Understanding Consumer Experiences and Insurance Outcomes Following Plan Disenrollment in the Nongroup Insurance Market

Abstract

Disenrollment from health plans purchased on Affordable Care Act (ACA) Marketplaces is frequent; little is known whether disenrollment from off-Marketplace plans is as common or about the experiences and consequences of disenrollment. Using longitudinal administrative data on 2017–2018 nongroup plan enrollment linked with survey data, we analyze plan disenrollment in one regional insurance carrier servicing three states. Overall, 71% of enrollees disenrolled from their 2017 plan. Disenrollment was associated with purchasing through an ACA Marketplace, the carrier making significant changes to an enrollee’s plan benefit design, being healthier, being younger, and paying a higher premium for their 2017 plan in 2018. Experiencing financial burden or poor access to preferred providers was not associated with disenrollment. Most disenrollees (93.2%) enrolled in other coverage, often at a lower premium, but lacked confidence that they could afford needed care. These results can inform policy to support enrollees through coverage transitions and foster stability in the nongroup market.

Introduction

Individuals who purchase health insurance for themselves and their families in the nongroup market have a wide array of plan choices. As of 2014, the Patient Protection and Affordable Care Act (ACA) created online health insurance Marketplaces where enrollees can shop among and enroll in plans from different carriers, and where enrollees who meet income and other eligibility requirements can receive premium and cost-sharing subsidies. Consumers also have the option to enroll in nongroup plans directly with insurance carriers (e.g., “off-Marketplace”). There are a few important differences between Marketplace and off-Marketplace choices. The off-Marketplace market includes more plan choices because consumers can enroll in Marketplace plans directly from the insurance carrier along with other plans not available on Marketplace, including some that do not meet ACA regulations because they have been “grandfathered” from the pre-ACA period (although enrollment in grandfathered plans is much more common in the group market than in the nongroup market (Keith, 2020)). In addition, subsidies are not available for off-Marketplace plans. For both Marketplace and off-Marketplace plans, individuals can shop for, enroll in, or switch plans once a year during open enrollment.

The aim of this market design is to harness the advantages of choice and competition in order to improve coverage, satisfaction, and affordability of nongroup market plans (Frank, 2019). If individuals consider plan price and quality to shop and choose among plans, then decisions to switch among plans can make individuals better off. Moreover, such consumer behavior creates incentives for insurance carriers to improve the quality of plans or to lower prices in order to attract enrollees. However, consumers in health insurance markets do not always behave in this way, with consequences for consumer welfare. In Medicare, enrollees have been found to overweight the premium paid for prescription drug plans relative to other cost sharing; as a result, many enrollees do not choose the plan that minimizes their costs (Abaluck & Gruber, 2011). In employer-sponsored insurance markets, consumers have been found to stay enrolled in the same plans over time even as different alternatives where they would be better off become available; this phenomenon is known as status quo bias and is driven in part by inertia (Samuelson & Zeckhauser, 1988; Sinaiko & Hirth, 2011). Thus, while staying in one’s plan over time could reflect satisfaction with one’s choice, it could also be due to consumer inattention and status quo bias. The latter can lead enrollees to miss out on plans that would better meet their needs and preferences and can interfere with competition (Afendulis et al., 2015; Ho et al., 2017; Polyakova, 2016; Samuelson & Zeckhauser, 1988).

Understanding the dynamics of plan disenrollment in the nongroup market can also inform the need for insurance regulations. For example, plan disenrollment could be driven by adverse selection (e.g., when higher risk enrollees switch to plans with more generous coverage) or adverse retention (e.g., higher risk enrollees are more likely to stay in more generous plans over time). Both adverse selection and adverse retention have created market instability and inefficient insurance outcomes in U.S. insurance markets (Altman et al., 1998; Cutler & Zeckhauser, 2000), and are a rationale for policy intervention such as risk adjustment (Meyers et al., 2019; Rahman et al., 2015; Riley, 2012).

Post–ACA evidence on plan disenrollment over the period from 2014 to 2018 suggests that plan disenrollment in ACA Marketplaces is common, ranging from 23% to 50% of enrollees per year (Crespin & DeLeire, 2019; Gordon et al., 2019; Rasmussen et al., 2019; Roberts et al., 2020). Reports of frequent plan disenrollment in the ACA marketplaces are difficult to interpret without evidence of disenrollment among off-Marketplace enrollees and without an understanding of enrollee experience during these plan changes or of insurance outcomes after plan disenrollment. We know of no evidence of levels of plan disenrollment from off-Marketplace plans since the ACA. Prior to the ACA in 2008–2011—when the entire nongroup market was off-Marketplace—52% of all enrollees experienced disenrollment over the course of 12 months (Sommers, 2014). In this article, we use longitudinal administrative data on 2017–2018 nongroup plan enrollment linked with survey data to analyze plan disenrollment in one regional insurance carrier servicing three states. We compare plan disenrollment among ACA marketplace with that among individuals who purchase their plans off-Marketplace, and present evidence on the experiences associated with disenrollment and on insurance outcomes after disenrollment.

New Contribution

Existing literature finds that carrier decisions to exit an ACA marketplace completely and major changes in plan premiums (e.g., the increased Silver Plan premiums imposed in 2018 to compensate for the withdrawal of federal funding for cost-sharing subsidies, also known as “silver loading”) are associated with plan disenrollment (Crespin & DeLeire, 2019; Rasmussen et al., 2019; Roberts et al., 2020). There is little understanding of whether enrollees who have negative cost or access experiences in their health plan disenroll and change plans, whether plan disenrollment is associated with becoming uninsured, and whether this varies for enrollees on the ACA Marketplaces versus off-Marketplace. Yet this evidence is needed to inform policy in order to support enrollees through coverage transitions and foster long-term stability in the nongroup market.

The dynamics underlying nongroup plan disenrollment remain unknown in part because the existing evidence stems largely from administrative enrollment data from ACA Marketplaces, which neither include data on off-Marketplace plans nor have information about enrollee health or plan experiences. Analysts also cannot observe downstream insurance outcomes if enrollees exit a Marketplace. To bring new evidence to bear on these dynamics, we analyze survey data on nongroup market enrollees’ experiences in plans, their health status, and their decisions regarding plan disenrollment along with administrative enrollment data. We sought to examine whether plan disenrollment was suggestive of adverse selection or adverse retention, of consumer inattention, or of consumers shopping and choosing higher quality or lower cost plans. We also sought to understand whether plan disenrollment was associated with uninsurance (as had often been the case prior to the ACA (Sommers, 2014)) or whether enrollees who experience plan disenrollment find more affordable plans.

Conceptual Model

Our conceptual framework is based on an economic model of consumer choice of health insurance, which posits that when choosing a health plan, families will consider their own characteristics, such as health status, expected utilization, and income, along with plan characteristics such as benefit design and plan quality, and choose the health plan that maximizes the expected utility for the family. For enrollees in the nongroup market, this choice involves staying in one’s current health plan, if still available, or switching to one of many other plan options. In this model, changes in one’s plan benefit design, in the cost of the plan, and changes in an enrollees’ own circumstances (e.g., income, health care needs) are predicted to have an impact on plan choice. In addition, enrollee satisfaction with their prior experience accessing, receiving, and paying for care also likely affect the decision to reenroll or switch plans each year. Thus, we hypothesize that the enrollees who have negative experiences in their health plan would be more likely to disenroll than would enrollees who have mostly positive experiences.

A few additional factors are important in the context of this framework. First, health insurance benefits are often confusing, and having access to usable information and tools to help understand plan options or plan changes is vital if consumers are to be able to make efficient choices. There are likely differences in the availability of these resources for enrollees who purchased their health plan on Marketplace versus off-Marketplace. To the extent that Marketplaces communicate with enrollees about open enrollment and plan options so as to foster shopping among plans, we expect that enrollees in Marketplace plans would be more likely to disenroll from their plans than would off-Marketplace enrollees. Second, previous research, described in more detail above, reveals that inertia, due to factors such as loss aversion and regret, is important to health insurance decisions and can lead enrollees to stay in their health plans over time despite better options being available. Whether there is substantial inertia in the nongroup market and, in particular, whether inertia is greater among enrollees who have been in their plans longer relative to those in their plans for only 1 year are not well-understood.

Data and Method

Study Setting and Sample

The study was set in a large, nonprofit regional insurance carrier in three states, Massachusetts, New Hampshire, and Maine, during 2017–2018. The number of insurance carriers in these states’ Marketplaces in 2017 ranged from nine in Massachusetts to three in Maine, decreasing by one or two in each state in 2018. The number of health plans offered by these carriers decreased substantially between 2017 and 2018, from 418 to 181 in Maine, from 2,463 to 858 in Massachusetts, and from 100 to 59 in New Hampshire. For the carrier that we study, the number of plans offered by the carrier increased from 29 to 35 in Massachusetts, stayed the same at 20 in Maine, and went down from 19 to 8 in New Hampshire (see Supplemental Appendix available online for additional demographics and non–group market characteristics in these states).

We selected a stratified random sample of subscribers aged 18 to 64 years enrolled in a non–group plan as of 2017 open enrollment and invited them to participate in a longitudinal survey; the sample was stratified by whether the enrollee purchased the plan through the ACA Marketplaces or directly from the carrier in each of the three states. The carrier represented 32% of the enrollees in ACA-compliant plans in the nongroup markets in New Hampshire, and 24% in Maine, in 2017 (The Kaiser Family Foundation, 2017). The sample in Massachusetts did not include enrollees in ConnectorCare, the state’s basic health plan for households with incomes below 300% of the federal poverty level (FPL), because the carrier does not participate in that program. The carrier represented 10% of unsubsidized non–group enrollees (e.g., non-ConnectorCare enrollees) in Massachusetts (Center for Health Information and Analysis, 2019).

Study Design and Data

We first fielded a baseline survey after 2017 open enrollment that collected data about enrollee demographic and socioeconomic characteristics; additional description of methods and findings have been published previously (Hero et al., 2019). Twelve months later, we fielded a follow-up survey to baseline survey respondents to collect data about experiences in their 2017 plan, health experiences during 2017, and plan selection for 2018. The survey asked about health care affordability, experiencing higher out-of-pocket costs than expected, access to desired physicians and hospitals, and how the premium of the plan in which they enrolled in 2018 compared with that in 2017. Among enrollees who left their 2017 plan, the survey asked why they disenrolled and about the type of health insurance they had in 2018. Both surveys were fielded by e-mail and mail, and we offered a $20 gift card as an incentive for each survey completion. The Harvard Pilgrim Health Care Institutional Review Board at Harvard Pilgrim Health Care approved this study (IRB Reference #912315), and the completion of the survey was an indication of participant consent. The baseline sample includes 2,029 enrollees (29% response rate), 60% of whom (n = 1,221) responded to the follow-up survey.

We obtained 2017–2018 administrative enrollment data from the health insurance carrier. These data included enrollee characteristics (i.e., age, sex, family size, state), 2017 plan characteristics (e.g., metal level, whether purchased through ACA Marketplace, cost-sharing requirements, narrow vs. broad network), and the enrollee’s paid premium net of income subsidy (subsidy was only relevant for Marketplace enrollees). Data also included which 2017 plans were “renewed” versus rolled into a “new” plan for 2018. More specifically, while most nongroup plans experience at least some changes in benefit design from year to year, if changes are minor then they are considered within-plan changes and the plan is “renewed.” If changes are more significant, with significance determined by state and Federal insurance regulations, then the 2017 plan is considered terminated and the 2018 benefits are offered as a “new” plan. When a plan is not renewed but is rolled into a “new” plan, enrollees are notified by mail that if they take no action during open enrollment then they will be auto-enrolled in this “new” plan for 2018.

Data also included the carrier’s state-specific average premium trend for 2018, and for each enrollee, and whether the enrollee disenrolled from their 2017 plan.

Method

Our main outcome of interest was whether the enrollee left their 2017 plan, which we term plan disenrollment. We measured plan disenrollment using administrative data in two ways. First, if the enrollee’s 2017 plan was renewed in 2018, then we defined plan disenrollment as the enrollee not staying in this plan for 2018. Second, if the enrollee’s 2017 plan was transitioned to a “new” plan, then we defined plan disenrollment as the enrollee not staying with the “new” plan for 2018.

Characteristics Associated With Plan Disenrollment

We linked baseline survey data to the administrative data and examined the association of plan disenrollment with purchasing on the ACA Marketplace, with having one’s plan transition to a “new” plan (as opposed to renewal), the 2017 plan characteristics described above, and estimated 2018 premium in the 2017 plan. For plan characteristics, our main specification includes the plan’s metal level as a summary measure of comprehensiveness of plan coverage. In sensitivity analyses, we substituted metal level with measures of plan deductible and of out-of-pocket maximum; results were similar (see Supplemental Appendix available online).

We estimate the enrollee’s 2018 plan premium in two steps. First, we assumed that enrollees’ income stayed constant from 2017 to 2018, and we used information on the amount of premium subsidy in 2017 to estimate subsidy in 2018. More specifically, for enrollees who received a premium subsidy in 2017, we use the amount of that subsidy and the amount of the second lowest cost silver premium for a similar family in the enrollee’s rating area to estimate the enrollee’s family premium contribution in 2017. We then use this family contribution amount and the second lowest cost silver premium for a similar family in the enrollees’ rating area in 2018 to estimate premium subsidy in 2018. This 2018 subsidy estimate thus accounts for silver loading. Second, we applied the carrier- and state-specific premium growth rate for 2018 to estimate 2018 plan premium. If enrollees purchased their plan off-Marketplace or did not get a 2017 subsidy, this is their estimated 2018 plan premium. For enrollees who received a 2017 subsidy, we subtracted their 2018 subsidy estimate from this 2018 plan premium to estimate their 2018 premium net of subsidy.

We also examined the association of disenrollment with characteristics of enrollees measured in the baseline survey, including household income as a percentage of the FPL (i.e., less than 250% FPL, 250% to 400% FPL, more than 400% FPL), health status (whether the enrollee or a family member on their plan had a chronic condition), race (White vs. non-White), health insurance literacy, and whether the enrollee had been in their 2017 plan in the prior year (e.g., 2016). These household income categories were measured using two survey items that asked about household income and household size, and categories were chosen so they align with the income-eligibility categories for cost-sharing and premium subsidies in Marketplace plans. Following methods used in the Health Reform Monitoring Survey, enrollees were measured to have high health insurance literacy if they were “very confident” or “somewhat confident” with seven health insurance terms and with six health insurance–related activities (e.g., figuring out if a doctor was included in your plan’s network; Holahan & Long, n.d.; Long et al., 2014).

We examined associations between disenrollment, individual/family characteristics, and plan characteristics using cross-tabulations; we assessed statistical significance using Pearson’s chi-square test (see Supplemental Appendix available online for results), and all variables significant at p = .20 or lower were included in logistic regression models. To assess whether transitioning to a “new” plan had a differential effect on disenrollment for enrollees on the ACA Marketplaces versus off-Marketplace, models included an interaction between these variables. Premium varies by number insured on a plan, so we controlled for family size, excluding families larger than five because of small numbers of these families in the sample. All analyses adjusted for poststratification weighting for sampling and nonresponse using methods published previously (Hero et al., 2019). In sensitivity analyses, we measured health status using self-reported health status instead of the presence of a chronic condition. Results were not significantly different from those of the main models (see Supplemental Appendix available online).

Plan Experiences and Disenrollment

We report weighted frequencies of enrollee reasons for plan disenrollment collected in the follow-up survey. All analysis of follow-up survey data used inverse probability weights constructed for the follow-up survey sample to account for sampling design and subject-level nonresponse (for details, see Supplemental Appendix available online).

We measured the relationship between having a negative plan experience during 2017 and plan disenrollment in cross-tabulations; negative plan experiences included higher out-of-pocket costs than expected, any financial burden, not being able to see a desired health care provider, or rating the plan as fair or poor. Financial burden was measured as having trouble paying medical bills; having difficulty paying for food, heat, or rent due to medical costs; or forgone care due to cost. We measured the relationship between having an unexpected health event or change in health status during 2017 and plan disenrollment in cross-tabulations. To assess whether reenrollment was associated with consumers actively choosing to stay in their plan versus inattention and inertia, we examined whether reenrollees said they considered other plans before choosing to reenroll. We examined this response overall, stratified by Marketplace enrollment, and stratified by length of time enrolled in the 2017 plan (2017 only vs. more than 1 year). We assessed statistical significance using Pearson’s chi-square test; we adjusted for multiple comparisons using Bonferroni’s correction, and p < .01 was considered statistically significant.

Insurance Outcomes

Enrollees’ insurance outcomes in 2018 were collected in the follow-up survey. For those who experienced plan disenrollment, we reported on their source of insurance coverage for 2018 (if any). We report the association of plan disenrollment with (1) enrollees’ survey responses as to whether their premium in their chosen plan for 2018 was higher, lower, or equivalent to that paid in 2017, using multinomial logistic regression and (2) whether they feel confident that they could afford their health care expenses should they become sick in 2018, using logistic regression. These models controlled for enrollee age, household income, family size, metal tier of 2017 plan, state, and whether a family member had a chronic condition. Results are presented as predicted probabilities from these models.

Results

Plan and Enrollee Characteristics Associated With Plan Disenrollment

There were 2,029 unique respondents to the survey across Massachusetts, Maine, and New Hampshire; characteristics of the study sample and their 2017 plan choices are shown in Table 1. A higher proportion of off-Marketplace enrollees had a 2017 plan that was not renewed and instead was transitioned to a “new” plan due to significant benefit design changes made by the carrier than did on-Marketplace enrollees.

Table 1.

Sample Characteristics (Weighted).

Individual characteristics	%
Male	51.7
Non-White	6.6
Age-group, years
18–35	27.6
36–45	15.7
46–55	22.8
56–65	33.9
Family on plan	36.9
Health measures
Enrolleea on plan with chronic condition	33.4
Enrolleea on plan in fair/poor health	11.5
High health insurance literacy	22.6
Education level
High school graduate or less	16.4
Some college	25.4
4-year college or more	58.2
Income level
≤250% FPL	41.1
250% to 400% FPL	20.3
400%+ FPL	38.6
Massachusetts	29.3
Maine	31.5
New Hampshire	39.2
2017 plan characteristics
Enrolled on Marketplace	70.3
Transitioned to “new” plan for 2018 (due to significant benefit design changes)
On-Marketplace enrollees	27.8
Off-Marketplace enrollees	37.2
Bronze plan	37.5
Silver plan	46.6
Gold plan	15.9
In same plan as in 2016–2017b	32.6
Observations	2,024

Note. Source. Authors’ analysis of weighted survey data. Respondents were considered health literate if they described themselves as “somewhat” or “very” confident in how well they understood seven common health insurance terms and six health insurance activities. FPL = federal poverty level.

^aEnrollee indicates either self or family member who is dependent on subscriber’s health plan.

^bFrequency of enrollees who responded “The same plan I have now” to the survey question, “What kind of health insurance did you have just before the most recent enrollment period (e.g., 2017 OE)?”

Overall, 71% of nongroup enrollees disenrolled from their 2017 plan. Among ACA Marketplace enrollees, 60.9% disenrolled from the carrier, 16.4% switched to a new plan from the same carrier, and 22.8% reenrolled in their 2017 plan. Among off-Marketplace enrollees, 47.2% disenrolled from the carrier, 8.2% switched to a new plan from the same carrier, and 44.5% reenrolled in their 2017 plan. Thus, plan disenrollment among enrollees who purchased their plan through the ACA Marketplaces was higher than for those who purchased off-Marketplace (77.3% vs. 55.5%, p < .001).

Results from adjusted analyses show that Marketplace enrollees were 27.3 percentage points more likely than off-Marketplace enrollees to disenroll from their plan (95% confidence interval [CI] [21.2, 33.3]), a 54.8% relative difference (Table 2). Enrollees in plans with more significant benefit design changes (i.e., plans that were transitioned to “new” plans by the carrier) were 17.0 percentage points (95% CI: [11.8, 22.4]) more likely than other enrollees to disenroll from their plan, a 26.0% relative difference. Enrollees who had been enrolled in their plan for more than 1 year were 14.1 percentage points (95% CI: [−19.6, −8.6]) less likely to experience plan disenrollment for 2018 (an 18.9% relative effect). The magnitudes of these three associations are larger than those of all other plan and enrollee characteristics.

Table 2.

Factors Associated With Plan Disenrollment Between 2017 and 2018 (Logistic Regression Results, Marginal Effects).

	Change in predicted probability of disenrollment (pp)	Relative difference versus reference group (%)
Enrollee or dependent has chronic condition (reference no chronic conditions)	−5.67*	7.9
Purchased plan on Marketplace (reference purchased off-Marketplace)	27.28***	54.8
Benefit design changes so plan became “new” plan (reference 2017 plan was renewed for 2018)	17.01***	26.0
Age, years (reference age 19–35 years)
36–45	−2.58	3.4
46–55	–11.15**	14.6
56–65	–10.52**	13.8
Household income (reference less than 250% FPL)
251% to 400% FPL	4.37	6.5
400% FPL or higher	4.19	6.2
Chose 2017 plan in prior enrollment yeara (reference in different plan in prior year)	−14.09***	18.9
Metal tier
Silver	−7.36*	10.0
Gold (reference bronze)	−13.57**	18.1
Plan had narrow provider network (reference plan had broad network)	8.13	11.9
Estimated premium for 2017 plan in 2018 ($100 increase)	0.01***	1.6
N	1,548

Note. Source. Authors’ analysis of weighted survey data. Results based on logistic regression models controlling for all variables listed in the table, an interaction between marketplace and nonrenewed plan, family size, and state. In cross-tabulations, we found no statistically significant difference in plan disenrollment by gender, race, health insurance literacy, or education level. See Supplemental Appendix available online for model results and from marginal effects including 95% confidence intervals. pp = percentage points; FPL = federal poverty level.

^aSurvey respondents who indicated the plan they chose for 2017 was the same plan as they were enrolled in for the prior (i.e., 2016) plan year.

^*p < 0.05.

^**p < 0.01.

^***p < 0.001.

Having a chronic condition in the family was associated with lower probability of plan disenrollment; plan disenrollment also decreased with age of enrollee. Greater increase in estimated 2018 premium was associated with increased plan disenrollment, though small in magnitude. Enrolling in a Gold plan was associated with lower probability of plan disenrollment.

Plan Experiences and Disenrollment

When asked why they changed plans, enrollees, including the majority of off-Marketplace enrollees, were most likely to report that it was because their premium had increased (Table 3). Approximately one third of those on the ACA Marketplaces reported that it was because they found a new plan that was more affordable. One third of enrollees who disenrolled from their plan but stayed with the same carrier reported that significant benefit design changes (e.g., their plan’s termination with auto-enrollment into a “new” plan) was an important reason for their switch. Only 1.5% of ACA Marketplace enrollees and 0.9% of off-Marketplace enrollees disenrolled from their plans because there would be no penalty for being uninsured.

Table 3.

Enrollee Reported Reasons for 2017 Plan Disenrollment.

	Disenrollees who left insurance carrier (%)		Disenrollees who switched plans within insurance carrier (%)
Reason why left plan (check all that apply)	On-Marketplace (n = 393)	Off-Marketplace (n = 283)	On-Marketplace (n = 105)	Off-Marketplace (n = 69)
Premium in 2017 plan increased for 2018	36.1	50.5a	45.4	55.5
Found more affordable plan	36.9	37.2	34.0	23.8
2017 plan not renewed	13.2	20	29.7	37.6
Got public or employer-sponsored coverage	28.2	18.2	b	b
Wanted better benefits/network	12.3	7.7	17.7	9.9
Moved	2.9	6.2	b	b
Subsidy eligibility changed	7.9	5.0	19.1	2.9
Thought no penalty if uninsured	1.5	0.9	b	b

Note. Source. Authors’ analysis of follow-up survey; N = 676 respondents who disenrolled from carrier; N = 174 respondents switched plans within carrier. For enrollees whose 2017 plan was transitioned to a “new” plan, we considered them to have disenrolled if they did not stay enrolled in the “new” plan in 2018.

^aDifference between on and off-Marketplace significant at the 1% level.

^bNot asked of respondents who switched plans within carrier.

Enrollees who experienced negative cost or access experiences in the plan, including any financial burden, higher out-of-pocket costs than expected, inability to see their desired physician, or fair/poor experience using the plan, and enrollees who had an unexpected health event in 2017 were not more likely than others to disenroll from their plans (see Supplemental Appendix Table 8 available online for results).

Half (50.4%) of those who reenrolled in their plans for 2018 reported that they considered other plans before choosing to stay in their 2017 plan. A higher proportion of ACA Marketplace enrollees than off-Marketplace enrollees considered other plans before choosing to reenroll, though the difference was not statistically significant (56.8% vs. 43.1%, p = .042). There was no significant difference in whether enrollees who had been in their plan for more than 1 year versus enrollees who were enrolled only 1 year reported that they considered other plans before choosing to reenroll (47% vs. 55%, p = .26).

Insurance Outcomes

The vast majority of enrollees who experienced plan disenrollment in 2017 had health insurance for 2018 (Table 4). Nearly two thirds (62.2%) of ACA Marketplace enrollees enrolled in another plan through the Marketplace. Among off-Marketplace enrollees, 41.8% remained in a non–group plan off-Marketplace, and 22.3% bought their 2018 plan through an ACA Marketplace. Approximately one fifth of all enrollees, 21.2% who disenrolled from an ACA Marketplace plan and 20.3% who disenrolled from off-Marketplace plans, gained employer-sponsored or union coverage. Just under 7% of disenrollees became uninsured.

Table 4.

Health Insurance Coverage Following Disenrollment, Weighted Frequencies.

	In 2017, plan was purchased:
Health insurance coverage in 2018	on ACA Marketplace (%)	Off-Marketplace (%)
Non–group plan purchased through ACA Marketplace	62.2	22.3
Non–group plan purchased off-Marketplace	1.7	41.8
Insurance through my employer or union	21.2	20.3
Public coverage (e.g., Medicaid, Medicare, other)	5.0	1.5
Other	3.2	7.2
No longer insured	6.9	6.8

Note. Source. Authors’ analysis of weighted survey data from follow-up survey. N = 498 ACA Marketplace disenrollees; N = 352 off-Marketplace disenrollees. ACA = Affordable Care Act.

Plan disenrollment was associated with a greater probability of having a 2018 plan premium that was lower than one’s 2017 premium obligation (Figure 1). However, plan disenrollment was not associated with enrollees having greater confidence in their ability to afford care in their 2018 plan if they or a family member got sick or had an accident.

Figure 1.

Health insurance premiums and affordability for 2018 (adjusted results).

Source. Authors’ analysis of weighted survey data.

Note. Results are predicted outcomes from logistic and multinomial logistic regression models controlling for enrollee age, household income, 2017 plan metal tier, family size, state, and whether a family member had a chronic condition.

^**Statistically significant difference across categories, p < .01.

Discussion

We examined administrative data linked to survey data for nongroup enrollees in plans offered by a large regional carrier across three states in order to provide a new understanding of the plan disenrollment dynamics in the non–group market. Over 70% of enrollees experienced plan disenrollment, either through changing plans within carrier or by disenrolling from the carrier altogether. High rates of plan disenrollment could be concerning if they were associated with spells of uninsurance, as was the case before the ACA (Sommers, 2014). We found that just under 7% of enrollees in our study became uninsured after plan disenrollment and even fewer reported that the elimination of enforcement of the individual health insurance mandate was a reason why they left their plans. This finding remained when excluding enrollees from Massachusetts, which was the only state of the three study states to maintain a state-based individual insurance mandate. We are unable to observe whether these plan changes are associated with any disruptions in relationships with clinical providers or the impact on quality of care. Further research is needed to provide evidence on clinical impacts of plan disenrollment.

There was greater disenrollment among enrollees who purchased their plan through the ACA Marketplace than those who purchased off-Marketplace. Our findings are associations, and our study is not designed to be able to answer why Marketplace enrollees are more likely to disenroll. ACA Marketplace enrollees could be more prone to consider and switch plans: They could have been responding to changes in the magnitude of premium subsidies due to policy (silver loading) or changes in their own income or disenrolling in response to functions of the Marketplace that promote comparison shopping.

Results of regression analyses find that benefit design changes that caused the carrier to notify enrollees of a transition onto “new” plans were associated with greater disenrollment. The state and federal insurance regulations that require this carrier notification may thus be important to activating enrollees to consider other plan options. Having already renewed plan enrollment in a prior year increased the probability that an enrollee stayed in their plan for 2018, suggesting inertia in plan choices over time. These choices are not likely all due to consumer inattention, however, as 47% of these enrollees reported that they considered other plan options before reenrolling.

Plan premiums were also very important to decisions to remain enrolled; when asked directly on the survey, 38% of enrollees reported that their plan disenrollment was because of high premiums. Those who disenrolled and found new coverage were more likely to report that they were paying lower premiums in 2018 (vs. 2017) than those who did not switch plans. It is unclear whether this is because these enrollees changed plans for the purpose of finding a lower premium or because reengaging in the shopping process allowed them to find a plan with a lower premium. We did not observe an association of disenrollment with household income. This may be because, while the lowest income households may be most likely to gain access to public coverage, higher income households may be more likely to have a new offer of employer coverage. Lower income enrollees also have less incentive to leave a Marketplace plan due to subsidies.

While plan premiums are very salient, insurance is meant to protect individuals from large losses were they to experience a negative health event. We find that after switching plans, sizeable minorities of enrollees lack confidence in their ability to afford care should they get sick. Helping individuals who have negative cost and access experiences (and who in our survey data were no more likely to disenroll from their plan) identify alternative plans where these experiences are less likely could improve satisfaction and affordability. This would also increase the incentive for plans to compete for new enrollees on overall cost and quality.

Evidence that enrollees with chronic conditions and older enrollees were less likely to have plan disenrollment is suggestive of adverse retention. This behavior could represent optimized choices of enrollees with more health care and health insurance experience for whom continuity with the same plan and providers is important. This pattern also suggests a longer run threat to competition in insurance markets and that risk adjustment remains important to the long-term stability of the nongroup market.

This study has several important limitations. Our study period included only 2 years from 2017 to 2018, so we were limited to examine plan disenrollment at one time point. This period included significant changes in premium cost for silver plans due to the discontinuation of cost-sharing reduction payments, which led to higher Silver plan premiums (e.g., Silver loading), and higher premium subsidies for those eligible. Thus, our findings may particularly reflect enrollee decisions to switch plans during a rapidly changing plan pricing environment and may limit their generalizability. While our study is multistate and includes both enrollees in the ACA Marketplaces (Federal and a state-based Marketplace) and off-Marketplace enrollees, we focus on one insurance carrier. In New Hampshire, the carrier stopped offering Bronze plans on the ACA Marketplace in 2018 (Bronze remained available off-Marketplace), and the carrier does not include enrollees in Massachusetts’ basic health plan for households below 300% FPL, so these results may not generalize to other carriers or states.

Our survey-based measure of self-reported income as a percentage of poverty may be imprecise, and some measurement error is likely. Survey nonresponse was also a limitation. However, our response rate is similar to rates achieved in similar work (Fung et al., 2017), and we were able to weight observations for nonresponse using administrative data. Finally, we report associations, and these findings should not be interpreted as causal effects.

Plan disenrollment in the non–group insurance market is common among all non–group enrollees and appears most sensitive to the type of enrollment modality (through the ACA Marketplaces vs. off-Marketplace), to significant benefit design changes, and to the duration of enrollment in a plan. The vast majority of those who disenrolled found another health plan, though 7% became uninsured. Disenrollees selected lower premium plans but did not always find plans that improved their overall cost or access. It remains to be seen whether state and federal policy strategies can support consumer transitions across non–group health plans in ways that improve their overall cost and access experiences along with affordability of plan premiums.