# Saumya Chatrath

Researcher, Mathematica

Dissertation title:  "Essays on Health and Vulnerable Populations"

This dissertation examines the impact of barriers to accessing care on the medical utilization incurred by vulnerable groups of patients.

Paper 1: This paper measures the health effects of a large, exogenous increase in the plan deductible on patients with diabetes and high cholesterol. I find that an average exogenous increase in the individual deductible of $1,000 does not affect the health of the average patient with these conditions. However, when I divide patients on the basis of their underlying disease severity, I find that the deductible increase worsens the health of patients that were initially in control of their condition. Among diabetics with initial glycemic control, the deductible change causes a 2.3 percentage point, or a 19% increase relative to the mean in the probability of having a high HbA1c value (>7.5%). Among patients with stable baseline cholesterol, the deductible change causes a 0.5 percentage point, or 12.5%, increase relative to the mean in the probability of having a high cholesterol value (>160 mg/dl). I investigate the mechanisms that drive these results and find that it is driven by the differential impact of the deductible increase on primary care use, and in short-term reductions of high value prescription drugs. Paper 2: For decades private insurers have increased cost-sharing to curb moral hazard based on the evidence that it causes patients to reduce care without corresponding effects on health. While there is a large literature on own-price elasticities, very little evidence examines how cost-sharing in medical care affects the full range of services patients use. I examine this question using a large data set with several hundred thousand observations. Such a large sample size allows me to also examine price responses for low vs. high income patients, and patients with vs. without chronic diseases. I reach several conclusions. First, there is no statistical difference in the own-price elasticities between groups, but the difference is clinically meaningful. Second, I find that compared to other groups, low-income patients are disproportionately affected by the compounding effects of care disruptions. When PCP copays increase, low-income patients experience a stronger decline in the use of downstream services such as referrals to specialists, laboratory tests, physical therapy, and radiation services. Finally, I do not find evidence for short-term harms from cost sharing that one can pick up with claims data such as emergency department visits or inpatient stays. Paper 3: We investigate adverse selection among Special Enrollment Period (SEP) members in the Affordable Care Act Marketplaces. As compared to Open Enrollment Period (OEP) members, SEP members are younger and have approximately 34% higher average monthly costs, primarily driven by inpatient spending. We use a set of representative procedures to classify inpatient care as being predictable & discretionary, predictable & non-discretionary, and non-predictable & non-discretionary in order to inform an optimal policy response. Utilization was higher among SEP members across all three categories indicating that these members are more likely to have predictable health care needs but are also more likely to require non-discretionary, acute care. We also test the impact of a 2016 policy to reduce adverse selection among SEP members in the Marketplace. We find that the policy, which increased paperwork requirements for SEP enrollment, caused a differential reduction among SEP members in the rates of predictable inpatient care (by 6.8 per 10,000 members for predictable & discretionary care, and 15.7 per 10,000 members for predictable & non-discretionary care) but did not affect rates of non-predictable care. When compared to OEP members, the policy caused SEP members to increase monthly inpatient spending by$91.18 but had no effect on total costs.