Timing and Duration of Pre- and Postnatal Homelessness and the Health of Young Children

Between May 1, 2009, and January 1, 2015, researchers with the ongoing Children’s HealthWatch study interviewed 20 571 caregivers of children <4 years old seeking medical care in emergency rooms or primary care clinics in Baltimore, Maryland; Boston, Massachusetts; Little Rock, Arkansas; Minneapolis, Minnesota; and Philadelphia, Pennsylvania. Institutional review board approval was obtained at each site before data collection and renewed annually. Interviewers elicited caregivers’ verbal responses to a survey of household demographics, experiences with homelessness, and child health and development. Eligibility criteria included the following: child age ≤48 months; residency in the state of the research site; caregiver ability to speak English, Spanish, or Somali (in Minneapolis only); respondent living in the child’s household; and consenting to be interviewed.

Of the 28 067 caregivers approached, 3108 (11.1%) were ineligible for the study, and 1912 (7.7% of the eligible) refused or were unable to complete the interview. Those with missing data about homelessness (n = 231) were excluded. As a proxy for low income, families with private health insurance were excluded (n = 2245), leaving a final study sample of 20 571 caregiver–index-child pairs with public or no insurance.¹³ 

Prenatal and postnatal homelessness were determined via affirmative responses to the questions, “Were you ever homeless, or did you live in a shelter when you were pregnant with this child?” and “Since your child was born, has she or he ever been homeless or lived in a shelter?” For the latter question, interviewers followed up on affirmative answers by asking about the duration of homelessness during the child’s lifetime (<6 months, 6–12 months, and ≥12 months). For these questions, homelessness was defined as living in a shelter, motel, temporary or transitional living situation, or scattered site housing (in Massachusetts, scattered site housing is paid for by the state emergency shelter system) or having no steady place to sleep at night. Our primary exposure was then defined such that families were categorized by homelessness timing: prenatal, postnatal, both, or never. In a secondary analysis, the duration of a child’s postnatal homelessness was characterized as either at least 6 months or <6 months. This analysis was stratified by child age: <1 year of age and >1 year of age. Stratification was performed to determine the presence of effect modification based on child age group. To examine the effects of longer-term homelessness (>6 months) in the youngest children, we needed to stratify at the 6-month cut point. For consistency, we used the same cut point for children ≥12 months of age.

The child health outcomes examined in this study were as follows: current child health status, history of hospitalizations, developmental risk, and anthropometric measures. Caregivers characterized their children’s health as excellent, good, fair, or poor using a question from the 1988–1994 NHANES. We administered the Parents’ Evaluation of Developmental Status (PEDS), a validated, 10-question, widely used screening instrument that is used to identify developmental risk. Because the PEDS sensitivity and specificity are better for children >4 months of age, we asked the PEDS questions of caregivers with index children ≥4 months of age. On the basis of the standard scoring of the PEDS, endorsed items are classified as significant or nonsignificant concerns depending on the age of the child. Children with ≥2 significant concerns are scored as being at developmental risk with clinical indication for diagnostic testing (“developmental risk”).^14,–16 

Caregivers also reported the number of the children’s previous hospitalizations, excluding those at birth. Children were categorized as having a history of any postnatal hospitalizations versus having none. Children’s weight and length were ascertained from their medical records on the day of the interview to determine whether they fell within normal anthropometric parameters. Anthropometric parameters were defined as whether a child was at risk for underweight status (weight for age less than the fifth percentile or weight for length <10th percentile for Centers for Disease Control and Prevention age- and/or sex-standardized growth trajectories) and whether the child was overweight (weight for length ≥95th percentile for children <24 months old and BMI >85th percentile for children ≥24 months old).

Descriptive statistics were used to characterize participants overall and by homeless status by using χ² testing and analysis of variance. Separate multivariable logistic regression models were fit to determine the association between the timing and duration of homelessness and child health and development. All timing models were adjusted for survey year and site; mother’s birthplace, race and/or ethnicity, education, employment status, and marital status; and child’s age (<12 months or ≥12 months), birth weight, and gestational age. In the secondary analysis, interaction models between age group (<12 months or ≥12 months) and duration of homelessness (≥6 months or <6 months) were conducted, and the duration of homelessness was then stratified by child age and adjusted for site; mother’s birthplace, race and/or ethnicity, education, employment status, and marital status; breastfeeding; and low birth weight. Propensity score matching was performed as a confirmatory analysis of the primary models to further control for potential confounding. Propensity scores were developed to account for factors that are associated with homelessness timing by creating a random sample of never-homeless families that had the same propensity score categories as the primary exposure: prenatal homelessness, postnatal homelessness, or both. Two-tailed tests and an α level of .05 were used for all tests. All analyses were performed by using SAS version 9.3 (SAS Institute, Inc, Cary, NC).

Of the 20 571 caregivers interviewed, 661 (3.2%) reported homelessness prenatally only, 758 (3.7%) reported homelessness postnatally only, and 729 (3.5%) reported homelessness both prenatally and postnatally (hereafter referred to as both). The remainder (18 423; 89.6%) were never homeless (hereafter referred to as never), as indicated in Table 1. Caregivers in the prenatally and postnatally homeless group had the lowest prevalence of being married and/or partnered (19.9%) compared with the prenatally homeless–only (22.9%), postnatally homeless–only (26.3%), and never-homeless (39.8%) groups. Slightly more than one-third of the sample was Hispanic (31.4%), approximately half was non-Hispanic African American (50.9%), nearly one-sixth was non-Hispanic white (14.2%), and 3.5% was other. The prenatally homeless–only, postnatally homeless–only, and prenatally and postnatally homeless groups are all compared with the never-homeless group.

Children in the both group had higher odds of the following adverse health outcomes than children in the never-homeless group: history of postnatal hospitalizations (adjusted odds ratio [aOR] 1.41; confidence interval [CI] 1.18–1.69; P < .001), fair or poor child health (aOR 1.97; CI 1.58–2.47; P < .001), and developmental risk (aOR 1.48; CI 1.16–1.89; P < .01). There were no statistically significant differences for risk of underweight (aOR 0.95; CI 0.76–1.18; P = .62) or overweight status (aOR 1.07; CI 0.84–1.37; P = .57) compared with children in the never-homeless group (Table 2).

Children in the prenatally homeless–only group were significantly more likely to have been hospitalized since birth (aOR 1.20; CI 1.00–1.45; P = .05) and be in fair or poor health (aOR 1.35; CI 1.06–1.74; P = .02). There were no statistically significant differences in developmental risk (aOR 1.12, CI 0.85–1.48; P = .42) compared with children in the never-homeless group. There were no statistically significant differences in risk of underweight (aOR 0.92; CI 0.73–1.15; P = .45) or overweight status (aOR 0.93; CI 0.71–1.23; P = .62) compared with children in the never-homeless group.

Children in the postnatally homeless–only group were also significantly more likely to have been hospitalized since birth (aOR 1.22; CI 1.03–1.45; P = .03), be in fair or poor health (aOR 1.37; CI 1.09–1.73; P = .01), and be at developmental risk (aOR 1.34; CI 1.07–1.69; P = .01) compared with children who were never homeless. There were no statistically significant differences in risk of underweight (aOR 0.83; CI 0.65–1.05; P = .12) or overweight status (aOR 1.09; CI 0.86–1.39; P = .49) compared with children in the never-homeless group. Confirmatory propensity score analysis yielded similar results to the primary analysis (data not shown).

A significant interaction (P < .05) was found between child age and the duration of homelessness for fair or poor health, developmental risk, and overweight status. Age-stratified analysis revealed that infants (≤12 months old) and toddlers (>12 months old) who experienced homelessness for >6 months in the postnatal period had higher odds of adverse health outcomes than infants and toddlers who were never homeless.

Infants who had been homeless for >6 months versus those who had never been homeless were significantly more likely to have a history of hospitalizations (aOR 1.51; CI 1.03–2.22; P = .04), be in fair or poor health (aOR 3.13; CI 2.05–4.79; P < .001), be at developmental risk (aOR 1.81; CI 1.11–2.94; P = .02), and have overweight status (aOR 1.74; CI 1.12–2.70; P < .01), as shown in Table 3. There were no statistically significant differences in being at risk for underweight status (aOR 0.90; CI 0.57–1.42; P = .64) compared with infants who were never homeless.

Toddlers who were homeless for >6 months were significantly more likely to have a history of hospitalizations (aOR 1.47; CI 1.15–1.87; P = .002) and to be in fair or poor health (aOR 1.89; CI 1.38–2.58; P < .001). There were no statistically significant differences in developmental risk (aOR 1.25; CI 0.93–1.68; P = .14) or being at risk for overweight (aOR 1.04; CI 0.74–1.46; P = .83) or underweight status (aOR 1.14; CI 0.83–1.54; P = .14) compared with toddlers who were never homeless.

In comparison, infants who had been homeless for <6 months were significantly more likely to be in fair or poor health; however, we found no significant association with history of hospitalizations, developmental risk, or being at risk for underweight or overweight status. Toddlers who had been homeless for <6 months were significantly more likely to be at developmental risk and at risk for underweight status; however, we found no significant association with history of hospitalizations, fair or poor health, or overweight status.

Previous researchers have found that there is an independent relationship between homelessness in childhood and poor health status.^17,18 Prenatal homelessness negatively affects birth outcomes and can be used to predict poor child health outcomes.^16,19,–22 Crowding, which is associated with low birth weight, is commonly found in shelters and short-term living situations and has been associated with increased food insecurity, which is a known child health risk. Limited social capital, which often precedes homelessness, may be associated with reduced material resources as well as a lack of emotional support that is essential to maternal self-care.^14,16,23 Women who are homeless tend to report higher levels of maternal stress, depression, and anxiety than women with housing; in turn, each of these conditions is related to hypertension, increased cortisol, and a higher risk of low birth weight and prematurity.^24,25 

The data reported here reveal 2 specific harmful associations of homelessness with child health and development: timing and duration. This reveals that both the developmental timing and postnatal duration of homelessness may compound the risk of negative child health outcomes. Thus, there is no safe time during early development for experiencing homelessness.

We acknowledge several limitations. First, the retrospective design reveals associations between homelessness and children’s health; caution needs to be exercised with assuming causality because of the potential for unmeasured confounding. Other unmeasured covariates may potentially alter the adverse child health outcomes described in this study, although confirmatory propensity score matching revealed similar results. Additionally, with the exception of child weight and height, all data were self-reported, including the caregivers’ disclosure of experiencing homelessness, which may have introduced reporting bias and shared-method variance. Data on the type and quality of living arrangements of those who were not identified as homeless before pregnancy, during pregnancy, or since the child’s birth were not available.

The findings of this study have clinical and public policy implications. Experiencing homelessness early in development, whether before or after birth, may contribute to a variety of negative outcomes for children’s health and well-being.¹² Greater health care use because of poor health status, increased hospitalizations, and the need for developmental interventions creates substantial family costs associated with caregiver, household, and economic stresses and societal costs, including increased health care expenses and loss of human capital.^26,27 

Interventions that are used to prevent family homelessness can be extremely effective from both public health and child development perspectives. By responding to the upstream needs of pregnant women who are at risk for homelessness, the risk of poor birth outcomes and later poor child health and development outcomes may be reduced. Practitioners should assess patients for housing instability, documenting its prevalence and associated health risks to advocate for more resources and drive innovations in addressing housing stability as a clinically important social determinant of health.¹³ Potential interventions include early identification of pregnancy and access to appropriate nutrition and medical care as well as targeting homelessness prevention services for at-risk families.^10,28 Additionally, hospitals and health systems have created permanent, supportive housing initiatives with the aim of reducing health care use among the chronically homeless.²⁹ For example, the Camden Coalition of Healthcare Providers in New Jersey and Hennepin County Health Center in Minnesota use housing vouchers to reduce health care costs; UnitedHealth Group has invested in new housing across the country; Bon Secours Health System in Baltimore, Maryland, and Nationwide Children’s Hospital in Columbus, Ohio, have built affordable housing units; and Boston Medical Center in Boston, Massachusetts, has invested in a variety of affordable housing projects throughout Boston.^30,31 Interventions that are used to rapidly house children and families are associated with reduced risk of poor birth and child health outcomes.³² In rapid rehousing programs, homeless families are relocated into permanent housing as soon as possible, and this acts as a preventive treatment to mitigate the health risks of prolonged child homelessness, resulting in smaller expenditures than other interventions and fewer reversions to homelessness, thus reducing homelessness levels overall.³³ Given the adverse impact of prolonged exposure to homelessness during early development, minimizing the duration of homelessness by making affordable housing more available to women and children, in addition to rapid rehousing that provides long-term subsidized housing, is a critical and likely cost-effective treatment to promote child health and well-being.

These new research findings can be used to refine our understanding of associations between homelessness experienced during early childhood and poor health and development among young children. The earlier and longer in development a child experiences homelessness may have a larger cumulative toll of poor health and development outcomes, which in turn can have long-lasting effects on the child, family, and community. This new medical knowledge creates an urgent rationale for policy makers to develop ways to disrupt homelessness as a source of toxic stress and effectively promote the healthy development of young children.³⁴ 

aOR

adjusted odds ratio

CI

confidence interval

PEDS

Parents’ Evaluation of Developmental Status