This project continues work estimating “returns to healthcare”: the relationship between healthcare inputs and health outcomes. With previous support, we made significant progress by developing a new empirical lens to compare similar patients treated at different hospitals. The main idea underlying the research program is that ambulance companies are effectively randomly assigned to patients when they are dispatched. We then discovered that ambulance company assignment matters for hospital choice – ambulance companies have preferences for hospitals where they usually take patients. Using this natural experiment, we found that emergency patients transferred by ambulance to hospitals that treat patients more aggressively have significantly lower mortality.
A feature of this empirical strategy is that its foundation is based on idiosyncratic preferences of over 2500 ambulance companies. As a result, we can run many “natural experiments” to answer related questions. Our first aim moves beyond inpatient spending to consider the entire profile of spending by patients after the (emergent) hospital admission. Preliminary results suggest that this allows us to resolve the difference
between our work and earlier cross-sectional evidence: large inefficiencies appear to arise at hospitals that rely on post-acute spending. In future work we can refine this result and explore the sources of productivity and inefficiency in healthcare spending. Our second aim is to unpack the bundle of characteristics that describe hospitals and how they are related to healthcare cost and outcomes. For example, we can use our strategy to provide some of the first causal evidence of the impacts of hospital volume, teaching status, and non-profit status. We can also assess the empirical validity of hospital quality measures that are taking an increasingly important role in efforts to transform the US healthcare system from one that pays for quantity to one that “pays for quality”. Our third aim focuses on the policy issues around ambulance transport itself, such as the optimal level of regionalization of care (trading off higher hospital quality against longer travel times and potentially hampered post-acute coordination). Our final aim turns our attention to physician characteristics within hospitals and their impacts on cost and health outcomes. This part of the project will employ a different approach: using variation in the attending physicians available to treat patients on the date of admission to focus on plausibly exogenous variation in physician characteristics due to scheduling variation. We will consider physicians’ treatment intensity, as well as qualifications such as specialty, board certification, and a particularly novel exploration of medical-licensing-exam scores obtained at the physician level.