Background Child homelessness has been associated with elevated mental health problems during early adolescence, a period of onset for psychiatric problems. Prior literature has relied on cross-sectional studies, limiting the understanding of temporality and trajectories of psychopathology. We extend prior literature by examining associations between child homelessness and internalising and externalising symptom trajectories in early adolescence, with consideration of timing and persistence of homelessness.
Methods Using population-based longitudinal data from the Avon Longitudinal Study of Parents and Children, we used multilevel models to examine the effects of homelessness prior to age 9, the timing of homelessness (eg, early vs middle childhood) and cumulative exposure to homelessness on internalising and externalising trajectories across ages <0–9 years. We also tested for sex differences in these associations.
Results Of the 8391 participants, 5.5% reported exposure to homelessness at least once before age 9. Children who experienced homelessness had elevated internalising and externalising symptoms compared with their consistently housed peers, with excess risk evident among children who first experienced homelessness in middle childhood (relative to early childhood) and children who experienced recurrent homelessness. We did not observe changes in symptom trajectories over the course of 4 years. Men who experienced homelessness displayed a more pronounced risk of internalising symptoms relative to women and men who did not experience homelessness.
Conclusion Childhood homelessness is associated with persistently elevated internalising and externalising symptoms across early adolescence compared with their consistently housed peers. Interventions and policies to address family homelessness may lead to better mental health among adolescents.
- MENTAL HEALTH
- HOMELESS PERSONS
- CHILD HEALTH
Data availability statement
Data may be obtained from a third party and are not publicly available. Access to Avon Longitudinal Study of Parents and Children data is through a system of managed open access (http://www.bristol.ac.uk/alspac/researchers/access/).
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
WHAT IS ALREADY KNOWN ON THIS TOPIC
Previous studies demonstrated that homelessness during childhood is associated with elevated psychiatric problems. However, longitudinal, population-based studies are scarce and often fail to consider the role of timing and repeated occurrence of homelessness on psychopathology.
WHAT THIS STUDY ADDS
The results of the present study demonstrate that children who experienced homelessness have consistently elevated internalising and externalising problems at age 9 and across early adolescence. These risks are pronounced for those who were homeless in middle childhood (ages 6–8) compared with early childhood (ages <0–5) and those who experienced recurrent homelessness (two times or more prior to age 9).
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
These findings point to the importance of preventing family homelessness and developing targeted programmes for preadolescents who experienced or are experiencing homelessness.
Childhood homelessness is a significant public health concern.1 2 More than 120 000 children in England were living in temporary accommodations in 2018, an increase of approximately 80% since 2011.3 Homelessness has a wide range of adverse effects on child development because it can contribute to pervasive disruptions within school and community relationships,4 substandard housing conditions,5 and unstable family dynamics and routines.6 Prior studies indicate that children experiencing homelessness display more emotional and behavioural problems than their consistently housed peers.7–9 For example, a 2015 meta-analysis showed that 10%–26% of homeless preschoolers and 24%–40% of homeless school-aged children had mental health problems requiring clinical evaluation.1 While most research on homelessness and children’s mental health is cross-sectional, some evidence suggests that repeated experiences of homelessness are important to consider. A study by Sandel and colleagues found that infants who were homeless both prenatally and postnatally were at the highest risk of developmental delays and poor health,10 consistent with research documenting the importance of the duration of other forms of childhood adversity, such as poverty.11 Relatedly, other studies indicate potential sensitive periods during which the impact of adverse experiences interact with dynamic brain systems to subsequently alter behaviour and mental health12–15; however, few studies have examined how the timing of the first episode of homelessness impacts mental health outcomes in early adolescence. Furthermore, few studies have examined sex differences in associations between homelessness and internalising and externalising behaviour trajectories across early adolescence, even though studies on other forms of childhood adversity suggest this is important to consider.16 17 It is possible that differences in the timing of the onset of puberty between men and women partially explains the sex differences in the impact of adverse experiences on the trajectories of internalising symptomatology.18 More specifically, women tend to exhibit internalising problems, whereas men are more likely to demonstrate externalising problems during adolescence19 20; yet, few studies have examined sex differences in the association between homelessness and mental health trajectories.
In this study, we use prospective data from the Avon Longitudinal Study of Parents and Children (ALSPAC) to examine longitudinal patterns of homelessness exposure in childhood on trajectories of internalising and externalising problems across early adolescence. Emerging evidence from several longitudinal studies suggests that early childhood may be a period during which neurodevelopmental processes (eg, neural development, DNA methylation patterns and cognitive functioning) are particularly vulnerable to the effects of severe adversity.21–23 Moreover, early adolescence represents a developmental period in which the onset of and changes in psychopathology are most often observed symptoms.17 Therefore, to better inform policies and practices, we sought to elucidate how the timing of exposure homelessness across the early childhood period impacts internalising and externalising symptomatology across early adolescence. We extend prior research on the associations considering childhood homelessness in three ways. First, advancing on studies that consider any life history of homelessness as a dichotomous exposure, we operationalise homelessness to consider (1) any experience of homelessness from prenatal to 9 years of age (ie, ever homeless); (2) the timing of the first episode of homelessness (ie, prenatal, early childhood and middle childhood); and (3) the cumulative frequency of homeless episodes before age 9. Second, building on cross-sectional studies that rely on shelter-based samples,7 24 25 we use a population-based cohort to examine the longitudinal associations of child homelessness on trajectories of internalising and externalising symptoms across early adolescence. Third, we assess whether these associations differ between boys and girls. Based on previous literature,10 we hypothesise that children whose first episode of homelessness occurred prenatally or during early childhood and children with repeated experiences of homelessness will have the highest levels of internalising and externalising symptoms. Given that evidence with regard to sex differences in the association between homelessness and psychopathology is limited, this aim is exploratory.
Data were drawn from ALSPAC, a prospective birth cohort study investigating genetic and environmental influences on health and development, conducted by the University of Bristol.26 27 Pregnant women residing in the Avon catchment area of South West England with expected delivery dates between April 1991 and December 1992 were recruited. The initial number of pregnancies enrolled was 14 541, and of these, there were 14 062 live births. Participating mothers, partners and study children were assessed repeatedly with in-person visits and mailed questionnaires (http://www.bris.ac.uk/alspac/researchers/our-data).
The analytical sample included 8391 children who had at least partial data on the homelessness variables. Specifically, participants were included if they had at least five complete responses out of 10 items on homelessness status from prenatal to age 9 years and at least one outcome score at ages 9, 11 or 13 years.
Mother-reported homelessness was measured by a dichotomous response to the question ‘You became homeless within (time)’ at each wave of data collection—specifically, during pregnancy and at child ages 0.67, 1.5, 2.5, 5.0, 6.0 and 9.0 years. We created three homelessness variables to capture potential differences in associations based on timing and accumulation. The broadest homelessness variable is binary, defined as ever experiencing homelessness from prenatal to 9 years of age versus never experiencing homelessness during this period. The timing of first exposure to homelessness was categorised into three developmental periods, including prenatal and infancy (ages <0–2), early childhood (ages 3–5) and middle childhood (ages 6–8). The variable for the accumulation of homelessness exposure was calculated as the sum of available homelessness reports across age. The sum was then transformed into a categorical variable, indicating none, one, or two or more exposures to homelessness by age 9. Online supplemental STable 1 summarises the coding of the homelessness variables.
When children were ages 9, 11 and 13 years, mothers reported on psychopathology symptoms in the past 6 months using the Strengths and Difficulties Questionnaire, a valid and reliable tool for measuring emotional and behavioural difficulties.28 We focused on symptomatology rather than a binary cut-off because a continuous variable allows for better representation of the distribution of the symptoms.29 Consistent with prior studies, we created a 10-item internalising problems scale (ie, emotional and peer relationship problems) and a 10-item externalising problems scale (ie, conduct, hyperactivity and inattention problems).30 Scores were standardised to have a mean of 0 and an SD of 1.
Potential confounders were measured via maternal self-report during and after pregnancy and included maternal education, number of financial difficulties at 32 weeks’ gestation (based on a checklist of five material needs), parity at the time of pregnancy, maternal age at child’s birth and child ethnicity (white vs non-white). We adjusted for ethnicity, which is correlated with child homelessness and psychopathology independently (reflecting structural and interpersonal discrimination), as a potential confounder. Maternal postnatal depression was measured by the Edinburgh Postnatal Depression Score questionnaire (continuous scale of 0–30).31 Child sex was obtained from the birth certificate.
Pearson’s correlation was used to establish the relationship between internalising and externalising symptom scores at each time point. We used restriction maximum likelihood estimation to estimate a series of multilevel linear regression models with growth curves (repeated psychopathology symptom measurements nested within individuals) to investigate the effects of homelessness on internalising and externalising behaviour trajectories in early adolescence. The intercept was set to the initial point, corresponding to symptom scores at age 9. First, we modelled trajectories of psychopathology symptoms across time with a linear slope term age and random effects for age. Confounders were included as fixed effects. Second, to test differences in the slopes of internalising and externalising symptoms over time, an interaction term between homelessness and age was included. Identical sets of models were repeated for each of the three homelessness variables and the two study outcomes. To examine sex differences, an interaction term between child sex and any experience of homelessness by age 9 years was added to the model.
To minimise selection bias, multivariate multiple imputation by chained equations was used to impute missing data on covariates and outcomes in all participants who met the study’s inclusion criteria. The imputation models included all variables used in the main analyses, as well as auxiliary variables (eg, marital status, parent alcohol use and number of major life events) that were predictive of incomplete variables and/or missingness. We generated 20 imputation datasets.32 The distribution of variables in observed and imputed data were similar (online supplemental table 2). We used Rubin’s rules to produce a single estimate unbiased to differential losses to follow-up.33
As a sensitivity test of our approach to missing data and modelling approach, after applying the same inclusion and exclusion criteria, we assumed that data were missing at random and used full information maximum likelihood (FIML) with latent growth curve modelling.34 Latent growth curve models estimate the mean trajectory (ie, intercept and slope) as well as individual variation. In these models, the internalising or externalising symptom scores at 9, 11 and 13 years of age were indicators of two latent factors: (1) an intercept representing initial levels of symptoms at 9 years of age and (2) a linear slope representing growth rates in internalising or externalising symptom scores across ages 9–13 years. The same confounders as previously described were included. We also estimated the covariance between the intercept and slope to reflect the strength and direction of the associations. Robust SEs were used to correct for skewness in the variables.
All statistical analyses were conducted in R V.3.6.2.
The distribution of demographic characteristics by homelessness are shown in table 1. Of the 8391 participants, 461 experienced homelessness by age 9. Among children who experienced any homelessness, the majority (72%) was first exposed to homelessness before age 3 years; 18% was first exposed between ages 3 years and 5 years; and 10% was first exposed between ages 6 years and 8 years. Of those who had been homeless by age 9, 83% reported experiencing homelessness once; 17% experienced homelessness two or more times. Internalising and externalising symptoms were correlated concurrently and over time (online supplemental SFigure 1).
Effect of homelessness, age of first episode and persistence on the development of psychopathology
Figure 1 presents the observed mean psychopathology symptoms by homelessness exposure. For both groups, we observed a slight increase in externalising and internalising symptoms across ages 9–13.
Children who experienced homelessness exhibited higher levels of internalising symptoms at 9 years of age compared with those who never experienced homelessness by age 9, after adjustment for individual and maternal characteristics and potential confounders (β=0.179, 95% CI 0.092 to 0.266; see table 2 and online supplemental STable 3 for the full model). The slope indicated that internalising symptom scores were stable during adolescence; the interaction between homelessness exposure and time was not statistically significant, indicating that childhood homelessness was not associated with changes in internalising symptoms over time.
With respect to the timing, children who first experienced homelessness during middle childhood had the highest internalising symptom scores compared with those who were never homeless (β=0.413, 95% CI 0.152 to 0.674). There were similar, but less pronounced, associations for those who first experienced homelessness in early childhood (β=0.206, 95% CI 0.017 to 0.395) and infancy (β=0.141, 95% CI 0.039 to 0.243). There was no interaction between changes in internalising symptoms over time and the timing of first episode of childhood homelessness.
Considering repeated homelessness, we observed a dose–response association between the number of homelessness experiences and internalising symptoms. Specifically, children who experienced homelessness two or more times by 9 years of age displayed the highest internalising symptom scores (β=0.406, 95% CI 0.204 to 0.607) compared with those who were never homeless. There was no interaction between changes in internalising symptoms over time and cumulative exposure to homelessness.
Children who experienced homelessness by age 9 had, on average, higher externalising symptoms than those who had never experienced homelessness by age 9, after adjusting for potential confounders (β=0.167, 95% CI 0.081 to 0.253); see table 3 and online supplemental file 1 for the full model). Externalising symptoms remained stable across ages 9–13 years.
In terms of timing of homelessness, children whose first exposure to homelessness occurred between ages 6 and 8 showed the highest externalising symptoms at age 9 (β=0.322, 95% CI 0.047 to 0.597) compared with those who had never experienced homelessness. Similarly, those whose first exposure to homelessness occurred between prenatal and 2 years of age also showed elevated externalising symptom scores at age 9 (β=0.159, 95% CI 0.059 to 0.259). No effect on externalising symptoms at age 9 was found when the first exposure to homelessness occurred between ages 3 years and 5 years (β=0.118, 95% CI −0.073 to 0.310). There was no interaction between changes in externalising symptoms over time and the timing of first episode of homelessness.
Considering cumulative exposure to homelessness, children who reported two or more exposures to homelessness by age 9 exhibited the highest externalising symptoms compared with those who had never experienced homelessness (β=0.353, 95% CI 0.156 to 0.551); similar trends were found among those who reported one episode of homelessness (β=0.130, 95% CI 0.037 to 0.223). There was no interaction between changes in externalising symptoms over time and cumulative exposure to homelessness.
Differences in associations by sex
We tested differences by sex in the association between the broadest definition of exposure to homelessness by age 9 and psychopathology symptoms (figure 2). No sex differences were observed for the association between homelessness exposure and internalising symptoms (β=0.031, 95% CI −0.006 to 0.068), whereas sex differences were observed for externalising symptoms. Specifically, homeless women had lower externalising symptoms relative to homeless men (β=−0.312, 95% CI −0.349 to 0.275).
Sensitivity analysis using latent growth curve analysis
Latent growth curve analyses of internalising and externalising symptoms showed consistent results as the results derived from multilevel linear models. Exposure to homelessness was associated with higher internalising symptoms at 9 years of age (β=0.219, SE=0.057, p<0.001), but it was not associated with the slope (table 4). Similarly, for externalising problems, exposure to homelessness was associated with higher externalising problems at 9 years of age (β=0.211, SE=0.057, p<0.001), but it was not associated with the slope.
Using prospective data from the ALSPAC cohort, childhood homelessness from pregnancy up to 9 years of age was associated with elevated internalising and externalising problems at age 9, and these associations were maintained after adjustment for individual maternal characteristics and potential confounders. Across both models examining the association between homelessness and internalising and externalising symptoms, the standardised effect size for homelessness was consistently larger than all other covariates associated with either increased internalising or externalising symptoms (online supplemental STable 3 and STable 4). However, given that we examined trajectories of symptomatology as opposed to clinical diagnoses, these differences in standardised effect sizes cannot be directly translated as clinically meaningful for mental health diagnoses. When we examined timing of homelessness, childhood homelessness first occurring between ages 6 years and 8 years demonstrated the largest association with internalising and externalising problems at age 9, relative to those who had never experienced homelessness, as compared with those with first experiences of homelessness earlier in the life course. As hypothesised, children with repeated experiences of homelessness showed the highest internalising and externalising symptoms compared with those who had never experienced homelessness. We observed neither changes in trajectories of psychopathology problems for the study population nor an effect of homelessness on these trajectories; the symptoms were relatively stable across the observation period. Our tests for effect modification revealed that the association between homelessness and externalising symptoms was more pronounced for boys than girls; differences in the association by sex were not observed for internalising symptoms.
Our results are consistent with a meta-analysis focused on younger children, which found that school-aged children (ages 6–11) who had experienced homelessness had elevated internalising and externalising problems compared with low income-housed children of the same age.1 Building on this meta-analysis, we focused on developmental trajectories across early adolescence and found that differences in internalising and externalising symptoms among youths with and without histories of homelessness were evident and stable into adolescence. Our findings related to age of first homelessness experience suggest that recent exposures are more strongly associated with elevated internalising and externalising symptoms during early adolescence relative to more distant experiences. Prior literature on the role of timing of first exposure is mixed. For example, consistent with our results, a population-based sample of Norwegian children with repeated information across childhood and adolescence found that in any given developmental period, high levels of externalising problems were concurrently associated with high family risk exposures.35 As such, psychopathology symptoms might be associated with recent stressors, including homelessness, regardless of the developmental period. In contrast, other evidence suggests that developmental timing of homeless experiences is relevant: a study by Castro and colleagues found that a younger age of first episode of homelessness is associated with elevated internalising symptoms relative to older ages of first episode.36 Finally, our results indicating that repeated experiences of homelessness is associated with the heightened internalising and externalising symptoms support prior literature that has linked increased frequency and duration of homelessness for youths with more mental health problems.36 Taken together, our analyses to evaluate timing and frequency of homelessness align with prior research using ALSPAC, which found that child psychopathology symptoms are better explained by recency and accumulation models relative to sensitive period models.12
Our study fills several gaps in the literature, including use of a population-based cohort to study homelessness, and consideration of symptom trajectories across early adolescence. Many previous studies of homelessness and children’s mental health draw on samples of families living in shelters. Our use of a population-based study allows for an enhanced understanding of the risk of homelessness on early adolescent psychopathology by minimising biases inherent to recruiting participants from homeless shelters (eg, over-representation of families with characteristics that select individuals to use shelters). Furthermore, our nuanced consideration of timing and repeated experiences of homelessness indicate that interventions should prioritise children with repeated experiences of homelessness and individuals with recent histories of homelessness.
This study has several limitations. First, we cannot rule out selection bias due to loss in follow-up. Participants retained in the sample were more likely to be wealthier and healthier than those lost from the study. To reduce selection bias, we used multiple imputation. Additionally, latent growth modelling using FIML as an alternative approach to missing data replicated results from the multilevel models. Second, the majority of participants in this study were white and resided in the former county of Avon; this limits the generalisability of findings to other populations. Furthermore, participants were recruited in the early 1990s, and given that homelessness rates and social safety net programmes have changed since then, future studies could advance the results of our study by using a more generalisable and more recent sample. Third, the cumulative exposure to homelessness was defined as the number of reported homeless experiences across waves, which captures the number of reoccurrences; however, this exposure does not capture duration of homelessness. Fourth, given that child psychopathology symptoms were reported by the mother, there is potential for reporter bias; studies have shown that parents often underestimate the severity of their children’s symptoms of depression and emotional difficulties.37 38 Fifth, due to lack of available ethnicity data and small numbers of minoritised children in the cohort, we were unable to examine race and ethnicity in more detail. Finally, given that we examined trajectories of symptomology as opposed to clinical diagnoses, future studies could assess if these results are clinically meaningful.
In conclusion, by longitudinally examining the role of early homelessness for emotional and behavioural development across preadolescence, this study advances understanding of the trajectories of symptomatology and documents the risk associated with timing and repeated occurrences of homelessness. In comparing young adolescents with and without histories of homelessness, and looking at specific patterns of homelessness across early and middle childhood, we found that homelessness is associated with stable, elevated internalising and externalising symptoms across early adolescence. Increased accumulation of homelessness and first episode of homelessness during middle childhood were associated with elevated levels of psychopathology symptoms, which suggests that attention to frequency and timing is important for future research to characterise the population more accurately and for interventions to better identify target populations. Additional research is needed to determine whether psychopathology symptoms remain elevated throughout late adolescence and adulthood.
Data availability statement
Data may be obtained from a third party and are not publicly available. Access to Avon Longitudinal Study of Parents and Children data is through a system of managed open access (http://www.bristol.ac.uk/alspac/researchers/access/).
Patient consent for publication
This study involves human participants and was approved by Avon Longitudinal Study of Parents and Children Ethics and Law Committee, local research ethics committees and the University of Exeter Medical School Research Ethics Committee. The participants gave informed consent to participate in the study before taking part. The authors obtained approval for secondary data analyses at their local university.
We are extremely grateful to all the families who took part in this study, the midwives for their help in recruiting them and the whole Avon Longitudinal Study of Parents and Children team, which includes interviewers, computer and laboratory technicians, clerical workers, research scientists, volunteers, managers, receptionists and nurses.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Contributors HHK, RK, NS, AT and CD conceived and designed the study. HHK conducted the data analyses. HHK and RK wrote the first draft of the manuscript and also revised the final draft of the manuscript. NS and AT supervised the analysis and writing. All authors critically reviewed the manuscript and approved the final version. HHK is responsible for the overall content as guarantor.
Funding The UK Medical Research Council and Wellcome (grant ref: 217065/Z/19/Z) and the University of Bristol provided core support for Avon Longitudinal Study of Parents and Children(ALSPAC). A comprehensive list of grants funding is available on the ALSPAC website (http://www.bristol.ac.uk/alspac/external/documents/grant-acknowledgements.pdf). This publication is the work of the authors, who will serve as guarantors for the contents of this paper. Partial funding/support for this research was provided by the Harvard T.H. Chan School of Public Health Initiative on Health and Homelessness.
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.