Moving Because of Unaffordable Housing and Disrupted Social Safety Net Access Among Children

We analyzed pre COVID-19 pandemic (2011–2019) survey data from the Baltimore, Maryland, and Philadelphia, Pennsylvania, sites of Children’s HealthWatch. Children’s HealthWatch is a serial cross-sectional study of families recruited from pediatric primary care clinics (Baltimore) and pediatric emergency departments (Baltimore and Philadelphia). Eligible caregivers had children aged <4 years, were proficient in English or Spanish, and resided in Maryland or Pennsylvania.

We limited the sample to children likely to be eligible for all 3 safety net programs (“social safety net eligible”); that is, who met 3 criteria: (1) caregivers or children ever received WIC, (2) caregivers or children ever received SNAP, and (3) the child was ever publicly insured in the past year. We excluded children whose caregivers reported transitioning out of WIC or SNAP because they became income-ineligible and children who became privately insured in the past year.

Disruptions in access to WIC and SNAP were defined as current nonreceipt of benefits among previously enrolled participants. Disruptions in Medicaid access were defined as noncontinuous receipt of public insurance over the past year, including coverage that was lost, lost and regained, or absent at any time.

Caregivers with children who lived in >1 place in the past year were asked, “The most recent time the child moved, which of the following was the most important reason for the move?” We defined cost-driven moves as moves because of: Issues related to paying the rent or mortgage, eviction or avoiding eviction, foreclosure, becoming homeless, or lack of housing subsidies. Noncost-driven moves included moves for any other reason (such as moving to a larger home or a neighborhood with more resources; see Prevalence and Reasons for Cost-Driven and Noncost-Driven Moves in the Results section). Past research on cost-driven moves and health has employed a similar categorization.¹⁴ Given the way the question was asked (“the [single] most important reason for the [most recent] move”), our categorization of cost-driven and noncost-driven moves is mutually exclusive, with no children experiencing both types. Children who lived in a single location over the past year were considered nonmovers.

Models included covariates that might simultaneously lead families to move and disrupt program access, including proxies for caregivers’ resources and bandwidth for navigating health care and social service systems. These included educational attainment (less than, greater than, or equal to a high school education), household income (increments of $12 000; top-coded above $48 000), marital status (single, married, separated/divorced/other, cohabiting), child age in years, caregiver age (<20, 20–25, 26–35, or 36+ years), and biologic mother’s nativity (United States-born or not). We also adjusted for caregiver race and ethnicity (Hispanic/Latino, non-Hispanic Black, non-Hispanic white, or another non-Hispanic race/ethnicity) as a proxy for racial discrimination in housing markets and services.^15,–17 We also included fixed effects for study site (Baltimore or Philadelphia) and year, recognizing that housing markets and safety net enrollment requirements might differ systematically between cities and over time. Finally, we controlled for visit type (acute care versus scheduled primary care). Because household income data were missing for 22% of our study population because of caregiver nonreport, we did not include income as a covariate in our primary analysis; we conducted a sensitivity analysis adjusting for income in the subsample with income data.

We first compared characteristics between children with cost-driven moves, those with noncost-driven moves, and nonmovers. We then modeled associations between each type of move (versus no move) and disruptions in program access using logistic regression. We modeled disruption in each type of program separately, as well as an overall indicator for experiencing disruption in at least 1 program. After regressions, we compared the estimated odds of disruption associated with cost-driven versus noncost-driven moves using a Wald test. We report unadjusted and covariate-adjusted results. We also report predicted probabilities of program disruption, derived from adjusted models with marginal standardization. All analyses were conducted in Stata SE, version 15.1.

The study was approved by institutional review boards at the University of Maryland School of Medicine and Drexel University. Caregivers gave consent to participation.

The Children’s HealthWatch study Baltimore and Philadelphia sites had a combined response rate of 92% during the study period. Of 13 675 children whose caregivers completed surveys during the study period, 12 118 (89%) had ever received WIC, 11 093 (81%) had ever received SNAP, and 12 932 (95%) had been publicly insured in the past year, with 10 240 (75%) meeting all 3 inclusion criteria. Among these, 767 (7%) reported becoming ineligible for SNAP and/or WIC because of increased income, and an additional 16 (<1%) became privately insured in the past year, leaving 9457 eligible children. We additionally excluded 113 children (1%) with missing covariate information (caregiver age, education, marital status, or nativity; remaining covariates had complete information), resulting in 9344 children in our analytic sample. Given our criteria, included children were more socioeconomically disadvantaged than excluded Children’s HealthWatch participants (Supplemental Table 2).

Thirty-six percent of study children were infants aged <1 year, 27% were aged 1 year, 20% were aged 2 years, and 17% were aged 3 years (Table 1). Eighty-eight percent of caregivers were aged 20 to 35 years, 31% were Hispanic or Latino, 59% were non-Hispanic Black, 93% were United States-born, 68% had a high school education or less, 80% had household incomes <$24 000 (among those with nonmissing income; n = 7250), and 63% were single. Forty-three percent were recruited from Baltimore and 56% from Philadelphia. Ninety-two percent were recruited from acute care settings or emergency departments.

A total of 1938 study children (21%) experienced a coverage gap for at least 1 social safety net program. Among these, 1461 (75%) had disrupted access to WIC, 390 (20%) had disrupted access to SNAP, and 304 (16%) had disrupted access to Medicaid (Fig 1).

Overall, 475 (5%) children experienced a cost-driven move in the past year and 2378 (25%) experienced a noncost-driven move. Reasons for cost-driven moves included issues paying the rent or mortgage (67%), eviction (17%), foreclosure (12%), entering a homeless shelter (4%), and issues with subsidy funding (1%). Primary reasons for noncost-driven moves included wanting a bigger or nicer house (26%), a family change (18%), moving into an independent household (17%), moving closer to work or school (10%), issues related to poor housing conditions (9%), and wanting a safer neighborhood (6%). The prevalence of cost-driven moves declined over the study period (likely reflecting recovery from the 2007–2008 financial crisis), whereas the prevalence of noncost-driven moves was fairly stable (Supplemental Fig 3).

Compared with children with no past-year moves (Table 1), children with 1 or more cost-driven residential move were older, had younger and less educated caregivers, and had lower household incomes. They were more likely to have United States-born caregivers, live in Baltimore, and participate in Children’s HealthWatch via an emergency or acute visit (as opposed to a primary care visit). Their caregivers were also more likely to have a history of depressive symptoms. Although children with noncost-driven moves resembled children with cost-driven moves with respect to child age, caregiver nativity, and visit type (Table 1), they had similar socioeconomic status to nonmovers.

Relative to children with no moves in the past year, children with cost-driven moves showed significantly higher odds of disrupted social safety net access (Fig 2, Supplemental Table 3). In adjusted models, we found cost-driven moves (versus no moves) were associated with 44% higher odds of disrupted access to at least 1 program (WIC, SNAP, or Medicaid; adjusted odds ratio [aOR] 1.44; 95% confidence interval 1.16–1.80). In absolute terms, these increased odds mean that 25.8% (22.2–29.4) of children with cost-driven moves experienced any program disruption, compared with 19.8% (18.9–20.8) among children with no residential moves. Analyzing each program separately, cost-driven moves were associated with 34% higher odds of disruption in WIC access (aOR 1.34; 1.05–1.71), 93% higher odds of disruption in SNAP access (aOR 1.93; 1.30–2.86), and 86% higher odds of disruption in Medicaid access (aOR 1.86; 1.24–2.78).

Noncost-driven moves were associated with a smaller but measurable increases in odds of disrupted access to any program (Fig 2, aOR 1.14; 1.01–1.28; P value comparing aOR to aOR for cost-driven moves = .045). In absolute terms, 21.9% (20.3–23.4) of children with noncost-driven moves experienced program disruptions 2 percentage points higher than the prevalence among children with no moves, but still 4 percentage points below the prevalence among children with cost-driven moves. By program, noncost-driven moves (versus no moves) were significantly associated with disruptions in WIC and SNAP (aOR 1.15 [1.00–1.31] and 1.38 [1.10–1.74], respectively), but not with disruptions in Medicaid (aOR 0.96; 0.73–1.27). These point estimates are consistently lower than point estimates measuring associations between cost-driven moves and program disruption, although tests comparing the strength of associations by move type (cost-driven versus noncost-driven) showed that the associations differed significantly for Medicaid (P = .004), but not WIC or SNAP (P = .241 and .113, respectively).

A sensitivity analysis adjusting for household income in the subsample (n = 7250) with nonmissing income data yielded similar results (Supplemental Table 3).

In this mid Atlantic sample of social safety net-eligible families with young children, children whose families experienced cost-driven residential moves were substantially more likely than children with no past-year moves to have experienced a disruption in social safety net services (WIC, SNAP, and Medicaid). Although noncost-driven moves were also associated with higher odds of service disruption, point estimates for odds of losing access to at least 1 safety net program were significantly attenuated relative to those for cost-driven moves. Results therefore suggest that cost-driven moves (versus moving per se) pose a particular challenge to program retention. This finding is concerning, because these programs are designed to protect children’s health during times of hardship.

Our findings are consistent with qualitative research suggesting that families may struggle to maintain safety net enrollment after a cost-driven move,⁷ as well as quantitative research finding an increased risk of disrupted Medicaid access among evicted patients.⁸ We saw variation in the prevalence of disruption, as well as the magnitude of associations, between the 3 social safety net programs. Specifically, disruptions in WIC were more common, but less strongly associated with cost-driven residential moves than SNAP or Medicaid. The finding of greater disruptions is consistent with data showing that WIC participation drops off substantially after infancy (ie, WIC disruption is highly prevalent in general).¹⁸ Further study is needed to understand the reasons underlying differential associations between cost-driven moves and disrupted access to each of the 3 programs.

There are several possible explanations for the identified associations. An address change might mean that caregivers miss communications about recertification or no longer know where to recertify regardless of what motivated a residential move. Stress surrounding a move may hinder caregivers’ ability to recertify for social safety net programs, especially acute stress and the urgency of finding new housing in the context of cost-driven moves such as evictions and foreclosures.¹⁹ It is also likely that, in the case of cost-driven moves, the observed association is at least partially driven by the reverse process: Disrupted social safety net access could lead to increased food and medical costs that compete with rent payments, necessitating a residential move. Studies finding that Medicaid expansion-reduced eviction rates support this hypothesis,¹¹ as do findings linking loss of SNAP benefits with reduced housing security.¹² Regardless of the direction of the associations, our results indicate that children experiencing cost-driven moves are more likely to experience social support disruptions; these interconnected deprivations may weigh heavy on the health trajectories of affected children.

Our results point to a need for interventions in clinical practice and social policy. First, child health providers should follow guidelines to screen for housing security,²⁰ recognizing that families may be at joint risk for disruptions to safety net access. Providers and clinic-based social workers might prioritize outreach to families experiencing cost-driven moves to ensure continuous enrollment in safety net programs and vice versa (ie, prioritize families with lapsed safety net access for referrals to housing assistance). Previous research finds that children coenrolled in housing and food subsidies have the greatest odds of secure housing.¹² On a policy level, social safety net programs should be cognizant of cost-driven moves and housing insecurity in program implementation. Reducing administrative burden,²¹ allowing for dual eligibility,²² or shifting to remote delivery models²³ may facilitate retention among housing-insecure households. To prevent cost-driven moves, policymakers must work to increase the supply of affordable housing and make housing assistance available to all who need it. Because structural racism places Black and Latino households at increased risk of cost-driven moves (in particular, evictions),¹⁵ future interventions ought to prioritize housing assistance and safety net retention in these communities.

Our research has several limitations. First, we could not determine temporality, because this is a cross-sectional study. Although our measure of disrupted Medicaid is specific to the past year, the timing of SNAP and WIC disruptions is more ambiguous: We could only define disruptions as current nonreceipt among children who had received the benefit previously. Longitudinal data are needed to understand the directionality of associations. Second, our approach to measuring social safety net disruptions is imperfect and could misclassify some voluntary or eligibility-related changes in participation as disruptions. For example, qualitative research on WIC demonstrates that eligible families sometimes opt out of the program.²⁴ Third, we classify evictions and entry into homeless shelters as cost-driven moves given evidence that the vast majority of evictions are because of nonpayment of rent,²⁵ and housing costs are the leading driver of homelessness in communities.²⁶ However, some of these moves may have been motivated by hardship unrelated to housing costs. Fourth, although we adjust for multiple covariates to proxy caregivers’ resources and bandwidth for navigating health care and social service systems, we lacked data on some potential confounders (eg, recent divorce or separation), and cannot rule out potential residual confounding. Fifth, because recruitment occurred in health care settings, families with access to health care and, by extension, to social services, may be overrepresented. Resulting selection bias may at least partially account for the null association between noncost-driven moves and disrupted Medicaid access. Finally, our data were collected in Maryland and Pennsylvania, 2 states with relatively robust social safety nets. Results may not generalize to states with weaker safety nets, where program access may be more vulnerable to cost-driven moves.

In this study of social safety net-eligible families with young children, cost-driven moves were associated with disrupted access to WIC, SNAP, and Medicaid. These results merit attention from policymakers, particularly given high levels of financial strain and housing insecurity among low-income families during the COVID-19 pandemic and subsequent period of inflation. Keeping children connected to housing, food, and health care is an urgent priority.

We thank Children’s Healthwatch participants, as well as the study staff and leadership, including the study investigators who reviewed this manuscript and provided useful feedback (Drs Deborah A. Frank, Stephanie Ettinger de Cuba, Megan Sandel, and Diana Becker Cutts).

aOR

adjusted odds ratio

COVID-19

coronavirus disease 2019

SNAP

Supplemental Nutrition Assistance Program

WIC

Special Supplemental Nutrition Program for Women, Infants, and Children