Essays on hospital technology adoption and health insurance

Local hospital systems, in which multiple hospitals within the same geographic market share common ownership, have been growing in the US in recent years. The first essay examines the impact of system membership on the adoption of MRI technology. I build up a structural model to incorporate market competition as well as collaboration among local hospital members. System hospitals without local members are found to behave similarly to independent hospitals and both are associated with higher MRI adoption rates. System hospitals with local members have lower adoption rates and this effect is stronger the closer they are located to local members who have already adopted the technology. This suggests that local but not remote system hospitals might exploit economies of scale by reducing duplicate equipment, and that local hospital acquisitions can potentially reduce costs.;The second essay examines the optimal health insurance, considering both multiple treatment goods and multiple time periods. Multiple goods and multiple time periods raise issues of both complementarities and of correlated losses. Substitutes and positively correlated demands over goods or over time are shown to reduce optimal cost shares on treatment. In a multi-period model, savings complicate the optimal insurance rules but positively serially correlated errors generally imply improved coverage is desirable. Further, the presence of positively correlated uncompensated costs also leads to reducing the cost sharing on the covered services. Using insurance claims data, we examine the empirical relevance of the contemporaneous correlations across goods and correlations over time, using three broad aggregates of health treatment spending: inpatient, outpatient, and pharmaceuticals.;The final essay studies the patterns of service-level spending in health plans. Using disaggregated commercial insurance claims we show that selection incentives measured by our selection indices are highly similar for four types of managed and unmanaged health plans. However, managed care plans spend proportionally less on services that are predicted to be more profitable to tightly ration. Conventional diagnosis-based risk adjusted premiums reduce selection incentives by about 50%. The incentives to ration care by health spending decomposed by type of service are stronger than incentives by provider specialty or place of service.