Effects of Medicaid Automatic Enrollment on Disparities in Insurance Coverage and Caregiver Burden for Children with Special Health Care Needs

Abstract

We analyze how Medicaid automatic enrollment policies for children with special health care needs (CSHCN) who are enrolled in Supplemental Security Income (SSI) reduce disparities in health insurance coverage and caregiving burden. Using the 2009–2010 National Survey of Children with Special Health Care Needs, we implement a difference-in-differences regression model comparing insurance enrollment rates between CSHCN receiving SSI and CSHCN not receiving SSI, in states with and without automatic enrollment policies. We find that Medicaid automatic enrollment has a meaningful impact on insurance enrollment for low-income CSHCN who participate in SSI and can be an effective method for mitigating disparities in insurance coverage (reducing uninsurance by 38%). Medicaid automatic enrollment also reduces caregiver burden among socioeconomically disadvantaged families with CSHCN. The effects of these policies are largest families who might be on the margin of eligibility or who face high administrative burden.

Introduction

Access to Medicaid offers crucial benefits for children with special health care needs (CSHCN). The high costs of care for CSHCN can exacerbate existing economic challenges (Newacheck & Kim, 2005), and Medicaid plays a key role in reducing financial distress for socioeconomically disadvantaged families (Galbraith et al., 2005; Ghandour et al., 2015; Jeffrey & Newacheck, 2006; Yu et al., 2008). Medicaid covers services for CSHCN that may not be covered by private insurance or have high out-of-pocket costs under private insurance, including therapies, screenings, assistive technology, home care, and respite care (Alker et al., 2020; Musumeci & Chidambaram, 2019; Newacheck & Kim, 2005). Medicaid also offers more generous coverage and increases utilization of inpatient, outpatient, and behavioral health services for many CSHCN (Jeffrey & Newacheck, 2006).

The majority of low-income families are eligible for Medicaid, but many still lack coverage or have inadequate coverage. The share of children with no insurance has been increasing since 2016, and noncitizen, Latino, Black and other racial and ethnic minority children and CSHCN have experienced disproportionately large declines in coverage (Abdi et al., 2020; Alker et al., 2020; Berchick et al., 2019; Garfield et al., 2019). Recent estimates before the pandemic found that 4% of CSHCN experience gaps in their insurance within a 12-month period and only 43% of families of CSHCN have a medical home, compared to 3% and 50%, respectively, among families of children without CSHCN (Abdi et al., 2020). Nearly one-third of families of CSHCN report that their health insurance was inadequate, with higher rates among children in non-English speaking households and children in households with higher levels of poverty (Kogan et al., 2010). These disparities have only grown since the pandemic has disproportionately affected the economic well-being of lower income and racial and ethnic minority households: material hardships are associated with higher unmet health needs for CSHCN (Fuller et al., 2019).

Many families face challenges accessing Medicaid due to a lack of knowledge, misinformation, low health literacy, stigma, language barriers, concerns about revealing immigration status, or racism (DeCamp et al., 2020; Kenney et al., 2015; Ngui & Flores, 2006, 2007; Son et al., 2018; D. R. Williams et al., 2019). These barriers are larger for low-SES parents of CSHCN who face higher financial and time costs to obtain and coordinate appropriate care for their children (Parish et al., 2009; Presser, 2003; Sanchez et al., 2004). Accessing adequate insurance and navigating the health care system increase the burden of care on families. Perhaps as a result of difficulties in navigating the formal care system, socioeconomically disadvantaged families of CSHCN are more likely to reduce their work and provide care at home, thus shouldering a larger caregiving burden (Romley et al., 2017). The presence of a stable care support system and medical home can mitigate caregiver burden and enable parents to work (Okumura et al., 2009). Challenges accessing insurance could lead to difficulties establishing a system of care, and disparities in access could exacerbate disparities in caregiving burden.

Medicaid automatic enrollment policies could mitigate these caregiving challenges for families of CSHCN. Automatic enrollment policies link enrollment in Medicaid to participation screeners in other public programs. Currently, seven states have express lane eligibility (ELE) policies that either grant eligibility or automatically enroll children in Medicaid if they participate in food assistance or welfare programs (CMS, 2021). Other strategies automatically enroll siblings or children of current Medicaid beneficiaries (DeLeire et al., 2012). These policies have been effective at increasing enrollment and at keeping children enrolled in Medicaid (Blavin et al., 2014; Dorn et al., 2018; Hoag, 2015). Yet, little is known about the extent to which reducing administrative burden may reduce disparities in insurance coverage or caregiving burden for socioeconomically disadvantaged or racial/ethnic minority CSHCN and their families.

In this article, we examine how Medicaid automatic enrollment policies affect insurance coverage and caregiving burden for CSHCN who receive Supplemental Security Income (SSI). SSI provides a monthly cash benefit (maximum of US$841 in 2022) and facilitates access to Medicaid for low-income families with children who “have a medically determinable physical or mental impairment which results in marked and severe functional limitations for […] at least 12 months” (SSA, 2021). Participation in child SSI has grown by over 40% since 2000, over 1 million children received benefits in 2019, and total annual expenditures exceed US$10 billion (Duggan et al., 2016; SSA, 2021). SSI’s income and asset tests target these benefits to families with incomes typically below 200% of the federal poverty level (FPL). CSHCN on SSI have higher severity health conditions and higher need for health services than other CSHCN (Houtrow et al., 2020). Currently, children with mental disorders represent 74% of the child SSI population, consistent with the growth in behavioral health conditions among low-income children in the United States (Boat et al., 2015; U.S. Government Accountability Office, 2012). Nearly 44% of the child SSI population are age 12–17, and two-thirds of child SSI beneficiaries are male (SSA, 2021).

A critical component of the SSI benefit package is access to Medicaid. In 34 states and D.C. (“1634 states”), all SSI beneficiaries are automatically enrolled in Medicaid. In seven states (“criteria states”), SSI beneficiaries are categorically eligible for Medicaid, but must complete a separate application. In the other nine states (“209b states”), SSI beneficiaries face additional criteria such as lower asset or income limits in order qualify for Medicaid (Rupp & Riley, 2016) (Figure 1). The 1634 versus 209(b) options for Medicaid date back to the creation of SSI in 1972. States were given the option to adopt the new federal eligibility and benefit levels for SSI and Medicaid, or to offer eligibility and benefits at least as generous as what the state offered in 1972 (Berkowitz & DeWitt, 2013; Gurny et al., 1992). Since then, SSI Medicaid policies have remained largely static to the option states chose initially, likely due to both funding constraints and policy inertia.

Figure 1.

Medicaid Automatic Enrollment Policies for SSI Beneficiaries, by State.

Source. Social Security Administration SSI Annual Statistical Report (2010).

Note. SSI = Supplemental Security Income.

Because SSI is means tested, a majority of beneficiaries are low-income and could qualify for Medicaid on the basis of income alone. However, Medicaid is an important and salient reason why adults and parents of children with disabilities seek out SSI, suggesting some may not have coverage outside of SSI (Burns & Dague, 2017; Levere et al., 2019; Yelowitz, 2000). Adult SSI beneficiaries who live in non-automatic enrollment states are 28% to 39% less likely to have Medicaid coverage than adults in automatic enrollment states (Rupp & Riley, 2016). Although families may have other options for coverage, Medicaid may offer more comprehensive coverage for the needs of disadvantaged CSHCN and their families. Yet little to no research has studied the effect of SSI-Medicaid automatic enrollment for CSHCN on insurance enrollment and caregiving outcomes, or explored the extent to which this means-tested program may improve outcomes specifically for socioeconomically disadvantaged families of CSHCN.

New Contributions

Although prior work has assessed the impact of other automatic enrollment policies, little is known about how these policies affect insurance coverage for vulnerable subgroups, including racial and ethnic minority CSHCN and CSHCN in low income households. Over one-third of low income and racial or ethnic minority children with significant special health care needs are inadequately insured, despite their high needs for care (Abdi et al., 2020). This article offers novel contributions by considering the role that automatic enrollment policies may play in mitigating disparities due to race/ethnicity or income in adequate coverage within the population of CSHCN. Furthermore, to our knowledge, no existing work examines how automatic enrollment policies may affect caregiving outcomes for families of these children. This article offers some of the first evidence of how automatic enrollment coverage expansions affect the disproportionate parental burden on families of color and low income families of CSHCN.

Conceptual Framework

The conceptual framework for this study is adapted from Andersen’s Model of Healthcare Services Utilization (Andersen, 1995) as shown in Figure 2. Health care use and health outcomes are determined by individual characteristics, including predisposing characteristics, enabling characteristics and need, and external factors in an individuals’ environment and community. Equitable access to health services occurs when demographic characteristics (e.g., age, gender) and need are the factors that drive health care use, and inequities occur when health beliefs or discrimination based on race, language or income drive access and use of care (Andersen, 1995; Han et al., 2015). At the same time, enabling resources like income and health insurance are typically the most mutable and have potential for policy interventions. Our study focuses on the role of automatic enrollment for CSHCN as an enabling factor to make health care access and caregiving more equitable.

Figure 2.

Conceptual Framework.

Two dimensions affect an individual’s ability to obtain health insurance: eligibility and enrollment. After Medicaid expansions for children in the 1990s and 2000s, the vast majority of CSHCN eligible for SSI qualify for Medicaid on the basis of income alone (Duggan et al., 2016). Yet, eligibility does not guarantee that a child and her family are able to use health insurance if they face barriers to enrollment. Disparities in access and enrollment in insurance are larger for low-income families of CSHCN and racial and ethnic minority CSHCN due to factors including information and language barriers, stigma, and structural and cultural racism (Abdi et al., 2020; Davidoff & Garrett, 2001; D. R. Williams et al., 2019). Both enrollment and eligibility are mutable factors that can be influenced by policy, but require separate interventions.

Our study examines the enrollment question separately from eligibility. We hypothesize that automatic enrollment in Medicaid reduces the administrative barriers in obtaining health insurance and eliminates some of the avenues where stigma or discrimination is likely to occur, thus increasing enabling factors for disadvantaged populations that previously did not have access to adequate health insurance. In so doing, automatic enrollment will lead to higher health insurance coverage and lower caregiving burden among these disadvantaged populations.

Methods and Data

We implement a difference-in-differences regression model that compares insurance enrollment rates and family caregiving outcomes between CSHCN receiving SSI and CSHCN not receiving SSI, in states with and without automatic enrollment policies. Because children on SSI in nonautomatic enrollment states typically are eligible for Medicaid on the basis of income, any difference in Medicaid enrollment between these groups will reflect the effect of automatic enrollment in Medicaid on insurance enrollment and outcomes, conditional on eligibility.¹

We analyze the 2009–2010 National Survey of Children with Special Health Care Needs (NS-CSHCN). This cross-sectional survey contains national- and state-representative samples of CSHCN, including detailed information on child and family demographics, health characteristics, utilization, insurance coverage, and SSI beneficiary status. The unit of analysis is a child with a special health care need. The survey is telephone-based with a complex survey design, stratified by state. Over 190,000 children are screened for special needs using the five-question CSHCN screener (Bethell et al., 2002), with a target sample size of at least 750 CSHCN per state. We use individual-level respondent weights to ensure that our analyses reflect a nationally representative sample of all CSHCN.²

The NS-CHSCN is, to our knowledge, the only large, nationally representative sample of CSHCN outside of administrative sources. It is also one of few sources with information on SSI receipt and sufficient sample sizes within each state that both enable us to link on other information about state-level policies and to exploit variation between states. Furthermore, the survey contains valuable data on caregiving outcomes for family members that are not measured in administrative records. Prior to 2016, the NS-CSHCN was a companion survey to the National Survey of Children’s Health (NSCH). In 2016, the two surveys were integrated into one survey, fielded on an annual basis. The combined survey still identifies CSHCN and many details on health and caregiving, but does not release information about SSI receipt. As a result, the 2009–2010 NS-CSHCN is the most recent survey year available with information about SSI receipt required for our analytic approach. We link information about SSI automatic enrollment policies in each state as of 2010 with data from the Social Security Administration (SSA, 2009). Only two states (Indiana and Ohio) have changed their automatic enrollment status for SSI since 2010 (SSA, 2021). As we discuss below, the characteristics of the current CSHCN population remain similar to 2009–2010 across states with and without automatic enrollment, and the health and caregiving needs remain substantial.

We define several outcome measures related to insurance coverage. The NS-CSHCN identifies Medicaid enrollment in only 29 states. Because it is not possible to directly identify Medicaid enrollment in all states, we instead define three primary insurance coverage outcomes including enrollment in any public insurance (which could be Medicaid or SCHIP, and includes those with both public and private insurance), only private insurance, and no insurance. After excluding 1,714 observations due to missing covariates, our final sample includes 38,528 children, with just over 3,400 SSI beneficiaries. We also examine several outcomes related to caregiver burden, including an indicator for whether the family provides care at home, whether the child’s health condition has caused financial problems for the family, and whether parents have stopped work or reduced their work hours due to their child’s health.³ The RAND Institutional Review Board approved this study.

Analytic Approach

Our approach exploits variation in SSI-Medicaid automatic enrollment policies across states and compares differences between a treatment group that is affected by the policy (CSHCN on SSI) with a control group of children who are not affected by the policy (CSHCN not on SSI). SSI-Medicaid automatic enrollment policies did not change in any state between the late 1990s and 2014; as a result, there is no state/time variation in policy to exploit for this analysis. Instead, we exploit variation in exposure to treatment across states. We compare the treatment group (states with a 1634 policy in place) and control group (including criteria states and 209(b) states), yielding a comparison of four groups for a valid difference-in-differences comparison.

We implement the following linear probability regression⁴:

$$\begin{gathered} y_{is}={\beta }_1+{\beta }_{2}SS{I_{is}}^{\ast }AUT{O}_s+{\beta }_{3}SS{I}_{is} \\ +{\beta }_{4}AUT{O}_s+X_i\theta +{\gamma }_s+{\epsilon }_{is}, \end{gathered}$$ (1)

where $y_{is}$ is an indicator for enrollment in a particular type of insurance (any public, only private, or uninsured) or caregiving outcome for a child $i$ who lives in state $s$. The coefficient ${\beta }_2$ is the main coefficient of interest: the extent to which the outcome varies for children on SSI in automatic enrollment states relative to children on SSI in nonautomatic enrollment states.⁵ We control for individual characteristics which vary between children within a state in $X_i$ (e.g., age, sex, race/ethnicity, health condition, household education level, income level) and include state fixed effects in ${\gamma }_s$ to account for differences in other state-level policies and demographic differences that could affect all children in the state. We cluster standard errors at the state level.

This difference-in-differences comparison, along with the inclusion of state fixed-effects to account for baseline differences in demographics or policy across the two sets of states, enables us to isolate the impact of the policy on insurance coverage and caregiving outcome among treated children (children on SSI in automatic enrollment states). We thus interpret the results as the casual effect of SSI-Medicaid automatic enrollment, with the assumption that there are no unobservable differences between treatment and control states that vary for CSHCN on SSI and not on SSI. In particular, the state fixed effects would control for other Medicaid policies in a state. As shown in Figure 1, the 18 states in the comparison group are distributed throughout different geographic regions and political ideologies of the country (including, among others, Oregon, Nevada, Oklahoma, Minnesota, New Hampshire, and Virginia), making it unlikely that we could attribute any remaining effects to unobserved differences of this nature. Because the SSI automatic enrollment policies rarely change, this eliminates concerns about the potential for endogenous policy change.

Given that access to insurance is more challenging for families facing language and information barriers, low health literacy, or discrimination (DeCamp et al., 2020; Han et al., 2015; Ngui & Flores, 2006; Son et al., 2018; Stuber & Bradley, 2005; Yin et al., 2009), we also stratify the model across poverty groups (<100% FPL, 100%–199% FPL, 200%–399 % FPL, 400%+ FPL), race/ethnicity (with categories of White non-Hispanic, Black non-Hispanic, Hispanic, or other non-Hispanic race), and family composition (single parent vs. two-parent household).

We use the model estimates to generate adjusted predictions (Manning et al., 1987; R. Williams, 2012). Holding all other covariates constant at the observed levels in overall sample of children with SSI (e.g., the grand means of children with SSI), we predict what the outcomes would have been if each child was in SSI in an automatic enrollment state versus in SSI in a nonautomatic enrollment state. We calculate standard errors using the delta method, via the margins command in Stata 17.

We perform a series of sensitivity analyses to test our assumption that our model captures the effects the enrollment margin, rather than the eligibility margin. We include an indicator for whether the child would qualify for Medicaid on the basis of family income, exclude higher income observations above 400% of FPL, and include an indicator for the 209(b) states. We also estimate a specification where we define SSI receipt based on the narrower category of families who indicate receiving SSI for a disability to test the robustness of our decision to use the less restrictive definition of SSI receipt in our main sample.

Results

Table 1 presents weighted summary statistics. Children on SSI are more likely to be minority, to live in lower income households, and to participate in special education programs. Due to the lower incomes among SSI households, children on SSI are more likely to be enrolled in public insurance in both automatic enrollment and nonautomatic enrollment states, although levels are higher in automatic enrollment states. Children on SSI are also more likely to be diagnosed with the health conditions specified in the survey, including (among others) attention deficit disorder, anxiety, behavioral problems, depression, autism, developmental delay, and mental retardation. The higher prevalence of diagnosed conditions and higher use of special education services confirm that CSHCN on SSI have more severe disabilities (Houtrow et al., 2020). This patterns is similar in both automatic enrollment and nonautomatic enrollment states.

Table 1.

Characteristics of CSHCN by SSI Participation and Automatic Enrollment, 2009–2010.

	SSI			No SSI
Characteristic	Auto	Nonautomatic	p value	Auto	Nonautomatic	p value
Demographics
Age	10.26	10.10	.43	9.84	9.90	.37
% Female	38.2	35.5	.17	40.8	42.0	.04
% Non-White	59.8	41.2	.00	41.4	30.9	.00
% 0%–99% FPL	46.1	49.4	.11	19.2	17.7	.00
% 100%–199% FPL	31.2	29.1	.27	20.5	21.0	.32
% 200%–99% FPL	16.6	15.8	.57	29.2	32.6	.00
% 400+% FPL	6.0	5.7	.72	31.0	28.7	.00
% households < high school degree	22.8	21.4	.42	9.6	8.2	.00
% not English speaking	6.9	2.9	.00	5.6	2.5	.00
% receiving cash welfare	12.1	15.3	.04	15.2	14.2	.14
% in Special Education	55.2	56.8	.45	25.7	28.0	.00
% with any public insurance	83.9	81.1	.06	38.2	34.2	.00
% with only private insurance	12.4	13.3	.50	54.8	59.1	.00
% with other insurance	2.6	3.5	.16	3.0	3.9	.00
% with no insurance	1.0	2.1	.02	3.9	2.9	.00
% currently diagnosed with the following health conditions
Attention deficit disorder	47.8	46.9	.69	32.2	31.5	.21
Depression	27.0	27.9	.60	12.9	13.6	.06
Anxiety	34.6	33.1	.46	19.3	20.1	.09
Behavioral disorder	34.1	28.7	.01	14.0	12.9	.01
Autism	23.1	22.0	.52	8.8	8.5	.34
Developmental delay	52.1	53.5	.51	20.5	20.6	.86
Mental retardation	20.5	21.6	.49	5.4	5.0	.14
Asthma	38.1	35.6	.20	44.0	41.4	.00
Diabetes	2.5	1.4	.07	1.9	2.2	.06
Seizure disorder	10.8	12.9	.10	4.2	3.5	.00
Migraines	16.0	17.7	.25	11.2	10.9	.41
Head injury	10.2	14.7	.00	6.5	7.0	.13
Heart condition	10.3	10.9	.64	4.8	4.8	.93
Blood condition	7.3	6.7	.54	3.8	4.0	.48
Cerebral palsy	5.2	7.0	.05	1.4	1.3	.38
Down syndrome	3.5	4.0	.47	0.8	1.0	.06
Arthritis	7.1	4.5	.01	3.1	3.4	.11
Allergies	47.8	49.2	.52	53.6	51.9	.01
Observations	2,441	968		22,578		12,541

Note. National Survey of Children with Special Health Care Needs, 2009–2010. Statistics calculated with survey weights. CSHCN = children with special health care needs; SSI = Supplemental Security Income; FPL = federal poverty level.

Table 1 also shows evidence of common patterns in insurance enrollment and demographics between automatic enrollment states and nonautomatic enrollment states, regardless of SSI status. Other CSHCN in automatic enrollment states are more likely to be minority, to live in households with lower education levels, and to live in households whose primary language is not English. Other CSHCN in automatic enrollment states are also more likely publicly insured.

Table 2 shows the coefficients from Equation 1 with a dependent variable for a different insurance enrollment or caregiving outcome in the overall sample. The first row shows the ${\beta }_2$ coefficients from each model, and the second row shows the coefficients for the main effect for SSI (${\beta }_3$). The lower rows show the predicted probabilities and confidence intervals generated from the model for children on SSI in automatic enrollment and nonautomatic enrollment states, respectively.

Table 2.

Effects of Automatic Enrollment on Insurance Coverage and Caregiver Burden, Overall Sample.

	(1)	(2)	(3)	(4)	(5)	(6)
	Any public	Only private	No insurance	Financial problems	Reduce work	Home care
SSI * Automatic	−0.009 (0.026)	0.034* (0.020)	−0.024** (0.009)	−0.015 (0.030)	−0.060* (0.030)	−0.053** (0.021)
SSI	0.153*** (0.020)	−0.136*** (0.016)	−0.019** (0.009)	−0.037 (0.026)	0.095*** (0.028)	0.044** (0.017)
Predicted probability—SSI and auto	0.884	0.083	0.011	0.262	0.425	0.518
Confidence interval	[0.862, 0.907]	[0.62, 0.103]	[0.005, 0.016]	[0.230, 0.294]	[0.387, 0.462]	[0.485, 0.550]
Predicted probability—SSI, no auto	0.876	0.09	0.029	0.27	0.502	0.556
Confidence interval	[0.834, 0.917]	[0.05, 0.131]	[0.005, 0.053]	[0.224, 0.316]	[0.439, 0.566]	[0.499, 0.614]
Observations	38,528	38,528	38,528	38,528	38,528	38,528

Note. National Survey of Children with Special Health Care Needs, 2009–2010. Displays select coefficients from a linear probability model of the outcome in the column header. Regression additionally controls for age, sex, race/ethnicity, household poverty level, child’s health condition, and state fixed effects. Huber–White robust standard errors are clustered at the state level. Predicted probabilities estimated with the margins command in Stata. Confidence intervals for predicted probabilities are estimated using the delta method. SSI = Supplemental Security Income.

^*p < .1.

^**p < .05.

^***p < .01.

Column 1 shows that SSI children in automatic enrollment states are not significantly more likely to be enrolled in any public insurance. However, Column 3 shows that automatic enrollment leads to a 2.4 percentage point decline in the probability that children on SSI are uninsured, a sizable effect size that is reflected in CSHCN on SSI in automatic enrollment states having a 38% lower predicted probability of not having insurance. Although the confidence intervals on the predicted probabilities overlap, the likelihood of not having insurance is estimated to be 1.1% among children on SSI in automatic enrollment states, compared with 2.9% in nonautomatic enrollment states. Column 2 shows a marginally statistically significant increase in children enrolled in private insurance in the overall sample, perhaps due to spill-overs in information about insurance or health needs. However, this result must be interpreted in context with the large negative association between receiving SSI and automatic insurance: as shown in the next row, children on SSI are 13.6 percentage points less likely to have private insurance. As a result, the predicted probabilities, which take into account the large negative main effects on SSI and automatic enrollment, suggest that private insurance rates are lower among CSHCN on SSI in automatic enrollment states (8.3% vs. to 9% in nonautomatic enrollment states, albeit with overlapping confidence intervals).

Columns 4 through 6 examine the effect of automatic enrollment on caregiving outcomes. The coefficients in the first row are all negative, suggestive that automatic enrollment reduces caregiving burden for low-income families of CSHCN on SSI. Furthermore, the effects in Columns 5 and 6 (on reducing work due to the child’s health and providing home care) are statistically significant. Automatic enrollment is associated with a 6-percentage-point decline in the likelihood that family members reduced work due to their child’s health, suggesting a 15% reduction in the predicted probability of these families reducing work (with predicted probabilities 42.5% for children on SSI in automatic enrollment compared with 50.2% in nonautomatic enrollment states—a difference of 15%). Furthermore, Column 6 shows that in automatic enrollment states families are 5.3 percentage points less likely to provide care at home. Predicted probabilities are 51.8% among CSHCN on SSI in automatic enrollment versus to 55.6% in nonautomatic enrollment states, but the confidence intervals overlap.

The results in the overall sample demonstrate that automatic enrollment for CSHCN on SSI has a statistically significant effect on reducing the likelihood of being uninsured and on reducing several measures of caregiving burden. To examine the effects on key subgroups who may face the largest barriers in accessing insurance, Table 3 presents the full slate of results from all of the stratified models across key subgroups. Each row in the table presents ${\beta }_2$ from a model restricted to the subgroup listed in the row header. Because the results from all models are presented, we do not adjust for multiple comparisons.⁶ The results across subgroups are consistent with the main results: there is a statistically significant decline in the likelihood of not having insurance, and statistically significant declines in the need to reduce work or provide care at home. Taken together, the results display a pattern consistent with the theoretical predictions from the conceptual model. For example, the effect size on reducing uninsurance is nearly double the result in the overall sample for Hispanic children (5.4 percentage points).

Table 3.

Effects of Automatic Enrollment on Insurance Coverage and Caregiver Burden, Subgroup Analyses.

	(1)	(2)	(3)	(4)	(5)	(6)
Subgroup	Any public	Only private	No insurance	Financial problems	Reduce work	Home care	Observations
Overall sample	−0.009 (0.026)	0.034* (0.020)	−0.024** (0.009)	−0.015 (0.030)	−0.060* (0.030)	−0.053** (0.021)	38,528
Black, non-Hispanic	0.022 (0.049)	−0.017 (0.042)	−0.025* (0.014)	0.001 (0.041)	−0.153*** (0.038)	−0.098** (0.038)	3,810
Hispanic	−0.007 (0.042)	0.023 (0.040)	−0.054* (0.029)	−0.050 (0.065)	−0.005 (0.066)	−0.003 (0.070)	4,227
White	0.001 (0.040)	0.029 (0.029)	−0.014 (0.011)	−0.016 (0.040)	−0.040 (0.046)	−0.040 (0.041)	26,931
Other race, non-Hispanic	0.092 (0.063)	−0.063 (0.063)	−0.012 (0.019)	0.115 (0.080)	0.004 (0.082)	0.048 (0.074)	3,560
Single parent	0.001 (0.029)	0.031 (0.023)	−0.023 (0.016)	0.013 (0.040)	−0.142*** (0.040)	−0.081* (0.048)	7,501
Households <100% FPL	0.026 (0.027)	0.001 (0.020)	−0.022 (0.018)	−0.041 (0.062)	−0.096** (0.046)	−0.133*** (0.039)	6,491
Households 100%–199% FPL	−0.022 (0.040)	0.030 (0.042)	−0.024* (0.013)	−0.031 (0.054)	−0.067 (0.046)	0.015 (0.039)	7,341
Households 200%–399% FPL	−0.051 (0.063)	0.108* (0.063)	−0.018 (0.012)	0.043 (0.067)	0.046 (0.084)	−0.006 (0.039)	12,089
Households 400%+ FPL	−0.015 (0.084)	−0.009 (0.094)	−0.044 (0.031)	−0.026 (0.076)	0.048 (0.087)	0.207* (0.109)	12,607

Note. National Survey of Children with Special Health Care Needs, 2009–2010. Each row displays the main coefficient of interest (${\beta }_2$) from a linear probability model of the outcome in the column header. Each regression sample is restricted to the subgroup listed in the row header. Regression additionally controls for age, sex, race/ethnicity, household poverty level, child’s health condition, and state fixed effects. Huber–White robust standard errors are clustered at the state level. FPL = federal poverty level.

^*p < .1.

^**p < .05.

^***p < .01.

The effects of automatic enrollment on caregiving outcomes, particularly the need to reduce work and provide home care, are substantially larger for Black non-Hispanic families (15.3 and 9.8 pp for reduce work and home care, respectively), single parent families (14.2 and 8.1 pp, respectively), and families with incomes below 100% of FPL (9.6 and 13.3 pp, respectively). We did not find any statistically significant effects for White children and families, children of other race, or among families with incomes above 200% of FPL.

Table 4 shows the predicted probabilities generated from the regression models in Table 3 for CSHCN on SSI in automatic and nonautomatic enrollment states for select subgroups. Although the relatively small sample size of CSHCN on SSI in nonautomatic enrollment states limits the statistical precision for many of these predictions, the predicted probabilities again suggest there are higher probabilities of public insurance, lower probabilities of not having insurance, and lower probabilities of reduced work and home care among CSHCN in automatic enrollment states. The difference in predicted probabilities of not having insurance is largest for Hispanic children, consistent with the regression results. The differences in caregiving outcomes are largest and statistically significant for Black non-Hispanic families, single parents, and children in families with incomes below 100% of FPL, again consistent with the regression results. The fact that the main results are driven by low-income and socioeconomically disadvantaged subgroups of the population suggest that SSI-Medicaid automatic enrollment is an effective policy to mitigate disparities for subgroups facing particular disadvantage in accessing insurance and providing care for their CSHCN.

Table 4.

Predicted Probabilities of Insurance Coverage and Caregiver Burden for CSHCN on SSI in Automatic Enrollment and Non-Automatic Enrollment States, for Select Subgroups.

	Automatic	Nonautomatic	Automatic	Nonautomatic	Automatic	Nonautomatic
Subgroup	Any public		Only private		No insurance
Black, non-Hispanic	0.889 [0.852, 0.925]	0.882 [0.813, 0.951]	0.068 [0.030, 0.106]	0.093 [0.044, 0.141]	0.010 [−0.002, 0.022]	0.007 [−0.006, 0.019]
Hispanic	0.876 [0.849, 0.903]	0.874 [0.807, 0.942]	0.091 [0.067, 0.114]	0.076 [0.013, 0.138]	0.017 [0.009, 0.025]	0.066 [0.014, 0.118]
Single Parent	0.839 [0.810, 0.868]	0.836 [0.792, 0.881]	0.118 [0.091, 0.144]	0.098 [0.070, 0.127]	0.016 [0.005, 0.027]	0.026 [−0.004, 0.056]
Households <100% FPL	0.936 [0.912, 0.959]	0.909 [0.879, 0.940]	0.033 [0.014, 0.051]	0.031 [0.009, 0.053]	0.008 [0.002, 0.014]	0.020 [−0.001, 0.042]
	Financial problems		Reduce work		Home care
Black, non-Hispanic	0.230 [0.190, 0.270]	0.230 [0.156, 0.304]	0.372 [0.341, 0.403]	0.526 [0.453, 0.600]	0.487 [0.443, 0.530]	0.577 [0.533, 0.621]
Hispanic	0.268 [0.234, 0.302]	0.315 [0.186, 0.443]	0.505 [0.447, 0.563]	0.508 [0.407, 0.609]	0.512 [0.458, 0.566]	0.512 [0.398, 0.626]
Single Parent	0.251 [0.207, 0.295]	0.235 [0.172, 0.297]	0.365 [0.326, 0.405]	0.515 [0.450, 0.579]	0.499 [0.456, 0.541]	0.578 [0.504, 0.653]
Households <100% FPL	0.268 [0.234, 0.302]	0.308 [0.193, 0.423]	0.416 [0.374, 0.457]	0.514 [0.444, 0.585]	0.490 [0.459, 0.521]	0.622 [0.560, 0.683]

Note. National Survey of Children with Special Health Care Needs, 2009–2010. Displays predicted probabilities from the regression models shown in Table 3. Regression additionally controls for age, sex, race/ethnicity, household poverty level, child’s health condition, and state fixed effects. Huber–White robust standard errors are clustered at the state level. Predicted probabilities estimated with the margins command in Stata. 95% confidence intervals for predicted probabilities are estimated using the delta method. CSHCN = children with special health care needs; SSI = Supplemental Security Income; FPL = federal poverty level.

^*p < .1.

^**p < .05.

^***p < .01.

We also conduct several sensitivity analyses, shown in the supplemental appendix. We estimate four separate models: including indicators for whether the child would have qualified for Medicaid based on the specific income eligibility thresholds in his or her state; excluding children in households with incomes above 400% FPL, including an indicator for 209(b) states with additional criteria for SSI children to qualify for Medicaid,⁷ and restricting the definition of SSI receipt families who report receiving SSI for their child’s disability. Across each of these specifications, the size and significance of the results remain the same. Finally, to ensure that the results are not driven by any one state in particular, we repeat the analyses excluding each state in the comparison group one at a time. The findings are also robust to this sensitivity check.

Discussion

Our primary results reveal that CSHCN on SSI in automatic enrollment states are 38% less likely to be uninsured than children on SSI in nonautomatic enrollment states. Furthermore, we find that families in automatic enrollment states are 15% less likely to reduce their work to care for the child, and 6% less likely to provide care at home.

Because these estimates are based on a comparison with SSI children in other states—who are also broadly eligible for Medicaid, the effect among families in this income category does not reflect changes in eligibility, but instead reflects changes in the ease of access to Medicaid, assuming families value insurance and are aware of their insurance status. Increased access to insurance can be particularly important for the SSI population, which tends to have a higher need for health services and therapies which may not be adequately covered under other insurance options (Alker & Corcoran, 2020; Musumeci & Chidambaram, 2019). As a result, expanded enrollment in Medicaid could result in significant changes in access to needed services for CSHCN on SSI. More comprehensive care for a child’s health condition could mean that parents do not need to reduce their work or provide additional care at home.

Importantly, we also find larger effects among lower income families and racial and ethnic minority families. We find the largest effects on insurance coverage among Hispanic children, and the largest effects on reductions in caregiving burden among single parents and Black non-Hispanic families. Racial and ethnic minority CSHCN have also been found more likely to have disabilities associated with a given condition (Houtrow & Okumura, 2011) and are more likely to face barriers, stigma and discrimination when attempting to access insurance and care (Rosen-Reynoso et al., 2016; D. R. Williams et al., 2019; You & Singh, 2009). As a result, our results suggest that automatic enrollment could be an effective mechanism to reduce disparities in insurance access and caregiving burden for these disadvantaged subgroups of CSHCN.

We find the largest effect on reductions in insurance among Hispanic children, but we do not find any statistically significant changes in caregiving burden for Hispanic families associated with automatic enrollment policies. By contrast, we find relatively modest effects on insurance coverage for Black non-Hispanic children, but large and statistically significant reductions in caregiving burden for Black families. One possible explanation could be differences in household structure: in the NS-CSHCN 2009–2010, while 44.5% of Hispanic families in households receiving SSI and earning less than 200% of the poverty level have a single parent, 66% of Black non-Hispanic families have a single parent. Differences in the caregiving burden results could result from differences in the number of adults available in the household to help with work and caregiving responsibilities, though this warrants further study.

For context, other work on related policies such as ELE find that these policies increase Medicaid enrollment by between 4% and 7% (Blavin et al., 2014). Although our data precludes us from examining Medicaid enrollment directly, approximately 42% of our sample receives public insurance. If we assume that the full 2.8% reduction in uninsurance translated into an increase in public insurance coverage and that all public insurance coverage is Medicaid, this would yield an increase in Medicaid coverage of approximately 6.6% in the population of CSHCN. Given that some portion of this public insurance coverage is likely other sources of coverage besides Medicaid, the percentage change in Medicaid enrollment could be larger than 6.6%. Earlier work estimates that about 15% of uninsured children in the general population could be covered through automatic enrollment (Kenney et al., 2010).

Our estimated effects on having insurance coverage within the population of CSHCN are relatively large compared to these earlier estimates from the general population of children. The difference between the predicted probability of not having insurance between CSHCN on SSI in automatic enrollment states and non-automatic enrollment states is approximately 38%, although the baseline rates of uninsurance among CSHCN are quite low (2.9% in the SSI nonautomatic enrollment cohort). Although small in absolute magnitude, our results on insurance coverage—and particularly the larger results among certain minority and low-income groups—demonstrate that these policies can nevertheless have meaningful impacts on high-need subpopulations who are difficult to reach. Furthermore, to our knowledge this is the first work to demonstrate the effects of automatic enrollment on downstream outcomes, and caregiving outcomes—which are particularly relevant for the population of families with CSHCN.

Data limitations present some challenges for this work. Because of the limitations on directly identifying Medicaid enrollment in the NS-CSHCN, our results focus only on the margins on enrollment in public versus private insurance or no insurance, and we are limited in exploring the details of insurance types for children enrolled in both public and private insurance. However, some analyses analyzing enrollment in the set of 29 states where we can directly identify Medicaid enrollment suggests a similar qualitative pattern in insurance enrollment as in our main results. Our approach also compares subgroups of CSHCN within a single cross section, due to the structure of the NS-CSHCN, the lack of variation in SSI automatic enrollment policies over time, and the omission of the SSI variable from more recent waves of the NSCH. We assume that CSHCN who are not on SSI can be used to adequately control for broader population differences between automatic enrollment and nonautomatic enrollment states, respectively. The summary statistics in Table 1 support this assumption. Yet, the cross-sectional data limits our ability to control for unobservable differences between children in these comparison groups.

Finally, our data are from 2009 to 2010, prior to the implementation of the Affordable Care Act (ACA), so many aspects about the health care environment have changed. Since the time period captured by our analysis, many families may have gained access to coverage through the health insurance exchanges and expansions in coverage for adults may have spilled over to children (Hudson & Moriya, 2017; Venkataramani et al., 2017). The rollout of ELE policies linking Medicaid access to other programs may have also expanded access to CSHCN whose families participate in multiple programs. However, only seven states currently have ELE policies (CMS, 2021) and SSI automatic enrollment policies have remained mostly static since 2010, meaning there is substantial scope for expanding automatic enrollment linked to SSI as well as other policies. Gaps in insurance coverage and adequacy remain for low-income and minority CSHCN in the post-ACA environment (Abdi et al., 2020), and will likely only be further exacerbated by the pandemic. As a result, there remains an opportunity to further expand access to coverage for CSHCN and other children through automatic enrollment policies that link Medicaid to SSI, or to other public programs. Indeed, streamlined eligibility and automatic enrollment are still touted as important strategies to increase insurance coverage and improve health outcomes (Shafer & Frakt, 2020). Furthermore, the characteristics of this population remain similar. Supplemental Appendix Table A1 compares demographic and health patterns between CSHCN in automatic enrollment and nonautomatic enrollment states in the 2016–2018 NSCH with the 2009–2010 sample used in our article. In both samples, automatic enrollment states have a higher share of non-White CSHCN, higher educated populations, and higher shares of the population on public insurance. Although there are some differences in the levels between these two sets of states across the two samples, the relative differences between automatic enrollment and non-automatic enrollment states are similar, making the differential effects captured in our analysis relevant in the current context.

In sum, we present new evidence that SSI policies for Medicaid automatic enrollment can have a meaningful impact on increasing insurance enrollment and reducing caregiving burden for low-income and racial and ethnic minority CSHCN. The effects of these policies are largest for families who might be on the margin of eligibility or who could face administrative, social, or racist barriers in navigating the enrollment process. Automatic enrollment could increase utilization of health services for these vulnerable populations, resulting in positive impacts on health. Future research should explore the effects of automatic enrollment on utilization, financial outcomes and unmet needs for CSHCN.