UC Berkeley

The first chapter, co-authored with Allyson B. Root and Benjamin R. Handel, studies how information and behavioral nudges impact patient behavior in end-of-life care. Despite the substantial economic and personal implications of end-of-life health care decisions, many fail to document their wishes or select a representative to make medical decisions on their behalf. Descriptive evidence suggests that this can result in sub-optimal outcomes including dissatisfaction and unnecessary medical spending, but it is not well understood why patients fail to engage in this high-value planning. We conduct an initial and subsequent intervention to facilitate advance directive (AD) completion in the patient population of our partner, Providence St. Joseph Health (PSJH). Using a randomized control trial, we find a significant 5 percentage point increase in AD completion with physical letter reminders tied to future primary care appointments, doubling the completion rate in the patient population from the start of the study. In addition, we find that including the physical AD form with paper letters as a nudge to decrease hassle costs increases AD completion 9 percentage points compared to no intervention. Our evidence also suggests that these interventions are more effective for older individuals, who are also less sensitive to the type of intervention. Back-of-the-envelope calculations suggest that it would cost $38 for every additional AD form completion using paper letters and included AD forms, compared to costless electronic reminders. However, we find no significant effects in AD completion from the initial intervention involving in-person AD drives and electronic videos.

In the second chapter, co-authored with Benjamin R. Handel, Lynn M. Hua, and Yuki Ito, we study health plan choice and health plan menu design with 13 years (2008-2020) of health claims and health plan choice data from the California Public Employees’ Retirement System (CalPERS). We develop a choice model that predicts the number and type of subscribers moving across plans under different plan environments, as a function of (i) plan premiums, (ii) plan cost-sharing, and (iii) plan brand. We find that (i) subscribers overweight premiums relative to out-of-pocket spending by a factor of roughly five to one, in line with other literature, and (ii) subscribers place meaningfully differential values on plan brands. We find that these preferences, especially for plan brand, depend crucially on whether a subscriber's family is sick or healthy. We also find some evidence of risk adjustment blunting adverse selection in our market while it was implemented from 2014-2018, although due to inertia the effects were relatively mild. Finally, we present counterfactual scenarios for future enrollment with and without risk adjustment and inertia.