practice patterns only for “eligible” (i.e., Medicare) patients and relatively more affected diagnosis (e.g., DRGs which faced the greatest price shocks) or whether there are spillover effects to other patients and service areas and hospital-level practices more broadly.