New frameworks for hospital quality and value profiling

The rapid expansion of payment models that incorporate information on individual providers' quality and efficiency of care has brought increasing scrutiny to performance measurement in health care. Summary measures of quality and value are integral components of pay-for-performance programs, but have also been used for setting patients' copayment levels, for regulation and accreditation purposes, and for studying the correlates and consequences of poor quality of care. Yet, by most accounts, these summary measures remain underdeveloped. In this dissertation, I describe new frameworks for estimating and comparing hospitals on summary measures of quality and value. In the first chapter I introduce a method to profile hospitals on the "value" of care for acute myocardial infarction. Using estimates of society's willingness to pay for survival gains to express each hospital's survival benefit in dollars, I combine in-hospital survival and cost outcomes on the incremental net monetary benefit scale to compare the performance of 69 hospitals. While common in cost-effectiveness analyses of new technologies, this approach has not been applied to provider profiling. In the second chapter, I propose and compare three value estimators that make different assumptions about the relationship between cost and quality. The estimators include one based on a regression framework, the cost-effectiveness estimator from Chapter 1, and the prevailing technique used by health plans, which compares performance on both domains independently. I find significant differences in the set of hospitals classified as high value across methods, and recommend against using the most common approach. In the third chapter I present a framework for profiling hospitals using a summary measure of health status following coronary artery bypass graft surgery, quality-adjusted life years (QALYs). I use eleven performance measures to predict two-year survival outcomes, and derive composite health utility estimates based on a patient's exposure to six potential surgical complications, which are used to weight survival time. I then assess differences in average quality-adjusted life expectancy over two years of follow up among 14 Massachusetts' hospitals. These frameworks provide general methods that are applicable to other conditions, and that can also serve as the basis for more detailed models.