The Affordable Care Act (ACA) extended coverage to millions of low-income consumers through Medicaid expansion and increased enrollment in the Individual market—the former targets adults with incomes below 138% of the federal poverty line (FPL), and the latter helps cover those above that line. Even though incomes for many consumers are known to rise above or fall below the 138% FPL threshold with some frequency, comparatively little is known about how Individual market enrollees and ACA Medicaid expansion enrollees transition in and out of coverage, or how those transitions affect care utilization and spending. Better insight into these issues could help both private- and public-sector stakeholders develop improved strategies for enabling consumers to maintain coverage, as well as for better managing healthcare costs and improving outcomes.
To develop a deeper understanding of coverage transitions among these populations, McKinsey conducted a detailed analysis of data from two states (that had expanded Medicaid to nondisabled adults up to 138% FPL) provided by a large health insurer that offers both Medicaid managed care and Individual market plans. The data covered a 30-month period from 2014 to 2016. Among the questions we investigated: How often do the transitions occur? Are demographic factors such as age or health status associated with the likelihood of transitions? Do healthcare utilization patterns change when consumers re-enroll?
The answers to these questions have important implications for many stakeholders throughout the healthcare system. For example, the findings could help private payers improve retention by identifying members with a high likelihood of dropping coverage and then systematically addressing common reasons for disenrollment. Government agencies could use the results and methods to help decrease the frequency of unintentional Medicaid disenrollment and to design more sophisticated analytic approaches for understanding the consumer experience.
In this article, we briefly review our research findings and the implications for stakeholders.
New insights into member populations
To date, few studies of actual consumer behavior in the Medicaid and Individual markets have been conducted, and thus we do not know how well our findings—drawn from a single large payer’s data set—reflect those markets overall. Our experience suggests that the patterns revealed in our findings are a fairly accurate representation of both markets. (The exact percentages might change slightly, though.) Nevertheless, we recognize that opportunities for further research into the interactions between the two markets exist and use the methods and findings below to highlight important remaining questions.
Changes in coverage are common. Roughly one-third of the Medicaid expansion enrollees in our data set, and about half of the Individual market enrollees, changed coverage type or disenrolled from the studied carrier within one year (Exhibit 1).
In both groups, disenrollment
rates were lower in the second year, and re-enrollments were common. About one-third of the expansion enrollees who left coverage, and one-fifth of the Individual market enrollees who did the same, re-enrolled with the payer within one year. Re-enrollment rates to the studied payer remained persistently higher among the Medicaid expansion population than among the Individual market population (Exhibit 2).
Transitions between the Individual market and Medicaid expansion market are thought to be common. However, these transitions are difficult to fully capture because members often change carriers when making a transition. Given the limitation of observations across only one carrier, we report having observed that approximately 3% of Individual and Medicaid expansion enrollees transitioned from their starting coverage type to a different coverage type within the studied carrier:
- Of the Individual market enrollees, approximately 5% transitioned to other coverage.
- Of the Medicaid expansion enrollees, about 1% transitioned to other coverage.
At the state or national level, the percentage of Individual market or Medicaid expansion enrollees who transition to different coverage is undoubtedly higher, given that our study was limited to one payer. As discussed above, roughly one-third of the Medicaid expansion enrollees in our data set, and about half of the Individual market enrollees, left their original coverage type within one year—and some of them may have changed carriers. For example, in McKinsey’s 2017 Individual Exchange Market Consumer Survey, about 7% of the respondents said they had switched from an ACA-compliant plan in 2016 to another ACA-compliant plan in 2017 with a different carrier.
Together, these findings suggest that there is additional churn within the market that could be identified through cross-payer research.
Temporal patterns in coverage transitions can be found. Among both Medicaid expansion enrollees and Individual market enrollees, the rate of disenrollment from the studied payer spiked around eligibility re-determinations. In addition, a disproportionate number of exits among the expansion enrollees occurred in months 1 and 2. (We hypothesize that the most likely explanation is that some members switched to another carrier after having been initially auto-assigned to the carrier we studied.) The timing of re-enrollment with the studied payer was different in the two groups: more than half of the expansion enrollees who left coverage and later re-enrolled did so within 60 days. In contrast, most of the Individual market members who left coverage and later re-enrolled waited 90 days or more, and many did not re-enroll until the next open enrollment period (OEP). These results could be expected given standard grace period requirements.
Tenure, risk status, and age influence coverage continuity patterns. In both groups, the likelihood of a coverage lapse decreased as the duration of a member’s coverage increased; this pattern was especially pronounced in the Medicaid expansion population (Exhibit 3). In addition, members with higher health-risk profiles were more likely to maintain continuous coverage than were those with lower risk profiles, and they were more likely to return to their plans after a coverage lapse. In the Medicaid population, children were more likely to maintain continuous coverage than adults were. This dynamic might arise because the income-based eligibility criteria are broader for children and because one state in our analysis has a 12-month continuous eligibility policy for children in Medicaid. Among the adults in both groups, older age was associated with a slight increase in the likelihood of maintaining continuous coverage.
Cost and utilization profiles vary depending on the pattern of coverage changes. In the data set we analyzed, costs and utilization differed significantly—before and after a coverage transition—depending on coverage type and the pattern of a member’s transition. Medicaid expansion enrollees who returned to the same type of coverage after a lapse had relatively similar pre- and post-churn costs and utilization. In contrast, Individual market members who followed this re-enrollment pattern had higher costs and utilization once they regained coverage (Exhibit 4), perhaps indicating “buy-to-use” behavior.
This finding was particularly noticeable among those who re-enrolled in months that normally fall outside the OEP—April or later.
Among members who transitioned directly from Individual market to Medicaid expansion coverage, post-churn costs were low initially but increased with each subsequent month. This dynamic may indicate a lag time in accessing care after a transition.
Among members who disenrolled from Individual market coverage but did not enroll in Medicaid coverage until at least two weeks had passed, claims costs rose quickly after the transition. It is not clear, however, whether this dynamic reflects new or pent-up demand, the onset of a new health need, or (possibly) Medicaid’s lower cost-sharing requirements.
How stakeholders can respond to coverage patterns
While our study was limited in scope to one payer and two states, our findings have several potential implications for private payers and government entities.
The frequent changes in coverage we observed suggest that these organizations have an opportunity to refine their tactics to improve retention and financial performance. Lapses in coverage and subsequent re-enrollment may result in higher costs and utilization, and could impair members’ health status if disruptions in coverage affect care continuity. Certain actions may enable payers to improve retention and financial performance, including the following:
Strengthen predictive modeling capabilities. The primary goal here is to identify and target members with a high likelihood of exit. Our research suggests that relatively predictable patterns can be found in how and when coverage transitions occur; a tailored analytic approach could reveal more about these patterns. Some patterns are associated with structural factors (e.g., pregnant women can lose Medicaid eligibility 60 to 90 days after giving birth). In these situations, payers could reach out proactively (to the extent permitted by state marketing rules) to help members obtain other coverage. These organizations could also influence nonstructural factors associated with transitions, such as eligibility re-determinations, where allowed by Medicaid or Individual market rules and consistent with state policy objectives. Predictive modeling could give payers insights to determine the likelihood of exit at the member level and to develop tailored, highly personalized outreach approaches to address the reasons for disenrollment before it occurs. Furthermore, payers may have the opportunity to develop predictive models that integrate information about underlying social determinants of health. Such models could inform nonmedical interventions that promote member health and engagement, potentially yielding higher retention.
Address common reasons for disenrollment systematically. The transition patterns we identified could help payers develop deeper insights into the reasons for disenrollment. In the Medicaid expansion population, for example, the high disenrollment rate in the first two months of coverage may result from members discovering that their preferred providers are not in network. To reduce the rate at which enrollees switch plans, payers could offer them more transparent and easily accessible information about a plan’s provider networks. In addition, payers might want to consider including select high-cost providers in their networks if a sizeable number of members are leaving because those providers are excluded (subject to an analysis of the return on investment of doing so).
In contrast, some of the disenrollment that occurs around eligibility re-determinations in both the Medicaid expansion and Individual market populations may be unintentional (for example, if members do not proactively renew). To minimize this risk, payers and states could collaborate to conduct targeted, proactive outreach (to the extent permitted by state marketing rules) to members to remind them of the value of continuous coverage and the need to re-enroll; such an approach could be particularly helpful if Individual market members face significant premium increases during the next OEP. A structured comparison of consumer journeys within and across states could identify unwanted variations in transition patterns, which could indicate gaps in payer networks or anomalies in state enrollment processes.
Consider the value of long-term relationships. As we have shown, different consumer segments have different patterns of disenrollment and re-enrollment. In both the Medicaid expansion and Individual market populations, older members and those with higher risk profiles are more likely to maintain continuous coverage. (The same is true for children covered by Medicaid.) Payers that can keep these members engaged in their care are likely to achieve better outcomes and derive greater value. This is not to say, however, that payers should ignore younger adults and those at lower risk, just because they are more likely to drop coverage. Here, the challenge is to find ways to develop lasting relationships. For example, if income fluctuations contribute to the decision of many younger adults to drop Individual market coverage, payers might want to help them better understand their eligibility for Medicaid.
Consider coverage transitions in the context of a broader member management strategy. The fact that many Medicaid expansion enrollees resume coverage after a lapse suggests that programs that accrue savings over the long term—including many care management programs—may have a higher return on investment than simple exit-constrained results might indicate. In contrast, the increases in utilization observed among Individual market members who lose and then regain coverage or transition from Individual to Medicaid coverage suggest that at least some of the members may have experienced disruptions in care associated with buy-to-use behavior. Payers could target these members for early interventions to address their needs while appropriately managing utilization. Payers could consider sharing patient data across lines of business (to the extent allowable) to improve care management and the completeness of risk coding. Payers could also consider cross-payer research opportunities to better understand Individual and Medicaid transitions.
Pay increased attention to consumers who re-enroll in Individual market plans during special enrollment periods. Our hypothesis that buy-to-use behavior may contribute to the increased post-churn costs and utilization among Individual market members is reinforced by results from our 2016 Individual Exchange Market Consumer Survey, which found that post-churn costs and utilization were higher among the consumers who purchased coverage during special enrollment periods than among those who regained coverage during the OEP.
In addition, among all Individual market enrollees who regained coverage, the proportion of consumers with increased post-churn costs and utilization was higher during special enrollment periods than during the OEP. A higher prevalence of buy-to-use behavior in off-cycle months might be expected, given the lack of marketing and outreach to a broad pool of potential enrollees at these times. Nevertheless, this pattern could have negative effects on payers as they seek to manage their medical expenses—since they do not have 12 months of claims data for these members, risk-adjustment payments may not adequately compensate for increased complexity.
Federal and state government agencies involved in healthcare must juggle several tasks: administer the relevant healthcare programs and manage costs, establish and enforce market regulations, and provide both market players and consumers with the information they need. Actions these agencies could consider include:
Reduce unintentional disenrollment among Medicaid expansion enrollees. Many of the changes in family circumstances that affect Medicaid expansion eligibility (e.g., pregnancies, income fluctuations) do not follow the annual Medicaid enrollment cycle; nevertheless, we observed high disenrollment rates around eligibility re-determinations. We also found that expansion enrollees are more likely to quickly re-enroll in coverage than Individual market members are, which suggests that many Medicaid disenrollments are unintentional. That Medicaid disenrollment rates do not appear to peak uniformly at the end of a Medicaid coverage year suggests either delays in managed care organization (MCO) enrollment (compared with the effective date of Medicaid expansion eligibility determinations), imperfect administration of annual eligibility re-determinations, or both.
State Medicaid agencies might want to evaluate their eligibility re-determination policies and how those policies are implemented to investigate whether the policies may inadvertently contribute to disenrollment. They could also consider using advanced analytics to find out whether specific populations or administrative jurisdictions within their states have unusually high disenrollment rates. In addition, the agencies might want to consider whether they could—within regulatory guidelines on marketing—allow payers to assist in more comprehensive outreach programs (particularly online and through mobile devices) to engage with lapsed members and help them regain coverage or to help current members maintain coverage (without any gaps) through assistance with the re-determination process. And, they could consider collaborating with payers to increase retention rates and avoid the adverse health outcomes that may result from coverage lapses.
The effect of member outreach was underscored when we compared Medicaid re-enrollment rates in two states. One of the states used a consumer-dependent re-enrollment process; the other did not. In both states, we detected an increase in disenrollments at the one- and two-year anniversaries—but the increase was much lower in the state with the less consumer-dependent process.
Empower payers by improving the quality of Medicaid data. Another cause of unintentional disenrollment may be the absence of accurate, comprehensive data about expansion enrollees (e.g., eligibility re-determination dates, updated contact information, or other demographic information that may correlate with retention rates). Efforts to improve the completeness and accuracy of member data that state agencies give to payers could enhance the effectiveness of outreach programs to encourage re-enrollment. In addition, the federal government could consider setting new standards for Medicaid data quality by making sure that eligibility re-determinations and member contact information (e.g., cell phone numbers, email addresses) are included in the data states share with payers. Both federal and state agencies may also consider whether there are any other ways they could improve data quality to make it easier for payers to use analytics to develop a more data-driven approach to member engagement.
Work with payers to equalize consumer profitability and member targeting. Our analyses show that certain Individual and Medicaid members may represent a higher lifetime value to payers, and that enrollee characteristics such as age, enrollment tenure, and health status can be correlated with both identifiable enrollment patterns and actuarial costs. Further analysis using methods similar to the ones we employed may reveal opportunities to incentivize positive payer behavior in member outreach.
Support multi-payer research to inform a market level view of transitions. The findings presented above support a number of insights that could be expanded and strengthened with enrollment records from additional payers and years. Expanding research to other payers that offer Medicaid and/or Individual market plans could help establish overall rates of exit and segment-to-segment transfer, enrich views of pre- and post-transition healthcare use, and enable a deeper understanding of initial enrollment and plan (re-)selection in Medicaid. Additional years of data could strengthen analysis of carrier affiliation and better characterize Medicaid member profiles. Government agencies are in the unique position to encourage and sponsor such research.
The insights and best practices in this article are intended to serve as a starting point for further analysis and discussion. Our hope is that others can contribute to this type of research over time and, more important, take action to achieve goals for coverage and retention and improve health outcomes for the broader population. The preliminary findings suggest a range of actions that could be taken to promote enrollment and reduce coverage lapses. These efforts should be rooted in data and analytics to ensure that engagement and interventions are tailored to the causes of coverage lapses for specific member segments.