Prevalence, Characteristics, and Associated Health and Health Care of Family Homelessness Among Fifth-Grade Students

Abstract

Objectives. We describe the lifetime prevalence and associated health-related concerns of family homelessness among fifth-grade students.

Methods. We used a population-based, cross-sectional survey of 5147 fifth-grade students in 3 US cities to analyze parent-reported measures of family homelessness, child health status, health care access and use, and emotional, developmental, and behavioral health and child-reported measures of health-related quality of life and exposure to violence.

Results. Seven percent of parents reported that they and their child had experienced homelessness (i.e., staying in shelters, cars, or on the street). Black children and children in the poorest families had the highest prevalence of homelessness (11%). In adjusted analyses, most general health measures were similar for children who had and had not been homeless. Children who had ever experienced homelessness were more likely to have an emotional, behavioral, or developmental problem (odds ratio [OR] = 1.7; 95% confidence interval [CI] = 1.1, 2.6; P = .01), to have received mental health care (OR = 2.2; 95% CI = 1.6, 3.2; P < .001), and to have witnessed serious violence with a knife (OR = 1.6; 95% CI = 1.1, 2.3; P = .007) than were children who were never homeless.

Conclusions. Family homelessness affects a substantial minority of fifth-grade children and may have an impact on their emotional, developmental, and behavioral health.

An estimated 23% of homeless persons in the United States are children younger than 18 years.¹ Of individuals who have experienced an episode of homelessness and used a homelessness assistance program over the course of a year, an estimated 38% are children living with parents.² Families defined as homeless may be literally homeless (spending the night in emergency shelters, abandoned buildings, cars, on the street, and so on) or precariously housed (at imminent risk of literal homelessness).^3,4 Data available on the prevalence of literal homelessness (henceforth referred to as “homelessness”) are generally based on emergency shelter use, requests for emergency shelter use, and nighttime counts of those living on the streets.⁵ The prevalence of homelessness in US cities ranges from a 1-day prevalence of 0.3% to a 3-year prevalence of 3% (prevalence rates increase as the period of time increases because of rapid turnover among the homeless).^4,6–8

Population-based studies have estimated the prevalence of episodic homelessness among adults.^9–11 One random-digit-dialing survey of US adults found a 7% lifetime prevalence of homelessness.⁹ A Philadelphia population-based study linked emergency shelter intake records to all adult women with a recorded live birth over a 2-year period and reported 11% homelessness between 3 years before and 4 years after the birth.¹⁰ Unlike studies of currently homeless people, these population-based estimates include individuals who have a history of homelessness.^6,8 There are no population-based data that estimate the lifetime prevalence of family homelessness among children.

Studies of children in sheltered, homeless families have shown that many have poor health status, high rates of asthma, high rates of emergency department visits, delays in obtaining preventive care, and high rates of emotional, developmental, and behavioral problems.^7,12–17 In most studies, health and health-related outcomes were worse for children in sheltered, homeless families compared with children in the general population. In most studies, homeless children also had worse outcomes compared with low-income housed children, although in a few studies the 2 groups had similar outcomes.¹⁸ In a study of 82 families using shelters and transitional housing in King County, Washington, 59% of children had no usual health care provider; these children experienced rates of emergency department visits higher than that of the national rate for children.¹³ In another study of 220 homeless families in shelters in Worcester, Massachusetts, homeless children were more likely to be reported in poor to fair health and made more frequent emergency department and outpatient visits than did low-income housed children.¹² Other studies have found high rates of developmental delay and emotional or behavioral problems among sheltered children.^13,14,19,20

Most data on the health status and needs of homeless children compare currently homeless children and their families residing in emergency shelters with low-income housed children or the general population.^12–14 Many poor families may experience 1 or more episodes of homelessness over a period of time, but most studies have only examined currently homeless children.^3,21 No published studies to date have described the lifetime prevalence of homelessness and the associated health and health-related problems of a school-aged, nonsheltered, population-based sample of children. We aimed to describe (1) the lifetime prevalence and characteristics of family homelessness among fifth-grade students and (2) the health and health-related concerns of these students.

METHODS

We analyzed data from Healthy Passages, a Centers for Disease Control and Prevention (CDC)–funded, multisite study of 5147 fifth-grade students that focused on risk factors, protective factors, health behaviors, and health outcomes.²² Interviews for the study were conducted from 2004 through 2006.

Sample

The study population consisted of fifth-grade students enrolled in regular classrooms in public schools with a minimum enrollment of 25 fifth-grade students. The eligible schools were located in 3 areas: (1) 10 contiguous public school districts in and around Birmingham, Alabama; (2) 25 contiguous public school districts in Los Angeles County, California; and (3) the largest public school district in Houston, Texas. This study population represented over 99% of all fifth-grade students in regular public school classrooms in the study districts. We selected a random sample of schools, using probabilities that were a function of how closely a school's racial/ethnic mix corresponded to the site targets for race/ethnicity. This sampling procedure (described in detail elsewhere²²) was used to ensure adequate sample sizes of Hispanic, non-Hispanic Black, and non-Hispanic White children.

A total of 118 schools with 11 532 enrolled fifth-grade students were selected with this procedure. Each student's primary caregiver or parent (henceforth referred to as “parent”) received a letter requesting permission to be contacted by study personnel. Of the 11 532 parents, 6663 who either agreed to be contacted or who were unsure were invited to participate in the study; 77% of them (5147) completed an interview. Interviews were conducted at the parent's home, the study center, or another location chosen by the parent. Prior to the interview, parents gave informed consent for their participation and their child's participation, and the child also gave assent.

Measures

Each parent–child dyad completed a computer-assisted personal interview and an audio computer-assisted self-interview in English or Spanish. The computer-assisted personal interview is a form of personal interviewing that allows the respondent to answer questions directly into a computer, aided by an interviewer. The audio computer-assisted self-interview is a modified form of the computer-assisted personal interview that is used for more sensitive survey items; respondents hear sensitive questions through headphones and privately respond on the computer. To describe the prevalence of homelessness, we focused on literal homelessness, excluding the precariously housed (parents staying with family or friends to avoid becoming literally homeless).^3,9 We excluded this group because these data do not always identify those who are staying with family or friends for other reasons, such as to provide care to an aging parent.⁷

Parent-reported homelessness and demographics.

To assess homelessness, we asked parents, “Have you ever had to stay in one of the following places with your child?” They could choose multiple locations from a list that included emergency shelters, abandoned buildings, public parks, and other nonresidential locations. They also reported the total amount of time they had had to stay in those locations. We also collected sociodemographic data, including parents' age, household income, and level of education, and children's age, race/ethnicity, and insurance status.

Some interviews took place after Hurricane Katrina hit the US Gulf Coast in August 2005. Because over 100 000 Katrina survivors were relocated to Houston, we examined the data for any increase in homelessness after August 29, 2005, that could be attributed to Katrina.

Parent-reported child health and health care measures.

We asked parents to rate their child's current state of health as excellent, very good, good, fair, or poor. Child health status was dichotomized into good health (excellent, very good, and good) and poor health (fair and poor) to examine differences in homelessness between children in good versus poor health. To assess physical health, we asked parents (1) if their child had had a well-child visit in the past 12 months; (2) if their child had had an emergency department visit in the past 12 months; (3) their child's usual location for preventive care; (4) their child's health insurance type, if any; and (5) if their child was injured seriously enough over the past 12 months that he or she was seen by a doctor or a dentist, seen in the emergency department, or admitted to a hospital. For emotional, developmental, and behavioral measures and mental health-related measures, we asked parents if their child (1) had any kind of emotional, developmental, or behavioral problem for which he or she needed or got treatment or (2) had ever received care for emotional, behavioral, or drug or alcohol problems.

Other parent-reported measures.

Parents also completed the Brief Symptom Inventory (BSI-18), a validated tool to identify psychological distress in adults.²³ This 18-item scale contains 3 subscales for depression, anxiety, and somatization; all items are summed and reported as standardized T scores derived from community samples to provide a global severity index. We used published cutoff values validated in previous community samples to identify parents with a level of psychological distress considered to require psychosocial intervention.²⁴

Children's interviews.

Children completed the Pediatric Health-Related Quality of Life Inventory 4.0, a validated tool used to measure health-related quality of life for children aged 2 to 18 years.²⁵ The total score for this 36-item tool measures physical, emotional, social, and school functioning. We dichotomized the total scores to identify children at risk for impaired health-related quality of life, defined previously as a score 1 standard deviation or more below the population mean.^26,27 Finally, each child reported whether, in the past 12 months, he or she (1) had been personally threatened or injured with a knife or a gun or (2) had seen someone else so threatened or injured.

Statistical Methods

All analyses employed design and nonresponse weights and accounted for the variance effects of both weights and the clustering of children within sites using Stata version 10 SE (Stata Corp, College Station, TX).^28–30 We calculated lifetime prevalence of homelessness in the overall sample and used bivariate analyses to compare the characteristics of ever-homeless and never-homeless children. To examine the racial/ethnic differences in homelessness more closely, we used multiple logistic regression, with homelessness as the dependent variable and other characteristics as covariates.

We also compared ever-homeless and never-homeless children on a variety of health measures, using logistic regression for dichotomous outcomes, ordinal logistic regression for ordered outcomes, and multinomial logistic regression for the unordered multicategory outcomes. We used multivariate versions of these tests to compare never-homeless and ever-homeless children while we controlled for covariates. Parent-reported child covariates were age (8–10 years vs 11–14 years; 93% of children were aged 10–11 years), gender, race/ethnicity (Hispanic, non-Hispanic Black, non-Hispanic White, other race/ethnicity), health status, and whether insured. Parent covariates included age, education (less than high school, high school graduate or general equivalency diploma [GED], some college or 2-year degree, and 4-year college graduate), household income, household composition (2-parent household, single-parent household, other composition), presence of psychological distress, and language of parent interview (English, Spanish). In these regression models, health and health-related outcomes were the dependent variables and homelessness was the independent variable.

RESULTS

Of 5147 children, 5024 parents answered the questions on homelessness; 7% had experienced homelessness. The prevalence of homelessness was similar across the 3 cities (Table 1). Prevalence was highest among Black children (11%), children from households with incomes under $20 000 (11%), and children whose parents had an education level of high school completion (11%). More children were ever homeless in single-parent households (10%) than in 2-parent households (5%). Parents of ever-homeless children were more likely to report psychological distress than were parents of never-homeless children (24% vs 17%; P = .01). There were no differences in the prevalence of homelessness before and after Hurricane Katrina, for Houston or the sample as a whole (data not shown).

TABLE 1.

Prevalence and Characteristics of Never-Homeless and Ever-Homeless Fifth-Grade Children: Birmingham, AL, Los Angeles, CA, Houston, TX; Healthy Passages; 2004–2006

	Prevalence of Homelessness, % (No.)	Children Never Homeless, %	Children Ever Homeless, %
Sample size, no.	5024	4662	362
Total	7 (362)
City
Birmingham, AL	6 (113)	32	28
Los Angeles, CA	7 (122)	33	34
Houston, TX	8 (127)	35	39
Gender of primary caregiver
Women	7 (340)	93	93
Men	6 (22)	7	7
Parent's age,** y
18–34	9 (154)	33	44
35–44	6 (154)	48	42
≥ 45	5 (54)	19	14
Child's age,* y
≤ 10	6 (135)	44	36
≥ 11	8 (227)	56	64
Child's gender
Girl	7 (189)	49	50
Boy	7 (173)	51	50
Child's race/ethnicity***
Hispanic	7 (123)	43	43
Non-Hispanic White	2 (32)	23	8
Non-Hispanic Black	11 (182)	28	44
Other	7 (25)	5	5
Annual household income,*** $
< 20 000	11 (192)	34	56
20 000–34 999	8 (82)	23	25
35 000–69 999	5 (47)	21	14
≥ 70 000	2 (28)	23	6
Parent's education level***
Less than high school	9 (110)	29	38
High school graduate or GED	11 (110)	20	31
Some college or 2-y degree	7 (96)	25	23
4-y college graduate or greater	3 (44)	26	9
Household composition***
2-parent household	5 (146)	59	44
Single-parent household	10 (203)	37	53
Other composition (foster, adoptive)	5 (13)	5	3
Parent's interview language
English	7 (272)	70	65
Spanish	8 (90)	30	35
Positive screen for parental psychological distress*
No	6 (276)	83	76
Yes	10 (86)	17	24

Note. GED = general equivalency diploma. Numbers are unweighted and percentages are weighted.

*P = <.05; **P < .01; ***P < .001.

In a more detailed examination of the association between child race/ethnicity and a history of homelessness, we used unadjusted regression analyses to conduct comparisons with respect to a reference category (White children). The unadjusted odds of homelessness for Black children (odds ratio [OR] = 4.7; 95% confidence interval [CI] = 2.7, 8.4; P < .001), Hispanic children (OR = 3.1; 95% CI = 1.7, 5.5; P = .001), and other children (OR = 3.2; 95% CI = 1.6, 6.1; P = .001) were greater than for White children. In adjusted analysis, the odds of homelessness remained greater for Black and other children, but not for Hispanic children when compared with White children (data not shown).

Places and Duration of Homelessness

Ever-homeless families reported having to stay in a variety of nonresidential locations, including shelters and transitional housing. Forty percent of ever-homeless families spent at least 6 months homeless (Table 2). Twenty-eight percent of ever-homeless families also reported having to stay with family or friends, whereas 14% of the total sample reported only having to stay with family or friends, which, by our definition, made them precariously housed, not homeless (data not shown).

TABLE 2.

Place and Total Duration of Literal Homelessness Among Fifth-Grade Children and Their Families (n = 396): Birmingham, AL, Los Angeles, CA, Houston, TX; Healthy Passages; 2004–2006

	Ever-homeless Children, No. (%)
Total time spent homeless
< 1 mo	72 (19)
1 mo to less than 6 mo	41 (11)
6–12 mo	11 (3)
> 1 y	130 (37)
Time not specified	108 (31)
Location of homelessnessa
Emergency shelter	148 (43)
Transitional shelter/housing	107 (30)
Welfare/voucher motel	61 (15)
Vehicle	46 (12)
Anywhere outside	22 (6)
Place of business	15 (3)
Transportation site	13 (4)
Abandoned building	11 (3)
Migrant worker camp	1 (< 1)
Other placeb	34 (9)

Note. Numbers are unweighted and percentages are weighted.

^aProportions for locations do not sum to 100%, because parents could report multiple locations.

^bDoes not include staying with family or friends.

General Health-Related Measures

Table 3 presents differences in general health-related measures between ever-homeless and never-homeless children. In unadjusted regression analyses, ever-homeless children were more likely to have a reported impaired health-related quality of life, to be insured by Medicaid or be uninsured, and to use a clinic or health center as their usual source of preventive care or to have no usual source of care. They were also more likely to have experienced a serious injury in the past 12 months. These differences in insurance type, health-related quality of life, and usual location for preventive care were fully accounted for by covariates in multiple and multinomial logistic regression (data not shown). However, after we controlled for covariates, ever-homeless children were more likely to have had a serious injury in the past 12 months (OR = 1.5; 95% CI = 1.1, 2.1; P = .03).

TABLE 3.

Frequency Distribution and Odds Ratios (ORs) of General Health, Access and Use of Health Care, and Injury History Among Fifth-Grade Children: Birmingham, AL, Los Angeles, CA, Houston, TX; Healthy Passages; 2004–2006

	Never-Homeless Children, %	Ever-Homeless Children, %	Unadjusted ORa (95% CI)
Child's health status
Parent-reported child health status of fair to poor	10	12	1.2 (0.8, 1.9)
At risk for impaired pediatric quality of life	16	21	1.4* (1.1, 1.9)
Health insurance typeb
Private	44	22	Reference outcome
Medicaid	28	49	3.5† (2.5, 4.9)
SCHIP	12	9	1.5 (0.9, 2.4)
Uninsured	13	17	2.6† (1.7, 3.9)
Other	3	4	2.5* (1.3, 4.5)
Usual location of preventive health careb
Doctor's office	70	58	Reference outcome
Clinic, mobile clinic, health center	27	36	1.6† (1.2, 2.2)
Other (school-based clinic, hospital outpatient)	2	3	1.6 (0.8, 3.2)
No usual source of care	2	4	2.3* (1.2, 4.7)
Use of general health care services
Had a well-child visit in past 12 mo	68	71	1.2 (0.9, 1.6)
Had at least 1 emergency department visit in past 12 mo	19	20	1.1 (0.8, 1.5)
Child had serious injury in past 12 mo	14	19	1.5* (1.1, 2.1)

Note. CI = confidence interval; SCHIP = State Children's Health Insurance Program.

^aUnadjusted OR of each outcome for ever-homeless children compared with never-homeless children.

^bMultinomial regression was used; relative risk ratios compared with the base outcome are reported.

*P < .05; †P ≤ .003.

Emotional, Developmental, and Behavioral Problems and Violence Exposure

Children in ever-homeless families experienced more emotional, developmental, and behavioral problems than did those in never-homeless families (Table 4). In bivariate and multivariate analyses, ever-homeless children were more likely than were never-homeless children to have a parent-reported emotional, developmental, or behavioral problem (OR = 1.7; 95% CI = 1.1, 2.6; P = .01), to have ever received care for an emotional, behavioral, or substance abuse problem (OR = 2.2; 95% CI = 1.6, 3.2; P < .001), and to have witnessed serious violence involving a knife (OR = 1.6; 95% CI = 1.1, 2.3; P = .007) or a gun (OR=1.5; 95% CI=1.0, 2.1; P=.03) in the past 12 months.

TABLE 4.

Frequency Distribution and Odds Ratios (ORs) for Emotional, Developmental, and Behavioral Health; Mental Health Care Use; and Exposure to Violence Among Fifth-Grade Children: Birmingham, AL, Los Angeles, CA, Houston, TX; Healthy Passages; 2004–2006

	Never-Homeless Children, %	Ever-Homeless Children, %	Unadjusted OR (95% CI)	Adjusted ORa (95% CI)
Child had emotional, developmental, or behavioral problem	8	14	1.8** (1.2, 2.7)	1.7* (1.1, 2.6)
Child ever received care for an emotional, behavioral, or substance abuse problem	9	14	1.7** (1.2, 2.4)	2.2*** (1.6, 3.2)
Child witnessed serious violence in past 12 mo
Witnessed someone threatened or injured with knife	9	18	2.3*** (1.6, 3.1)	1.6** (1.1, 2.3)
Witnessed someone threatened or injured with gun	11	21	2.1*** (1.5, 3.0)	1.5* (1.0, 2.1)
Child was victim of serious violence in past 12 mo
Was threatened or injured with knife	2	3	2.0 (0.9, 4.6)	1.6 (0.8, 3.5)
Was threatened or injured with gun	1	2	1.4 (0.4, 4.7)	1.1 (0.4, 3.5)

Note. CI = confidence interval.

^aAdjusted OR for each mental health outcome for ever-homeless children compared with never-homeless children; covariates included parent's age, child's age, child's gender, child's race/ethnicity, child's health status, child's insurance status, parent's education level, household income, household composition, parental psychological distress, and language of parent's interview.

*P < .05; **P < .01; ***P ≤ .001.

DISCUSSION

The lifetime prevalence of homelessness among fifth-grade students and their families was 7%. Among Black children and children in the poorest households, the prevalence of homelessness was 11%. After control for sociodemographic factors, indicators of general child health were similar for ever-homeless and never-homeless children, whereas indicators of emotional, developmental, and behavioral health were not. Ever-homeless children were significantly more likely to have a parent-reported emotional, developmental, or behavioral problem and to have received mental health-related services than were never-homeless children. This association, however, should not be interpreted as causal, because we do not know the temporal relationship between the emotional, developmental, and behavioral problems and the episodes of homelessness.

Although there are no other population-based studies that describe the lifetime prevalence of homelessness among children, our 7% prevalence is similar to other related estimates and studies. Researchers at the Urban Institute used estimates of the number of homeless persons using homelessness assistance programs in an average week in 1996 to calculate a series of projections for homelessness. They projected that 9% of low-income children experienced family homelessness over a 12-month period.² Our finding of a 7% lifetime prevalence of homelessness for children is also similar to findings from earlier population-based studies of adult homelessness. Webb et al., in their study of perinatal women in Philadelphia, reported an 11% prevalence of homelessness during the 3 years before and 4 years after the birth of the child,¹⁰ similar to the 9% we found for parents aged 18 to 34 years. That study also showed that Black women experienced more homelessness than did White and Hispanic women.¹⁰

In another study of public shelter use in New York and Philadelphia, less than 1% of White children had spent time in a shelter in the previous 3 years, compared with 7% of Black children.⁸ A random-digit-dialing study of US adults, however, did not find a significant difference in the prevalence of homelessness by race/ethnicity, but it did find that Black adults experienced homelessness of longer duration than did White adults.⁹ This same study found that most adults with a history of homelessness had been homeless for 4 weeks or more during their lifetime, which is similar to our finding that a majority of families reported cumulative homelessness of over 1 month. Although we know the total duration of homelessness experienced by the fifth-grade students with their families, we do not have information on the number of episodes of homelessness experienced or the duration of each.

Although some data suggest that health and health-related outcomes are worse for homeless, sheltered children than for low-income housed children or the general population of US children, to our knowledge, no previous studies have examined this association among a population-based sample of children.^{12–14,19,20} Our study provides the first description of the association between a history of family homelessness and health-related measures among a population-based sample of school-aged children. We found that children with a history of homelessness were more likely to report impaired health-related quality of life; however, this difference was not significant after control for sociodemographic factors. Similarly, differences in usual source of care and insurance were explained by sociodemographics. One possible explanation for the differences between our findings and those of other studies that used convenience samples is that shelter-based convenience samples overrepresent families with longer durations or more-frequent episodes of homelessness.⁹ Children in these families may differ in health and health care use from children in families that use shelters less frequently or who are not currently homeless. These measures may be worse for children during the actual period of homelessness and improve once families secure a place to live.

Even after we controlled for sociodemographic factors, we found that ever-homeless children were more likely to have emotional, developmental, and behavioral problems. Some studies of sheltered, homeless children have described high rates of emotional, developmental, and behavioral problems.^14,15 These findings may be related to the stress that episodes of homelessness place on families, or even to the experience of homelessness (or the conditions that led to it). For instance, we found that ever-homeless children were more likely to have witnessed serious violence and were more likely to have had a serious injury in the past 12 months, even when we controlled for a variety of covariates. The experience of homelessness may itself expose children to stressors that in turn affect their behavioral and emotional development.

It is also possible that the children's emotional, developmental, and behavioral problems are related to their parents' mental health. Findings from studies of currently sheltered mothers vary; some have found no differences in the rates of psychological distress, depression, and posttraumatic stress disorder between homeless mothers and low-income housed mothers, but others have found higher rates of these mental health problems among homeless mothers than among low-income housed mothers or the general female US population.^19,31–33 We found a positive association between parental psychological distress and homelessness; however, when we controlled for parental psychological distress in multivariate analysis, ever-homeless children were still more likely to have emotional, developmental, and behavioral problems than were never-homeless children.

We found that ever-homeless children were more likely to have received care for an emotional, behavioral, or substance abuse problem, which may seem counterintuitive. Although we do not have data to explain this finding, we can hypothesize that children may have increased access to some social and health services as a result of accessing homelessness services. It is also possible that children with a history of homelessness have more severe symptoms, which increases the likelihood of receiving mental health services. Increasing access to mental health care is probably only a part of the solution for these families; addressing the housing instability itself will likely be an important part of reducing this mental health disparity.

There are limitations to this study. First, our sample was limited to 3 sites and may not generalize to the US population. Nevertheless, these sites contain urban, suburban, and rural schools in 3 different US regions. Second, there could be nonresponse bias caused by parental nonparticipation; however, our nonresponse weighting likely addressed much of this concern, and it is unlikely that any remaining bias overestimated the prevalence of homelessness. Third, we do not know if the episodes of homelessness occurred before, after, or during many of our outcomes of interest. For example, although ever-homeless children were more likely to have experienced a serious injury in the previous 12 months, the injury may not have coincided with the episodes of homelessness. Fourth, we did not count precariously housed families as homeless. By only considering literal homelessness, we may be underestimating the true prevalence of homelessness that would be captured by the broader definition.

Our findings have important clinical and policy implications. First, we found that a small but significant proportion of fifth-grade students experienced homelessness with their families, with a prevalence as high as 11% in some subpopulations. Our results suggest that in a school of 500 students, 35 of them would have a history of family homelessness, and in a classroom of 28 students, 2 students would have been homeless at some point in their lives. It is critical that health professionals who care for children consider the living environments and limited resources of families that may affect numerous aspects of care, including prescription and over-the-counter medication choice, timing and location for follow-up visits, and parents' ability to adhere to recommendations.³⁴ In addition, pediatricians and other health professionals may need to have some knowledge of local resources for families who are homeless or precariously housed.

Second, we found a positive association between a history of family homelessness and emotional, developmental, and behavioral problems. The experience of homelessness may contribute to emotional, developmental, and behavioral problems among school-aged children. In addition to ensuring access to appropriate health services, efforts to help families maintain stable living arrangements may be an important part of providing comprehensive primary care. Finally, a majority of ever-homeless families in our study relied on emergency shelters and transitional housing. Many major urban areas, however, may not have the capacity to provide shelter for them. In 2006, emergency shelters in 20 of 23 surveyed major US cities reported having to turn away homeless families because of a lack of resources.³⁵ Policies on homelessness should ensure that states have the necessary funds to meet the emergency housing needs of poor families.