Health Expenditures and Health Outcomes: A Longitudinal Analysis in the Medicare Population

Are we getting our money’s worth for the marginal health care dollar?  There is little consensus to this question, with estimates that range from a strong yes to a resounding no. In this paper, we argue that medical inputs exhibit differential marginal effects, meaning that standard regressions in which expenditures “explain” outcomes are uninformative – it’s about how the money is spent, and not how much. Following Chandra and Skinner (2012), we first categorize healthcare inputs into three groups: (i) highly effective, (ii) with heterogeneous benefits, or (iii) costly but with little known health benefit.  We then test the model in a 100% sample of heart attack (acute myocardial infarction) patients in the Medicare fee-for-service population during 2007-08. We first show that estimates of the association between expenditures and outcomes are highly sensitive to specification, and depend on how the health outcome is measured, whether one uses an instrument for expenditures, the risk-adjustment approach, and limiting the sample to “tourists.” By contrast, when we estimate our theoretically appropriate model, the results are quite stable.  Briefly, we find (i) large gains in heart attack survival for effective treatments such as aspirin, beta blockers, and reperfusion, (ii) heterogeneous benefits for “Category II” treatments such as stents, and (iii) little observable benefits for the “Category III” such as the frequency of scans (including the practice of “double scans,” recognized to be harmful to patients), ICU/CCU use, and home health care utilization.  Conversely, overall risk-adjusted expenditures are negatively associated with effective care and positively associated with Categories II and III.