Insurance and Health Care Outcomes in Regions Where Undocumented Children Are Medicaid-Eligible

This study was conducted using the NSCH–19, a periodic, cross-sectional survey conducted by the US Census Bureau with sponsorship from the Health Resources and Services Administration’s Maternal Child Health Bureau. The NSCH is a nationally representative survey of caretakers of noninstitutionalized US children aged ≤17 years in the 50 US states and Washington, District of Columbia. It assesses multiple domains of health and well-being, including access to care.²² 

The NSCH is available in English and Spanish. Outreach is conducted via mail, and participants can respond via a web-based interface or paper, with telephonic and e-mail support available. An initial screener is followed by a detailed survey on 1 randomly selected child per household, with an oversampling of children aged ≤5 years and of those with special health care needs. Sampling weights are adjusted for nonresponse, and survey data are weighted to allow for generalizability at the state and national level. The NSCH–19 was administered between June 28, 2019, and January 17, 2020, and had a sample size of 29 433 (representing 77.6 million children nationally) with an overall weighted response rate of 42.4%. Additional information about the NSCH and methodology are available elsewhere.²² 

The NSCH–19 is a publicly available, de-identified data set. Its use is not considered human subjects research, and we obtained a determination of this status from the Yale Institutional Review Board.

The extended-eligibility states/territories included the 6 states, California, Illinois, Oregon, Massachusetts, New York, and Washington, and Washington, District Columbia, that provide public health insurance eligibility to children regardless of documentation status. All other 44 states were defined as nonextended eligibility.

The primary outcome was caregiver’s proxy-report of their children’s uninsured status at the time of response to the survey. Secondary outcomes were caregivers’ report regarding their child’s health care utilization in the previous year, including: (1) forgone medical care (not receiving needed health care), (2) forgone dental care (not receiving needed dental care), (3) lack of a preventive visit, (4) not having a personal physician or nurse, and (5) having an emergency department (ED) visit.

Based on previous studies of disparities in health insurance and in access to care for immigrant children, selected covariates included: being a child in an immigrant family (CIF, further defined below), sex, age, race and ethnicity, household income (reported as percentage of federal poverty level), caregiver’s education, caregiver’s employment, household structure (single mother, 2 parents, or other), and primary household language.^23,24 

The NSCH includes a “household generation” variable: first-generation children and their caregivers were born outside the United States, second generation were born in the United States but at least 1 caregiver was not, and third-generation children and their caregivers were born in the United States. Using this household generation variable, CIF were defined as first- and second-generation households, consistent with previous literature.^18,25 

Simple descriptive statistics and Pearson χ² tests were used to compare the social and demographic characteristics of children from extended- and nonextended-eligibility states/territories.

We used Poisson regression modeling with robust standard errors to examine the extent to which the primary exposure variable of residence in extended- versus nonextended-eligibility states/territories was associated with the primary and secondary outcomes of insurance and health care utilization.²⁶  Using Wald test, we assessed for statistical interactions between being a CIF and being in an extended- versus nonextended-eligibility state for all outcomes, followed by a stratified analysis for CIF and non-CIF subgroups. We adjusted these associations for a priori chosen covariates in multivariable models; the final adjusted models always retained the exposure variable and had all covariates except for (1) primary household language, and (2) caregiver education. Primary household language was not included in the final adjusted model because of known cooccurrence with CIF; according to the American Community Survey, over half (53.8%) of CIF have at least 1 parent with limited English proficiency, compared with 2.1% of non-CIF.¹³  Similarly, in the NSCH–19, 55.1% of CIF compared with 97.1% of non-CIF reported English as the primary household language. Additionally, because CIF and caregiver education were highly associated with each other in our sample, but CIF was our primary exposure variable; the final adjusted model in the combined sample of children did not include caregiver education as a covariate. Results are summarized as adjusted estimates (relative risk, [RR] and prevalence obtained via postestimation exponentiation of relevant parameters and use of marginal means to determine predicted probability, respectively, in Stata) with 95% confidence intervals (CI).

Given the magnitude of our sample size, we did not rely on a strict 2-sided α level for a hypothesis test, and P values are presented in the results as supplementary statistics. We determined a meaningful RR based upon the magnitude of the effect size and 95% CI, which do not include the value of 1.0.^27,28  If that criterion was met, we further defined a meaningful magnitude of difference in the prevalence of uninsured children at the 1% point, which reflects a difference of about 770 000 US children, a number deemed meaningful in analyses of changes in child enrollment in Medicaid and CHIP over time.²⁹ 

We conducted analyses with Stata Version 15 (StataCorp College Station, TX). Survey weights were accounted for by using the svy(subpop) command.

As shown in Table 1, of the 29 433 NSCH–19 survey respondents, 4035 were from extended-eligibility states/territories. The extended-eligibility and nonextended-eligibility states/territories accounted for 27.2% and 72.8% of weighted survey respondents, respectively. We did not observe a meaningful difference between extended- and nonextended-eligibility states/territories in terms of sex or caregiver employment. A higher proportion of children in extended-eligibility versus nonextended states/territories were in the 12 to 17 age range, lived in a 2-parent household, identified as CIF, had a non-English primary household language, and identified as Hispanic/Latino or non-Hispanic/Latino Asian; fewer identified as non-Hispanic/Latino African American/Black or non-Hispanic/Latino White (Table 1).

Fewer children in extended-eligibility states/territories, compared with children in nonextended-eligibility states/territories, were uninsured (adjusted prevalence: 3.7% vs 7.5%, Fig 2) with an RR of 2.0 (95% CI, 1.4–3.0) in the adjusted model (Table 2 and Fig 3).

CIF, compared with non-CIF, had a higher risk of uninsurance (adjusted RR, 1.5 [95% CI, 1.1–2.1]) (Table 2). Although we did not observe a significant difference in the RR of being uninsured in nonextended-eligibility versus extended-eligibility states/territories between CIF and non-CIF (P = .86 for interaction), Fig 4 displays that CIF had higher adjusted prevalence of uninsurance than non-CIF regardless of residence in nonextended (11.1% vs 5.5%, RR, 2.0 [95% CI, 1.1–3.6]) versus extended- eligibility states/territories (6.2% vs 3.1%, RR, 2.04 [95% CI, 1.2–3.4]).

More children in nonextended versus extended-eligibility states/territories had forgone medical (3.1% vs 2.2%, adjusted RR, 1.4 [95% CI, 0.9–2.2]) and dental care (20.5% vs 17.1%, adjusted RR, 1.2 [95% CI, 1.0–1.4], Figs 2 and 3).

We did not observe a meaningful difference in the RR of forgone medical (adjusted RR, 1.1 [95% CI, 0.7–1.7]) or dental care (adjusted RR, 1.1 [95% CI, 0.9–1.3]) between CIF and non-CIF (Supplemental Tables 3 and 4). We found no meaningful interaction between forgone medical or dental care and residence in an extended-eligibility state/territory (P = .94 and P = .56, respectively, for interaction). Both CIF and non-CIF were more likely to have forgone medical and/or dental care in the nonextended-eligibility versus extended-eligibility states/territories (Fig 4).

More children in nonextended-eligibility states/territories, compared with children in extended-eligibility states/territories, went without preventive visits in the previous year per caregiver report (adjusted: 17.1% vs 14.2%, RR, 1.2 [95% CI, 1.0–1.4]) (Figs 2 and 3, Supplemental Table 5).

CIF, compared with non-CIF, had higher risk of having no preventive visit in the previous year (adjusted RR, 1.2 [95% CI, 1.0–1.4]) (Supplemental Table 5). We found no meaningful interaction between CIF status and residing in an extended versus nonextended-eligibility state for lack of previous-year preventive care (P = .88 for interaction). Both CIF and non-CIF were more likely to lack a preventive visit in nonextended-eligibility versus extended-eligibility states/territories, but the rates were higher for CIF regardless of living in nonextended versus extended-eligibility regions (Fig 4).

A similar proportion of children in nonextended and extended-eligibility states/territories lacked a personal doctor or nurse, per caregiver’s report, in the previous year (adjusted prevalence 27.2% vs 25.3%, RR, 1.1, [95% CI, 1.0–1.2]) (Figs 2 and 3, Supplemental Table 6).

Overall, for CIF compared with non-CIF, the unadjusted RR of not having a personal doctor or nurse was 1.2 [95% CI, 1.0–1.3], but this association was attenuated in the adjusted model (RR, 1.0, [95% CI, 0.9–1.1]) (Supplemental Table 6). We did not observe evidence of moderation by CIF status on the association between not having a personal doctor or nurse and being in the extended-eligibility state/territory (P = .34 for interaction). For CIF, 31.8% did not have a personal doctor or nurse in the previous year in the nonextended-eligibility states/territories compared with 29.5% in the extended-eligibility states/territories (RR 1.2, [95% CI, 0.94–1.43]). By comparison, for non-CIF, 25.6% and 23.8% (RR, 1.0, [95% CI, 0.9–1.2]) did not have a personal doctor or nurse in the nonextended- and extended-eligibility states/territories, respectively (Fig 4).

Children in nonextended- and extended-eligibility states/territories had similar rates of ED visits in the previous year (19.1% vs 18.8%, RR, 1.0 [95% CI, 0.9–1.1],) (Figs 2 and 3, Supplemental Table 7).

Overall, CIF were less likely than non-CIF to have an ED visit in the past year (adjusted RR, 0.76, [95% CI, 0.65–0.90]) (Supplemental Table 7). We found no meaningful interaction for CIF status on the association between residing in an extended-eligibility state/territory and previous-year ED visit (P = .67 for interaction); for CIF, the RR of previous ED visit was 0.9 (95% CI, 0.7–1.3), and for non-CIF, the RR was 1.0 (95% CI, 0.8–1.2) (Fig 4).

In this cross-sectional analysis using data from the nationally representative NSCH–19, we found that children living in states/territories with public insurance eligibility for children, regardless of documentation status, were less likely to be uninsured, have forgone medical or dental care, and were more likely to not have preventive visits in the previous year per caregiver report. There were no meaningful differences between extended- versus nonextended-eligibility states/territories in caregivers’ report of their children having a personal physician/nurse or ED visits. Additionally, CIF were more likely to be uninsured compared with non-CIF, although uninsurance was lower for both groups in extended-eligibility states/territories.

We found that, for all children, as well as for subgroups of CIF and non-CIF specifically, there were more uninsured children in states/territories without policies that provide insurance eligibility to children, regardless of documentation status, even when controlling for multiple covariates that may affect insurance status. This builds upon the evidence that policies which expand insurance access can improve enrollment within and beyond the target expansion demographic through a “welcome mat” effect.^30,31  California’s 2016 expansion to undocumented children via its Health4All Kids initiative, for example, resulted in a 26% reduction in uninsured status among noncitizen children.³² 

Children in extended-eligibility states/territories were less likely to have forgone care and more likely to have preventive visits, which reflects findings that policies which increase insurance eligibility and enrollment are associated with decreased unmet medical care needs and increased preventive medical and dental care use; 1 systematic review and metanalysis of studies evaluating outcomes after Medicaid and/or CHIP expansion found that all included studies found reductions in unmet medical and dental care.³⁰ 

Limitations in this cross-sectional study included noted baseline differences in the sociodemographic characteristics of children in extended-eligibility compared with nonextended-eligibility states/territories, which may warrant further investigation. Children in extended-eligibility states/territories were more likely to be in the 12 to 17 age range, live in 2-parent households, identify as CIF, have a non-English primary household language, identify as Hispanic/Latino or non-Hispanic/Latino Asian, and to be in the highest income bracket as measured by percentage of federal poverty level. Although we adjusted for many of these covariates, there may be additional unmeasured confounders that underlie these state-level differences that could not be controlled for in the model. For example, states/territories which provide health care to children regardless of documentation status may have other policies that improve health and access to care, including policies to support health-related social needs like housing, employment, peer support, and case management services. Example policies, which were funded via Medicaid Section 1115 waivers and occurred in extended-eligibility states/territories, include Washington’s “Accountable Communities of Health,” New York’s Delivery System Reform Incentive Payment waiver to support housing, and California’s “Whole Person Care” pilots.³³  Additional policies, such as state driver’s license and sanctuary policies for undocumented residents, have also been associated with improved access to care for children.³⁴ 

In addition to potential unmeasured confounders related to state-level differences that were not controlled for in the model, there were additional limitations to this study. CIF, especially undocumented CIF, were likely underrepresented in this survey. The CIF variable also has limitations, as it includes citizen, US-born children with at least 1 parent/guardian born outside of the United States and does not distinguish by documentation status. The NSCH is offered only in Spanish and English, and there is limited support for those who may have limited literacy. Furthermore, the demographics of US immigrant populations continue to shift, Asian-Americans are expected to be the predominant immigrant group by 2055, thus adding to the limitations of the NSCH and CIF variable.³⁵  Additionally, undocumented residents may avoid responding to surveys for fear of affecting immigration status, part of the “chilling effect” of antiimmigrant policies.^36,37 

Although there were limitations with this cross-sectional study, it is important to assess differences in outcomes in extended-eligibility and nonextended-eligibility states and to highlight geographic inequities in access to care. Future prospective studies can evaluate changes in insurance enrollment, health care utilization, and health outcomes for both CIF and non-CIF as more states/territories expand enrollment eligibility to undocumented children. Several states, including Vermont, Maine, Connecticut, and New Jersey, have recently passed legislation to expand eligibility to undocumented children (Fig 1).^38–42  Longitudinal studies are needed to assess outcomes over time related to these policies, not only for undocumented children, but for all children. Although outcomes related to improved access to preventive care may not be apparent in the immediate time frame surrounding implementation,⁴³  public health effects, such as access to preventive care and coronavirus disease 2019 vaccinations, may be readily apparent across populations.

Residence in states/territories which offer public insurance eligibility to children regardless of documentation status is associated with higher health insurance enrollment and preventive care utilization. Children in extended-eligibility states are less likely to be uninsured, have forgone medical or dental care, or have gone without preventive care per caregivers’ report. As more states/territories continue to expand health care access for undocumented children and families, new and emerging data can be harnessed to investigate short-term and long-term outcomes for all children and families.

CHIP

Children’s Health Insurance Program

CI

confidence interval

CIF

children in immigrant families

ED

emergency department

NSCH–19

National Survey of Children’s Health 2019

RR

relative risk