Examining selection in Affordable Care Act (ACA) Marketplaces: special enrollment periods

TL;DR

People who join ACA health plans through special enrollment periods spend more on healthcare, suggesting a selection bias that requires careful policy solutions.

Contribution

The study identifies adverse selection in ACA special enrollment periods and highlights the need for nuanced policy interventions.

Findings

SEP enrollees had higher inpatient spending compared to OEP enrollees.

SEP members showed higher utilization rates for predictable, discretionary, and nonpredictable care.

A 2016 policy change did not reduce utilization rates among SEP members.

Abstract

Affordable Care Act (ACA) Marketplace members who enroll through a special enrollment period (SEP) have significantly higher average monthly spending than members who enroll through the annual open enrollment period (OEP), driven primarily by higher inpatient spending.

Using data from a large national insurer that participated in the federal ACA Marketplace from 2015 to 2016 in 24 US states, we examined differences between SEP and OEP Marketplace enrollees in time from enrollment to inpatient use of predictable and discretionary care (ie, hip and knee replacement), predictable and nondiscretionary care (ie, childbirth), and nonpredictable and nondiscretionary care (ie, acute myocardial infarction and stroke). We examined whether a 2016 policy that increased SEP eligibility verification requirements was associated with changes in utilization.

When compared with OEP Marketplace members, SEP members had significantly higher rates of care in all 3 categories. The 2016 policy was not associated with changes in utilization rates.

Our results provide evidence that there is adverse selection in the SEP of the ACA Marketplaces. However, since SEP members were more likely to seek care for services that are predictable and nonpredictable, and discretionary and nondiscretionary, the optimal policy response to reduce adverse selection needs to be nuanced and multipronged.