Differential Mortality in Medicare Advantage

Consumers increasingly choose among differentiated health insurance plans with varied networks and premiums. The consequences of such heterogeneity for consumer health outcomes are, however, unknown, as is whether consumers perceive and weigh these potential health impacts when selecting plans. Past research has shown that consumer demand can affect insurer incentives to innovate (Gaynor et. al. 2010, Cooper et. al. 2011) and contract with different providers (Ho 2009). Nevertheless, if consumers are unable to identify which plans make them healthier, or if they place little value on relative plan health effects, then insurers may have little incentive to invest in such quality improvements.
In this project we analyze whether some insurance plans reduce mortality more than others, what drives this heterogeneity, and whether consumers take into account relative health effects when selecting plans. Our setting is the Medicare Advantage (MA) market in which beneficiaries choose from a broad array of privately-managed and government-subsidized plans. The program is large and growing, currently covering nearly one in three Medicare beneficiaries, and its model of subsidized private provision is the forerunner of the Affordable Care Act health exchanges. Our study data consists of the entire universe of MA enrollees from 2006-2011.
Our analysis proceeds in three steps: first, we construct observational measures of plan health effects by comparing patient mortality across MA plans, adjusting for differences in observable characteristics of enrollees. Second, we use quasi-experimental variation arising from unexpected plan terminations to determine the relationship between these observational estimates and the causal effect on mortality of enrolling in different plans. Third, we analyze whether consumers take into account varying health impact when choosing among plans; and we estimate the implied value of statistical life (VSL) from choices in the MA market.
In a preliminary exploration of the data, we find large differences in plan-level mortality which appear to be driven by causal effects on health. The magnitude of these effects is enormous: moving beneficiaries from plans in the bottom quartile of health effects to top-quartile plans would prevent 18,000 deaths each year. Consumers slightly prefer plans with larger mortality added, valuing a 1 percentage point difference in annual mortality at around