Who benefits from health signals? Estimating marginal returns to medical care
Abstract
Getting information about your health has become easier through wearable devices, routine checkups, and wellness visits. Such health signals are intended to increase the use of preventive care, improve people's health, and reduce medical costs. However, consumers (i.e., patients) with limited medical knowledge are delegated the decision to actually see a doctor, leading to a self-selection problem: Do those who are most responsive to such signals benefit the most?We investigate the marginal returns to health signals that people receive after routine health check-ups. We use administrative data on mandatory check-ups and healthcare utilization from the largest health insurers in Japan, covering one-third of all Japanese (about 40 million people). We exploit the fact that individuals who are just above a clinical threshold of fasting blood glucose (FBS) receive health warnings about diabetes (DM) to see a doctor in a regression discontinuity design.
We have five main findings. First, we find strong evidence that crossing the threshold (FBS of 126) increases healthcare utilization, as measured by DM-related physician visits and outpatient expenditures. Second, although the magnitude is relatively small, we find that the additional care improves several biomarkers, including FBS, BMI, and blood pressure. Third, we show that those who "respond" the most to health signals and visit doctors are not the ones who "benefit" the most, suggesting reverse selection on gains based on observables. Fourth, we estimate marginal treatment effects that account for unobserved heterogeneity and document reverse selection on benefits beyond the compliers. Finally, the policy simulation of targeting health signals by observables does not successfully induce sick patients, who are reluctant to go to the doctor but could benefit greatly from doing so, to see a doctor, suggesting that reliance on self-selection into treatment may be limited.