Insurance coverage during transitions: Evidence from Medicaid automatic enrollment for children receiving supplemental security income

Abstract

Objectives

To analyze relationships between Medicaid automatic enrollment for child Supplemental Security Income (SSI) recipients and health insurance coverage during transitions.

Data Sources and Study Setting

Medical Expenditure Panel Study, 2000–2020 and National Survey for Children with Special Health Care Needs, 2001–2010.

Study Design

Leveraging variation in SSI‐Medicaid automatic enrollment status across regions and over time, we estimate a regression model to quantify associations between automatic enrollment and insurance coverage. We validate our findings in the NS‐CSHCN.

Data Collection

Our sample includes children receiving SSI for a disability. We also analyze a subsample of children newly enrolled in SSI.

Principal Findings

Automatic enrollment is associated with a statistically significant increase in insurance coverage. Expanding automatic enrollment to all states is associated with increases in Medicaid enrollment of 3% (CI 0.9%–6.7%) among all SSI children and 7% (CI 1.1%–13.9%) among children newly enrolled in SSI. We find similar decreases in uninsurance. Analysis in the NS‐CSHCN replicates these findings.

Conclusions

Medicaid automatic enrollment policies are associated with increased insurance coverage for SSI children, particularly those transitioning into the program. Medicaid policy defaults could play an important role in reducing administrative burdens to improve children's coverage and access to care.

What is known on this topic

• Gaps in insurance have significant consequences for children's health, particularly for children of color and children with complex health needs.

• Automatic enrollment and express‐lane eligibility policies increase Medicaid coverage in the general population of children.

• During the 2000s, Medicaid automatic enrollment for children on Supplemental Security Income (SSI) led to declines in the uninsured rate and declines in caregiver burden for their families.

What this study adds

• The uninsured rate is higher (average 5.6%) among children transitioning into SSI than in the general child population (average 4%)

• SSI‐Medicaid automatic enrollment continues to have a strong relationship with insurance coverage for children on SSI in the post‐ACA era

• Expanding SSI‐Medicaid automatic enrollment nationwide would be associated with a 3–7% increase in Medicaid enrollment (means 82–86%) among all SSI children and new child SSI enrollees

1. INTRODUCTION

Medicaid provides critical health benefits to low‐income and disadvantaged children, yet administrative burden prevents many eligible children from accessing needed care. ¹ , ² Defined as the learning, compliance, and psychological costs entailed in accessing public benefits, administrative burden leads to significant declines in child Medicaid enrollment. ³ , ⁴ , ⁵ The costs of administrative burden are top of mind during the unwinding from the coronavirus pandemic Public Health Emergency (PHE). As of August 2023, approximately 763,000 children were disenrolled from Medicaid in the 34 states reporting data. ⁶ An estimated 73% of these disenrollments were due to procedural reasons. ⁷ Even short periods without insurance are associated with an increase in unmet health needs, ⁸ , ⁹ fewer office‐based doctor visits, ⁹ , ¹⁰ increases in the risk of hospitalization, ¹¹ , ¹² and high administrative costs. ¹³ The risk of losing insurance and experiencing these negative outcomes is more likely to fall on children of color. ¹⁴ Consequences could be particularly acute for children with unique and complex health needs. ⁹ , ¹⁴

While some administrative procedures disrupt access to health insurance, others, including automatic enrollment, can streamline access. ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ Under Title 16 Section 1634, Medicaid agencies in 34 states and the District of Columbia have agreements enabling the Social Security Administration (SSA) to make Medicaid determinations for Supplemental Security Income (SSI) recipients. ¹⁹ In these “1634 states”, SSA sends eligibility information about SSI recipients to Medicaid agencies, who automatically enroll them. In 7 “criteria” states, SSI recipients are automatically eligible for Medicaid but must apply separately. In 9 “209(b)” states, SSI recipients must meet additional criteria, such as asset or income limits, to qualify or enroll in Medicaid (see Figure 1).

FIGURE 1.

Supplemental security income (SSI)‐Medicaid automatic enrollment policies by state. Map shows variation in Medicaid enrollment policies for SSI beneficiaries by state. Dark blue 1634 states have set up agreements between SSA and state Medicaid agencies to automatically enroll SSI recipients in Medicaid. Blue criteria states grant SSI recipients automatic eligibility but require separate enrollment paperwork. Light 209(b) states do not automatically grant Medicaid eligibility or enrollment to SSI recipients; they must meet some additional income or asset criteria. Data compiled from SSI Annual Statistical Reports. SSI, supplemental security income.

Prior work leveraging this state‐level variation found SSI‐Medicaid automatic enrollment led to declines in the uninsured rate for SSI children and declines in caregiver burden for their families. ²⁰ Using a different data source, this paper offers two new contributions. First, we offer new evidence about the relationship between SSI‐Medicaid automatic enrollment and insurance coverage for children after the passage of the ACA. Second, we examine Medicaid enrollment among children newly enrolled in SSI, who could experience the highest administrative costs. It is not possible to identify new enrollees in the data sources used in prior work.

1.1. Data and variables

We use the Medical Panel Expenditure Survey (MEPS) panels from 2000 to 2020, focusing on 2010–2020 for our primary analyses. ²¹ The MEPS two‐year panel tracks health insurance coverage, utilization, and unmet needs on a monthly and annual basis. Income, including SSI income, is reported annually for each household member. Our primary sample includes 1246 children (2469 child‐years) ages 0–17 reporting receipt of SSI income. Our second sample is the subset of children (392 children, or 769 child‐years) who are new enrollees. We classify a child as a new enrollee if they report SSI income in their second year on the panel but not in the first.

We replicate our analysis in the National Survey of Children with Special Health Care Needs (NS‐CSHCN). This cross‐sectional survey includes large representative samples of CSHCN, information on demographics, health, utilization, insurance coverage, and SSI status. Our sample includes 9325 children receiving SSI from three survey waves (2001, 2005–2006, and 2009–2010). Rennane and Dick 2023 provides more detail on using the NS‐CSHCN for analyzing children on SSI and Medicaid. ²⁰ The NS‐CSHCN has state identifiers but ends in 2010. The NSCH was redesigned in 2016 to incorporate information from NS‐CSHCN but does not include SSI receipt from 2016 to 2021. Thus we use the MEPS for our primary analysis and the NS‐CHSCN to validate our findings. The RAND Institutional Review Board approved this study.

Because state indicators are unavailable in the public use MEPS, we generate a regional measure (using Census regions) of the share of SSI children living in a 1634 state using publicly available data on states' automatic enrollment policies ²² and the number of children enrolled in SSI in each state and year. ²³ There is geographic clustering in automatic enrollment policies: Figure 2 shows that from 2000 to 2020, the share of SSI children in a 1634 state is lower (25%–60%) in the Midwest than in other regions (80%–90%). The discrete increases in the share of children in 1634 states in the Midwest reflect that Indiana and Ohio changed from 209(b) to 1634 states in 2014 and 2016. ²⁴ , ²⁵

FIGURE 2.

Share of child supplemental security income (SSI) population living in 1634 states, by region. Figure shows the share of the child SSI population in each region living in a state with Medicaid automatic enrollment for SSI beneficiaries. Denominator is the total number of SSI children in each region (averaged over the time period) and numerator is the number of SSI children in states with automatic enrollment (1634 states in the region. Data aggregated from SSI Recipients by State and County, SSA Publication No. 13–11,976 and SSI Annual Statistical Reports, 2000–2021. We classify states to regions using the U.S. Census definition of region. SSI, supplemental security income.

We collect annual data on other state enrollment policies from 2000 to 2020 using the Kaiser Family Foundation 50‐state survey on Medicaid and CHIP eligibility policies, including whether the state has presumptive eligibility for children, requires a face‐to‐face interview at enrollment, or has 12‐month eligibility for children. ²⁶ We compile data on states' Medicaid expansion status beginning in 2014. We create similar regional share measures of these variables weighted by the child SSI population. As in similar applications of state policy data, ²⁷ we code policies so higher shares indicate that more states in the region adopt policies encouraging enrollment. Figure A1 shows the share of each region adopting these policies fluctuates from 2000 to 2020, and has a low correlation (ranging between −0.22 and 0.04) with regional 1634 shares.

1.2. Methods

We compare insurance rates, demographic, and health characteristics between child SSI recipients in different regions to understand the relationship between SSI‐Medicaid automatic enrollment and insurance coverage for SSI children. Then, we estimate a regression:

$$y_{\mathit{irt}}=\alpha +\beta {\mathit{\text{pctauto}}}_{\mathit{rt}}+X_{\mathit{it}}\theta +P_{\mathit{rt}}\varphi +{\delta }_t+{\epsilon }_{\mathit{irt}},$$

where $y_{\mathit{irt}}$ indicates insurance coverage for child i in region r and year t (uninsured all year, ever have Medicaid during the year, ever have Medicaid managed care (MC) during the year, only have private insurance during the year), ${\mathit{\text{pctauto}}}_{\mathit{rt}}$is the share of SSI children in region r who live in a 1634 state, and $X_{\mathit{it}}$ includes demographic characteristics (age, sex, race, household poverty level, household education, primary household language, household marital status, and indicators for severe difficulty seeing or hearing). ${\delta }_t$are year‐fixed effects. Because there is little within‐region variation in automatic enrollment policies we include time‐varying regional measures of related Medicaid policies in $P_{\mathit{rt}}$instead of region‐fixed effects. Conditional on these observed characteristics and regional policy controls, we assume any unobserved differences are uncorrelated with our outcomes of interest. The $\beta$ coefficient reflects the association between changing the share of children with access to automatic enrollment in a given region from 0 to 1 (corresponding to all states being 1634 states) on the outcome. We incorporate survey weights and cluster standard errors at the primary sampling unit (PSU).

We first estimate this model with all children on SSI. To understand the role of automatic enrollment during periods of transition, we repeat the analysis with new SSI enrollees. We estimate the model in the NS‐CSHCN with the indicator for the state's 1634 status as well as the regional share, including the same controls.

1.3. Results

Figure 3 compares unadjusted uninsured rates across regions for all children in the MEPS (left gray bars), all children enrolled in SSI (middle white bars), and new child SSI enrollees during their second year in a MEPS panel (right navy bars) from 2010 to 2020. Uninsured rates in the general population are 2.3 percentage points higher than in the SSI population in the West (4.0% vs. 1.7%), and 1 percentage point higher in the South (5.3% vs. 4.3%) and Midwest (3.3% vs. 2.2%). There are no children on SSI who are uninsured in the Northeast in our sample; recall from Figure 2 that nearly all states in these regions have automatic enrollment. All of these differences are statistically significant at the 5% level with the exception of the South (p‐value 0.07). Given that children on SSI have complex health conditions, ²⁸ their families may be more likely to seek out and maintain health insurance.

FIGURE 3.

Share of children who are uninsured, by region. Data from the Medical Panel Expenditure Survey, 2010–2020. Analysis conducted using survey weights. All children are between the ages of 0–17 (n = 52,859), all children on SSI are children ages 0–17 with annual income from SSI (n = 1264). Children new to SSI are children ages 0–17 with income from SSI in their second year on the panel (n = 392). We classify states to regions using the U.S. Census definition of region. SSI, supplemental security income.

The right blue bars show uninsured rates among new SSI enrollees. The uninsured rate among new SSI enrollees is nearly triple the rate in the overall child population in the South (11.6%), and more than double the rate of the overall child population in the Midwest (7%).

Of course, other characteristics of SSI children vary across regions. Table A.1 shows variation in race and ethnicity (compared SSI children in the Midwest and South, those in the Northeast and West were more likely to be of Hispanic ethnicity), income (SSI children in the Northeast had slightly higher income); and family composition (children in the Midwest and West were more likely to live in households with married parents). We account for these characteristics in our regression model. Table 1 Panel A shows coefficients on the regional share of SSI children in a 1634 state from equation 1 (${\mathit{\text{pctauto}}}_{\mathit{rt}}$) on the full sample. Automatic enrollment is associated with a statistically significant and sizeable decline in uninsurance. The coefficient reflects a change in the share of 1634 states in a region from zero to 1 (corresponding to all states being 1634 states). A change of this magnitude would be associated with a 7 percentage point decline (CI −13.8 to −0.1) in the likelihood that a child is uninsured for the entire year and an 18.5 percentage point increase (CI 4.3 to 32.7) in the likelihood that a child is covered by Medicaid during the year. There is an associated 11.6 percentage point decline in the likelihood that the child is covered by private insurance (CI −24.4 to −1.2), suggesting some crowd‐out.

TABLE 1.

Association between Medicaid automatic enrollment and insurance coverage for children on Supplemental Security Income (SSI).

	Uninsured	Any Medicaid in the year	Any Medicaid MC in the year	Only private insurance
Panel A: All SSI (MEPS)
Share of SSI kids in region in 1634 states (pctauto)	−0.0697** (0.0347)	0.185** (0.0720)	0.220 (0.139)	−0.116* (0.0650)
Observations with SSI (person‐year)	2469	2469	2469	2469
Predicted change in dependent variable moving from mean to 100% automatic enrollment	−0.013	0.033	0.040	−0.021
Y‐mean	0.025	0.867	0.561	0.106
Panel B: New to SSI (MEPS)
Share of SSI kids in region in 1634 states (pctauto)	−0.176* (0.0960)	0.354** (0.154)	0.606** (0.246)	−0.172 (0.123)
Observations with SSI (person‐year)	769	769	769	769
Predicted change in dependent variable moving from mean to 100% automatic enrollment	−0.031	0.062	0.105	−0.030
Y‐mean	0.0562	0.823	0.491	0.114
Panel C: All SSI (NS‐CSHCN)
Share of SSI kids in region living in an automatic enrollment state	−0.0175 (0.0227)	0.0484 (0.0460)		−0.103*** (0.0383)
Observations (person)	9294	6475		9325
Y‐mean	0.076	0.769		0.301
Panel D: All SSI (NS‐CSHCN)
Indicator for automatic enrollment state	−0.0338*** (0.0106)	0.0596*** (0.0207)		−0.00923 (0.0154)
Observations (person)	9294	6475		9325
Y‐mean	0.076	0.769		0.301

Note: Data from the Medical Panel Expenditure Survey, 2010–2020 and the National Survey of Children with Special Health Care Needs, 2001–2010. Analyses conducted using survey weights and standard errors are clustered at the primary sampling unit (PSU) level. Regression also controls for age, sex, race, household poverty level, household marital status, highest household level of education, household language, disability indicators and year‐fixed effects and other state‐level Medicaid enrollment policies aggregated to regional shares. Key independent variable (pctauto) reflects the share of children in each region living in a 1634 state, on a scale from 0 to 1 (where 1 indicates all states in the region have automatic enrollment). The predicted change re‐scales the coefficient by the baseline share of children in 1634 states in each region. Standard errors in parentheses.

Abbreviations: MC, managed care; MEPS, Medical Panel Expenditure Survey; NS‐CSHCN, National Survey of Children with Special Health Care Needs; SSI, supplemental security income.

^*** p < 0.01;

^** p < 0.05;

^* p < 0.1.

Because 34 states and the District of Columbia are already 1634 states, the relevant margin for policy change at the national level is not to move the share of 1634 states from 0 to 1. Instead, we scale the coefficients in each region by the current baseline automatic enrollment rate in the region and take the population‐weighted average of these regional predictions to reflect the associated change in insurance rates if automatic enrollment increased from current rates to 100%. (e.g., in the Northeast, 95% of children are in automatic enrollment states, so we multiply the coefficient by 5% but multiply by 56% in the Midwest, where an average of 44% of children are in automatic enrollment states during this time period). Increasing from current rates to 100% automatic enrollment would be associated with a 1.3 percentage point decline (CI −2.5 to 0.0 pp) in the likelihood of being uninsured (52% relative to the mean uninsured rate) and a 3.3 percentage point increase (CI 0.8 to 5.9) in the likelihood of having Medicaid coverage at some point during the year (3.7% relative to the mean).

Panel B shows that the coefficients are even larger for new SSI enrollees. In this subgroup, changing from 0% to 100% 1634 states in a region is associated with an 18‐percentage point decline in the likelihood that a child is uninsured for the entire year (CI −36.5 to −1.3) and a 35‐percentage point increase that the child is enrolled in Medicaid (CI 5.1 to 65.7). Compared to the regional baseline automatic enrollment rates, moving to 100% automatic enrollment from current rates would be associated with a 3.1 percentage point decline (CI −6.3 to 0.0) in the likelihood that a new SSI enrollee is uninsured for the entire year (55% relative to the mean), and a 6.2 percentage point increase (CI 0.9 to 11.4) in the likelihood of Medicaid enrollment during the year (7.5% relative to the mean).

Panels C and D replicate the analysis in the NS‐CSHCN. The direction and significance of coefficients are largely the same, although the coefficients are smaller than in the more recent period in MEPS. In the version of the model with the state indicators, automatic enrollment is associated with a 3.4 percentage point decline in the likelihood of being uninsured for the entire year (CI −5.4 to −1.3), and a 6.0 percentage point increase in the probability of having Medicaid in the year (CI 1.9 to 10.0). These results are broadly consistent with prior work (though different specifications) on this topic using NS‐CSHCN. ²⁰

Table A.2 shows robustness checks. We estimate a logistic regression model on our primary sample from 2010 to 2020 and report the marginal effects. Next, we repeat the linear analysis in MEPS on the sample from 2000 to 2020, and from 2000 to 2009. Across all specifications, the results are similar in magnitude and significance to our main results.

2. DISCUSSION

Children on SSI have complex and significant health needs ²⁸ and live in low‐income families, ²⁹ meaning most are eligible for Medicaid based on household income. Their SSI status demonstrates that their families are able to navigate bureaucratic enrollment processes, and they likely have had health insurance at some point in the past when receiving care to diagnose their disability. ³⁰ Nevertheless, substantial shares of these children do not have consistent insurance: from 2010 to 2020, on average 2.5% of all children on SSI and 5.6% of new SSI enrollees were uninsured for an entire year. The lower uninsured rates among children who have been on SSI longer than 1 year could suggest that these disruptions may not be permanent, and some families do find their way to health insurance coverage. Yet in the interim, the gaps in insurance coverage could result in significant health consequences for children without access to needed care. ³¹ , ³²

Families and children risk losing health insurance due to administrative burden ³³ —even if they have high needs for care. Our regression analysis shows that automatic enrollment facilitates access to insurance coverage: expanding SSI‐Medicaid automatic enrollment to all states would be associated with a reduction in uninsured rates by over 50% relative to current levels among all child SSI enrollees and new enrollees. We find these large associations in a population with high need and demonstrated ability to navigate complex enrollment processes; the relationship could be even more important among groups less likely to seek care or facing difficulty navigating administrative burdens. The results are larger than similar work examining earlier time periods, which is consistent with other work finding that increased administrative burden has reduced child Medicaid enrollment over the same time period. ⁴ , ⁵

There are limitations to this study. Our regional measure of automatic enrollment is less precise than using a state's actual 1634 status and does not fully account for other Medicaid policies that vary at the state level. Because we cannot fully account for state‐level variation, this analysis is not causal. We control for time‐varying Medicaid policies aggregated to the regional level to address this limitation to the extent possible. Second, the small sample sizes in MEPS limit the statistical power of the analysis and we are unable to examine variation across subgroups or examine downstream impacts on utilization. Finally, insurance status in surveys is self‐reported and could be prone to error, though self‐reports in MEPS have been found to be accurate. ³⁴

Our analysis shows that simple changes in administrative actions have large associations with insurance coverage, and streamlining insurance enrollment is particularly important during periods of transition in program participation or large household income changes. As Medicaid policy is redefined after the pandemic, there is an opportunity to introduce good administrative defaults which could lead to meaningful improvements in children's coverage and access to care.

AUTHOR CONTRIBUTIONS

Drs. Rennane and Dick conceptualized and designed the study. Dr. Rennane conducted the data analyses, drafted the initial manuscript, and Drs. Stein and Dick and Ms. Sobol reviewed and revised the manuscript. Ms. Sobol led the literature review and assisted with other aspects of the study.

All authors approved the final manuscript as submitted and agreed to be accountable for all aspects of the work.

FUNDING INFORMATION

Research reported in this publication was supported by the NIMHD Grants R03MD013951 (PI: Rennane) and R01MD017062 (PI: Rennane).

The NIMHD had no role in the design or conduct of this study. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

CONFLICT OF INTEREST STATEMENT

The authors have no conflicts of interest to disclose.

Supporting information

Data S1. Supporting information.

Rennane S, Sobol D, Stein BD, Dick A. Insurance coverage during transitions: Evidence from Medicaid automatic enrollment for children receiving supplemental security income. Health Serv Res. 2024;59(3):e14261. doi: 10.1111/1475-6773.14261