Community Mental Health Journal

, Volume 42, Issue 5, pp 437–448

The Mental Health of Children Exposed to Maternal Mental Illness and Homelessness

  • Ilan Harpaz-Rotem
  • Robert A. Rosenheck
  • Rani Desai

DOI: 10.1007/s10597-005-9013-8

Cite this article as:
Harpaz-Rotem, I., Rosenheck, R.A. & Desai, R. Community Ment Health J (2006) 42: 437. doi:10.1007/s10597-005-9013-8


In recent years a number of reports have documented an increase in the number of homeless families in the US. Using a sample of 195 mothers who were veterans of the US armed forces we assessed the association of maternal homelessness and clinical status, with measures of children’s mental health, school enrolment and attendance. Although maternal homelessness had no significant association with children’s reported emotional problems it had a profound effect on school enrollment and attendance. Mothers’ mental health status, history of incarceration and cumulative history of trauma, as well as children’s exposure to trauma and their self-esteem were the factors most strongly associated with measures of children’s emotional problems. The study suggests that preventive interventions are needed to minimize the effects of exposure to trauma, both in the community and at home, as well as family interventions to address both the mother’s and child’s emotional and physical needs. A longitudinal study is needed to understand better the association between residential instability and children’s mental health.


child adolescence homelessness mental health. 


The current study examines the relationship between mothers’ psychiatric symptoms, homelessness, and their children’s mental health. More specifically, it examines factors correlated with psychiatric symptoms and distress among children who are exposed to the environmental risk factors of maternal homelessness and psychiatric symptoms.

Understanding risk factors associated with psychiatric symptoms among children exposed to homelessness is of great concern since research has indicated that during the past decade homeless families have been the most rapidly expanding subgroup of homeless Americans (Institute of Medicine, 1988; Wright, 1991, 1993). A survey of 25 US cities in 2000 found that families with children constituted 36% of the homeless population (U.S. Conference of Mayors, 2000). The survey also confirmed concerns about the growing number of families with children requesting emergency shelter. Among the population of homeless mothers who are caring for their children, women exposed to domestic violence have been the largest subgroup (Vissing, 1996).

It has further been documented that homeless children, compared to children in the general population, experience a variety of developmental problems including delays in expressive language, visual motor skills, and reading skills (Bassuk, Rubin, & Lauriat, 1986; Bassuk & Rosenberg, 1990; Finkelstein & Parker, 1993; Fox, Barrnett, Davies, & Bird, 1990; Molnar, Rath, Klein, Lowe, & Hartmann, 1991; Parker et al., 1991; Rescorla, Parker, & Stolley, 1991). This is not surprising, since the lack of residential stability has been found to be strongly associated with missed days of school (Bassuk & Rosenberg, 1988). Research on children of homeless families has also found that as many as 38% report behavioral and emotional disorders of clinical significance (Buckner & Bassuk, 1997b; Drennan & Stearn, 1986; Fox et al., 1990).

More recent studies have questioned the direct effect of homelessness on children’s mental health. One study of preschool children found that neither homelessness nor residential instability were strong predictors of poor psychological outcomes but rather a history of physical abuse, poor parenting practices and specific life stressors were all strongly associated with adverse externalizing behavior measured by the Child Behavior Checklist (Bassuk, et al., 1997). However, Buckner, Bassuk, Weinreb, and Brooks (1999) found that among school-age children, housing status was associated only with internalizing behavior problems but not with externalizing behaviors on the Child Behavior Checklist.

In 2001 the Department of Veterans Affairs (VA) established 11 specialized programs across the country designed to provide comprehensive medical and mental health care to homeless female veterans. As part of the evaluation of that program, veterans were invited to participate in an outcome study, and those who consented were interviewed in depth at the time of entry into the program, and every three months thereafter for up to a year. The current study utilizes data from baseline interviews conducted with the sub-sample of women who had minor children. We examine risk factors for increased psychiatric symptoms in children of homeless mothers, as well as for problems with school enrollment and attendance.



A total of 582 women entered the evaluation at 11 VA sites nationwide. Of these, our study sample consisted of 195 women (33%) who reported having at least one child under the age of 18, of whom 47.2% were African American, followed by 25.1 % Caucasian, 9.2% Hispanic and 16.9% of other race/ethnicities. About half (49.3%) of the mothers had only one child and the mean number of children was 1.61 (SD=.94), with a maximum of five children. Mothers’ mean age was 40.21 (SD=6.97) and 20% were married, 17.9% were single and 62.1% were separated or divorced. On average, respondents reported to have been exposed to 9.15 (SD=3.14) traumatic events during the course of their lives.

At the time of the baseline interview 14.7% of the sample was housed, 24% lived in a residential facility, 34.6% were categorized as at risk for homelessness, as they lived in temporary housing arrangements with friends or relatives, and 26.7% were literally homeless. Current housing status was determined by where mothers had primarily resided in the previous 90 days. The average cumulative time spent homeless as an adult was 18.03 months (SD=24.26) and the mean monthly income was $663.37 (SD=778.86).

At the time of assessment 35.9% of the children lived with their mothers, 30.3% with their fathers, 25.6% with other family members, 3.1% in foster care, .5% in institutions and 4.1% in other settings. Altogether, 45% of the children were reported to have been exposed to violence either in their homes or communities, and 19% were directly victims of violence and/or abuse. Only 66.2% of the children ages 5 and older were enrolled in school.


Both child and maternal data were collected from the mothers by their clinicians through a face-to-face interview. Mothers reported on their own age, race, marital status, income, current living arrangement (literally homeless, doubled up with friends/family, institution, or housed), total days homeless in the previous 30 days, social support, as measured as the number of people they could rely on for instrumental or emotional support (Vaux & Athanassopulou, 1987), and history of incarceration (yes/no). They also documented psychiatric risk factors which included history of trauma, as measured by the total number of lifetime traumatic events (range 0–17); emotional problems as measured by the Symptoms Checklist-Revised (SCL-90-R; Derogatis, 1994) a measure that assesses the degree of distress experienced in the past month on 30 psychiatric symptoms (using a five-point Likert scale); and drug and alcohol use/abuse as measured by the Addiction Severity Index (ASI; McLellan, Luborsky, Woody, & O’Brien, 1980) a composite measure that incorporates questions on both use and abuse and ranges from 0, meaning no abuse/dependence evident, to 1, indicating extremely severe abuse/dependence.

Mothers also reported on their youngest child’s age; school enrollment (yes/no); typical number of days of school missed in a month; and the child’s exposure to violence in the home or community (yes/no either as witness or victim). The primary dependent variable of interest was the presence of reported psychiatric symptoms among the children. Mothers completed the Emotional Problems and Self-Esteem sub-scales of the Children’s Problems Checklist (CPC; Schinka, 1985) for their youngest child. This instrument presents a series of descriptions of child behavior and asks the mother to indicate (yes/no) whether the description applies to her child. The Emotional Problems sub-scale includes 32 descriptive statements, while the Self-Esteem sub-scale includes 18 descriptive statements. Our self-esteem measure consisted of the sum of positive responses to the self-esteem sub-scale items, so that higher scores indicated lower self-esteem. We calculated the Cronbach’s Alpha (internal consistency) for both the emotional problems sub-scale and the self-esteem sub-scale of the CPC. The emotional problems sub-scale Alpha was .89 and the self-esteem sub-scale Alpha was .84, indicating that the sub-scales had strong internal consistency.

We conducted an Exploratory Factor Analysis (EFA) with Varimax rotation of the CPC Emotional Problems Scale in an attempt to identify independent symptom factors. The EFA identified four major factors that characterized the children in our sample. We have labeled these: (1) Depression/Anxiety; (2) Clinging Behavior/Separation Anxiety; (3) Obsessive Traits, and (4) Emotional Withdrawal. Items with high loading on more than one component or low loading on all components were excluded because they had no discriminative value between factors. Subsequent analyses also examined the total emotional scale (the sum of all 32 items), and each of the 4 independent factor sub-scales, which consisted of 18 items.

Data Analyses

Initially we conducted bivariate linear regressions to identify the independent variables associated with children’s total emotional problems as measured by the CPC and later included these variables in a multivariable linear regression to model the children’s total emotional problem score. The best model fit was identified by conducting a backward stepwise analysis where items which did not contribute significantly to the explained variance of the model were removed. The same procedure was also used to assess the association between the independent variables and the four factors identified in the exploratory factor analysis.

Among school-age children, we were interested both in their school enrolment and, among those enrolled in school, their attendance. Multivariable logistic regression was used to model the likelihood of school enrolment and multivariable linear regression was used to assess the association between the independent variables and the number of days absent from school. The total CPC Emotional Problems score was included as an independent variable in this model. Again, a backward stepwise model was used.


Children’s Emotional Problems

In the analysis of the measure of child’s total emotional problems, the backward multivariate regression analysis showed that the mother’s housing status was not significantly associated with children’s emotional problems. Major significant risk factors for poorer child mental health included the mother’s own emotional problems as measured by higher scores on the SCL β=1.57, SE=.46, p<.01), and a history of maternal incarceration (β=2.60, SE=.81, p<.01). In addition, children of married mothers were reported to suffer from fewer emotional problems than those of unmarried or separated mothers (β=−3.55, SE=1.43, p<.05).

This analysis also showed that children who were living with their mothers experienced more emotional problems than children who were living in other settings (β=1.73, SE=.80, p<.05). Mother’s current homelessness status did not add significantly to the explained variance of the model. Children who were reported to experience low self-esteem or who witnessed violence either in the home or the community also experienced a higher degree of emotional problems (β=1.14, SE=.11, p<.01; β=1.79, SE=.78, p<.01 respectively). The adjusted R2 for the best fitting model assessing children’s total emotional problems was .55.

Next, using backward multivariate analysis to achieve the best fitting model we examined correlates of the four symptom sub-scales identified by the exploratory factor analysis using independent variables that were significantly associated with these scales in the bivariate analysis. The children’s depression–anxiety symptoms factor was found to be positively associated with (1) mother’s score on the SCL (β=.22, SE=.08, p<.01), such that the more psychiatric symptoms the mother experienced the more depressed/anxious was her child, (2) with mother’s history of incarceration (β=.31, SE=.14, p<.05) and (3) with mother’s cumulative exposure to trauma (β=.05, SE=.02, p<.05). Again, mother’s housing status did not contribute additional explained variance to the model. Children of Hispanic mothers were less likely to report depression symptoms among their children than other ethnic groups (β=−.46, SE=.20, p<.05). Both children’s poor self-esteem (β=.15, SE=.02, p<.01) and being a victim of violence were found to be significantly associated with symptoms of depression and anxiety (β=.05, SE=.02, p<.05). The adjusted R2 for the best fitting model assessing children’s depression–anxiety symptoms was .47.

The clinging behavior/separation anxiety factor was not associated with any of the mothers’ characteristics but was significantly associated with both children’s self-esteem and age, where both younger children and children with lower self-esteem were reporting more clinging behavior (β=−.03, SE=.01, p<.05; β=.11, SE=.02, p<.01 respectively). Children who lived with their fathers were reported to exhibit fewer symptoms associated with clinging behavior when compared to children living in other settings (β=−.37, SE=.16, p<.05). The adjusted R2 for the best fitting model assessing children’s clinging behavior/separation anxiety was .17.

No risk factors were found to be significantly associated with the obsessive traits factor (factor 3) or the emotional withdrawal factor (factor 4).

School Enrolment and Attendance

Using independent variables that were significantly associated with school enrolment in the bivariate logistic regression, backward multivariable logistic regression indicated that children of homeless mothers, regardless of their current custodial status, were significantly less likely to be enrolled in school (OR=.23, CL(95%)=.06–.82, p<.05). Children who resided with their father (OR=6.15, CL(95%)=1.51–25.07, p<.05) or with other family members (OR=4.26, CL(95%)=1.01–17.96, p<.05) were more likely to be enrolled in school than those living with their mothers, in foster care or in other institutions. Children who were reported to have more intense emotional problems were also more likely to be enrolled in school (OR=1.14, CL(95%)=1.03–1.27, p<.05). The adjusted R2 for the best fitting model assessing school enrolment was .27.

Finally, using backward multivariable linear regression model we evaluated, utilizing the sub-sample of those children enrolled in school (n=130), the average number of days that they had missed school in a typical month. Again, only those variables that were significantly associated with school attendance in a bivariate logistic regression were entered into the model. Children of homeless mothers (β=6.93, SE=1.70, p<.01) and married mothers (β=4.68, SE=1.68, p<.01) were reported to have missed significantly more days of school. The analysis also indicated that children living in foster care were reported to miss significantly fewer days of school than other children (β=−8.05, SE=3.68, p<.05). Finally, children suffering from more intense emotional problems on the CPC missed significantly more days of school (β=.13, SE=.05, p<.01). The school enrolment best fitting model’s adjusted R2 was .23.


Relatively little is known about exactly how parental homelessness and poverty affect children’s mental health (see: Center for Mental Health Services, Annual Report to Congress on the Evaluation of the Comprehensive Community Mental Health Services for Children and their Families Program, 1999 for more details). However, family factors are known to be strong predictors of children’s well-being. During World-War II Dorothy Burlingham and Anna Freud (1949) reported that the effects of war on children’s well-being were negligible compared to family factors. This study augments the literature on the health status of children of homeless mothers, finding that while mothers’ housing status was not a primary risk factor for emotional problems, maternal homelessness and residential instability had an adverse effect on children’s school enrolment and attendance, regardless of whether they reside with their mothers.

In addition, there was a significant risk to children posed by residing with their mother compared to residing in other settings (i.e., father, other family members or foster care). These results add to other studies which have reported that residential instability did not predict behavioral or emotional problems among school-age children as well as findings that parental distress and recent adverse life events other than homelessness, were associated with less well-being among children (Buckner et al., 1999; Masten, Miliotis, Graham-Bermann, Ramirez, & Neemann, 1993). Our study found that independently of the children’s current custody arrangements, children of mothers who exhibited elevated emotional problems or had a history of incarceration were at a greater risk for developing emotional problems. However, children who did reside with their mother exhibited greater emotional problems than other children, regardless of mother’s current housing status or history of homelessness. Children of married mothers experienced significantly less emotional problems compared to children of separated or single mothers, suggesting that an intact family may provide some protection against overall emotional problems, even in the context of homelessness.

When we examined maternal characteristics associated with symptoms of depression and anxiety in their children, we again found that mother’s housing status was not a significant independent risk factor. In addition to maternal emotional problems and past incarceration, mothers’ cumulative trauma was significantly associated with the symptoms of depression and anxiety. The fact that Hispanic mothers reported lower rates of depression and anxiety symptoms among their children may be related to cultural differences in reporting style or child rearing practices (Wasserman, Brunelli, Rauh, & Alvarado, 1994).

Children who witnessed, but who were not necessarily subjected to, violence (either in their homes or communities) were reported to have more emotional problems than children who were not exposed to such events. However, when examining specific symptoms of depression and anxiety the study found that children who were reported to be victims of abuse had a significantly higher depression score. In all cases, children’s low self-esteem was significantly associated with more severe emotional problems.

Similar to the findings of Buckner et al. in a low income school age population, this study suggests that traumatic life events such as exposure to violence and abuse, and mother’s incarceration are strongly associated with poor emotional status among children (Buckner & Bassuk, 1997a). Furthermore, it highlights the link between mothers’ psychiatric problems and their children’s psychiatric condition, independently of their mothers’ current housing status or whether or not they reside with the mother or in another setting. When examining school enrolment and attendance among school-age children, children who either were under the care of a homeless mother or a mother who was at risk of becoming homeless were less likely to be enrolled in school. Surprisingly, children who experienced more emotional symptoms were more likely to be enrolled in school. It is possible that parents of children with emotional problems find schools to be a source of help in caring for them. It has been documented that the vast majority of children who receive any mental health care do so through the educational system (Burns et al., 1995). Alternatively it is also possible that school imposes greater emotional challenges to children of distressed mothers and thus enrolled children are reported to have more psychiatric symptoms.

Children of homeless mothers also had a greater number of days absent from school. For reasons that are not clear, children of married mothers also had more absentee days compared to single or separated mothers. We can only speculate here that the burden of care may be higher on single mothers and therefore, they may do their utmost to send their children to school in order to reduce this burden (see also Kurz, 1997). It is also possible that single mothers are less sensitive to their child’s distress and may send them to school whether or not the child is overwhelmed, as it has been documented that single mothers respond less positively to their children than married mothers (Rosenkrantz & Huston, 2004).

These findings have several clinical implications. First, they suggest that community providers serving homeless populations should be concerned not only with the physical and mental health of homeless women, but with the health and well-being of their children, even when those children are not literally homeless, or even when children are not currently residing with their mothers. It also suggests that it is important for homelessness providers to link their services to local school systems, since the major adverse effect of maternal homelessness was on children’s school enrolment and attendance. Finally, these data suggest that when considering children’s well being, service providers should not exclusively be concerned with their housing status, but also with reducing stressors such as abuse and violence, as homeless children are more likely to be exposed to such events (Buckner & Bassuk, 1997a).

Our findings demonstrate the need for family preventive interventions aimed at minimizing the effect of exposure to traumatic experiences—either in the community or at home. Family interventions that address both the mother’s and child’s emotional and physical needs may prove to be useful in reducing emotional problems. The federally sponsored Comprehensive Community Mental Health Services for Children and their Families Program (System of Care) initiated in 1993 advocates for providing services to children with serious mental health disorders directly in their homes and communities, and may provide funding for the establishment of such interventions (President’s New Freedom Commission on Mental Health: Achieving the promise: Transforming mental health care in America. Final report, 2003). As suggested by other researchers, interventions targeting homeless and low-income families should also include a service component that would promote children’s self-esteem and competence (Bassuk, Weinreb, Dawson, Perloff, & Buckner, 1997; Weissberg, Caplan, & Harwood, 1991).

This study has several limitations. First, since the study sample is comprised of homeless female veterans who suffer from a wide range of psychiatric symptoms, the generalizeability of these findings to other populations is uncertain. Second, since the study makes use of maternal self report measures it is possible that mothers who reported more psychiatric symptoms were also more likely to endorse such symptoms among their children, thereby inflating the association between mothers’ and children’s mental health problems. It is also possible that children who are being raised by their father are reported by mothers to show less clinging behavior than other children only because mothers failed to observe this behavior. However, mothers reported no significant difference in clinging behavior between children residing with them and children residing in other settings. Third, in this study we limited mothers’ report to their younger child who we judged to be in the greatest need for maternal attendance. This restriction may have biased the results in ways that are unclear.

Fourth, the CPC, which was used to record children’s psychiatric symptoms, has high face validity but lacks the necessary information on other psychometric properties such as reliability. Finally, the fact that only 36% of the children in this study were living with their mothers during the evaluation may affect the statistical power to observe associations between current maternal homelessness and children’s mental health. Nevertheless, the study indicated that whether or not the children reside with their mothers, maternal emotional factors have a profound effect on their children’s wellbeing. Further research is needed that independently assesses child mental health status among homeless families, without relying on parental report.


The findings reported by this study suggest that maternal homelessness per se may not be a significant risk factor for emotional problems in children of homeless mothers. However, other risk factors correlated with homelessness, such as maternal mental health, history of incarceration, and exposure to violence and/or abuse have a noticeable negative adverse effect on their children’s mental health. These results are consistent with findings among children of domiciled mothers, but may be exacerbated in a setting that includes maternal residential instability and homelessness.

Copyright information

© Springer Science+Business Media, Inc. 2006

Authors and Affiliations

  • Ilan Harpaz-Rotem
    • 1
  • Robert A. Rosenheck
  • Rani Desai
  1. 1.Department of Psychiatry, School of MedicineYale University, VA Healthcare SystemWest HavenUSA

Personalised recommendations