Maternal Depression as a Risk Factor for Family Homelessness

Abstract

Objectives. We estimated the effects of maternal depression during the postpartum year, which is often an unexpected event, on subsequent homelessness and risk of homelessness in a national sample of urban, mostly low-income mothers.

Methods. We used logistic regression models to estimate associations between maternal depression during the postpartum year and both homelessness and risk of homelessness 2 to 3 years later, controlling for maternal and family history of depression, prenatal housing problems, and other covariates. Risk factors for homelessness included experiencing evictions or frequent moves and moving in with family or friends and not paying rent.

Results. We found robust associations between maternal depression during the postpartum year and subsequent homelessness and risk of homelessness, even among mothers who had no history of mental illness, whose own mothers did not have a history of depressive symptoms, and who had no previous housing problems.

Conclusions. This study provides robust evidence that maternal mental illness places families with young children at risk for homelessness, contributes to the scant literature elucidating directional and causal links between mental illness and homelessness, and contributes to a stagnant but important literature on family homelessness.

Homelessness is a significant and often glaring social problem in the United States, particularly in urban areas.¹ A 2012 point-in-time assessment revealed that there were more than 600 000 homeless people in the country, among whom 38% were members of families with children.² In 2011, more than 500 000 individuals in families used emergency housing at least once, representing an increase of almost 14% since 2007.²

Although most homeless individuals are single male adults, children and families make up a larger percentage of homeless individuals today than they did in the past.¹ One study of annual prevalence rates (rather than point-in-time measurements) revealed a higher risk of sheltered homelessness among young children than among men in the 1990s.³ Recent data show that 1.6 million children (1 in 45) experienced homelessness in the United States during 2010.⁴ A recent systematic review showed that studies of family homelessness have been on the wane despite the fact that the problem has been worsening.⁵

The federal government recently broadened the official definition of homelessness to recognize and serve individuals and families previously not considered homeless but who are at risk. The 2009 Homeless Emergency Assistance and Rapid Transition to Housing (HEARTH) Act (Pub L No. 111-22) expanded the definition of homelessness to include the “imminently homeless,” outlining several categories under which one can qualify for homelessness assistance. These categories include experiences such as eviction, residential instability, and not having one’s own home.⁶ A recent study showed that 3% of mothers with young children experienced homelessness in the preceding 12 months but that a much larger share (10%) experienced evictions, multiple moves, and doubling up with family or friends without paying rent or sharing living quarters with an adult other than a spouse or partner.⁷

Homelessness is strongly associated with mental illness. According to the US Department of Housing and Urban Development, 29% of homeless adults in the United States suffer from some disabling form of mental illness,² and data from the Substance Abuse and Mental Health Services Administration indicate that 6% of American women are afflicted with a serious mental illness.⁸ More refined mental illness prevalence data (e.g., among women or mothers) allowing comparisons between homeless and nonhomeless individuals in the United States are scarce, and existing published figures are based on small or very specific samples.

Most existing studies exploring links between mental illness and homelessness were based on data from the 1980s and 1990s or did not focus on women or families. This literature reveals strong positive associations between mental illness and homelessness.^9–13 Robertson and Winkleby synthesized and critically reviewed studies focusing specifically on women and found that the literature was largely cross sectional and descriptive.¹⁴ Some studies produced evidence of increases in mental illness among homeless mothers in the 1980s and 1990s^15,16 and pointed to inadequate availability of community-based mental health services as an exacerbating factor.^15–17 Recent studies provide evidence that mental illness is associated with subsequent family homelessness,^18,19 although it is difficult to rule out that the observed associations reflect unobserved confounding factors.

Overall, it is widely assumed that homelessness and mental illness are linked and that the connections may be bidirectional. However, there is little strong evidence of causal links in either direction. Furthermore, more research on mental illness and family homelessness is needed. We addressed these gaps by estimating the effects of a form of mental illness that is often unexpected and has a known timing of onset—depression during the postpartum year—on subsequent homelessness in a national sample of urban, mostly low-income, childbearing women. We estimated effects among mothers overall as well as among subgroups of mothers with no prenatal history of mental illness, whose own mothers had no history of depressive symptoms, and who had not experienced housing problems or homelessness prior to the episode of mental illness. In the spirit of the HEARTH Act, we considered both homelessness itself and risk of homelessness as outcomes.

METHODS

Births were randomly selected from birth logs in 75 hospitals in 20 US cities with populations greater than 200 000 as part of the Fragile Families and Child Wellbeing study, a longitudinal birth cohort survey. A mother was eligible for the study if both she and the baby’s father were at least 18 years old or, if they were minors, the hospital allowed recruitment of minors into the study; if she was able to complete the interview in either English or Spanish; if the father of the newborn was living; and if the couple was not planning to place the child for adoption. Births among nonmarried individuals were oversampled.²⁰ A total of 4898 mothers (86% of those eligible) were interviewed while they were still in the hospital after giving birth, between spring 1998 and fall 2000. Fathers were also interviewed.

Of the mothers who completed initial (baseline) interviews, 89% were reinterviewed 1 year later and 86% were reinterviewed when their children were 3 years old. Additional information was collected from medical records of mothers and newborns in 61 of the 75 hospitals. The availability of medical record data depended primarily on a hospital’s administrative processes rather than decisions of respondents to make their records available. Of the 3684 mothers with available medical record data (needed to control for prenatal mental illness and other covariates), 3024 completed both follow-up surveys; 50 of these 3024 women were excluded from the analysis sample because they were missing data on key variables, leaving 2974 cases for analysis. Comparisons (at baseline) between the 2974 mothers who were included in the analysis sample and the 1924 mothers who were not indicated no significant differences with respect to education, relationship status, race/ethnicity, insurance status, prenatal living arrangements, or parity.

Measures

Exposure variable.

We assessed depression during the postpartum year with an indicator of whether the mother met diagnostic criteria for major depression in the preceding 12 months according to the Composite International Diagnostic Interview Short Form, an instrument widely used in epidemiological and population research studies. Our main measure of depression was based on number of depressive symptoms experienced (range = 0–7), with a major depressive episode defined as the experience of 3 or more symptoms of dysphoria or anhedonia (e.g., feeling sad, losing interest in activities one usually enjoys) for most of the day during a period of at least 2 weeks. We also used a less stringent measure in supplementary analyses described later. Both measures, which were assessed at the 1-year follow-up interview, have been validated in the literature.²¹

Although our measures of maternal depression were not based on the most commonly used postpartum depression screening instruments (e.g., the Edinburgh Postnatal Depression Scale), they were based on a validated instrument used to assess diagnostic criteria for major depression and were focused on the postpartum year. However, we cannot assume that our measures characterize depression specifically related to childbearing. Mitchell et al. found that rates of depression (based on scores from the Composite International Diagnostic Interview Short Form) during the postpartum year among mothers taking part in the Fragile Families and Child Wellbeing study fell in the expected range for postpartum depression.²² Nevertheless, we refer to “depression during the postpartum year” rather than “postpartum depression.”

Outcomes.

First, we considered homelessness as it is traditionally defined, as lack of a fixed, regular, and adequate nighttime residence or residence in a temporary accommodation or space not intended for residence. In the 3-year follow-up interview, the mother was asked where she currently lived. Response choices included living in a temporary shelter and being homeless. She was also asked whether she had stayed at a shelter or in an abandoned building, automobile, or any other place not meant for regular housing, even for one night, in the preceding 12 months. If she responded affirmatively to either question, she was coded as having been homeless.

Second, we considered a measure of risk of homelessness based on the HEARTH Act, which considers the following groups to be at risk for homelessness: individuals or families who will imminently lose their primary nighttime residence within a 14-day period, wherein evidence of eviction is provided and the individual or family has no other resources and no other residence has been identified, and families with young children in which the parents have not been named on a lease within the preceding 60 days.

Our measure of risk of homelessness, the best approximation possible with our data, included (1) an affirmative response to the 3-year interview question “Were you evicted from your home or apartment for not paying the rent or mortgage?” (in the past 12 months); (2) situations in which the mother moved 3 or more times between the 1- and 3-year interviews, which seemed to be a reasonable criterion given findings by others that moving more than once per year is a risk factor for homelessness^23,24; or (3) positive responses at the 3-year follow-up to questions about living with family or friends without paying rent or living with an adult other than a spouse or partner.

We coded 80 mothers in our analysis sample as having experienced homelessness at 3 years and 284 as being at risk for but not having experienced homelessness. All summary statistics and regression analyses involving the at-risk outcome excluded mothers who had experienced homelessness at 3 years.

Covariates.

Demographic and socioeconomic factors included mother’s age, race/ethnicity, nativity (foreign-born vs not foreign-born), education, type of health insurance (Medicaid vs private or unknown), employment (whether the mother had worked in the 2 years before the birth), relationship status with the child’s father, and parity, all measured at baseline (racial/ethnic, education, and relationship status categories are shown in Table 1). We also included measures of the mother’s prenatal mental and physical health: whether there was any documentation of previously diagnosed mental illness (excluding substance abuse diagnoses) and whether there was a preexisting physical health condition in the mother’s prenatal medical record. In addition, we included a measure of whether the mother used illicit drugs during her pregnancy. This measure, used in past research,²⁵ was coded as positive if there was any indication of prenatal drug use from the medical record (e.g., from drug testing of the mother or infant) or the mother reported at baseline that she had used illicit drugs during the pregnancy.

TABLE 1—

Characteristics of Respondents Participating in the Fragile Families and Child Wellbeing Study Who Gave Birth Between 1998 and 2000: United States

Characteristic	Full Sample (n = 2974), % or Mean (SD)	Women With Depression During Postpartum Year (n = 375), % or Mean (SD)	Women Without Depression During Postpartum Year (n = 2599), % or Mean (SD)
Outcome
Homeless**	3	6	2
At risk for homelessness**	10	14	9
Maternal characteristics
Age, y*	25.0 (6.0)	24.4 (5.8)	25.1 (6.1)
Race/ethnicity
Non-Hispanic White	21	20	21
Non-Hispanic Black	48	52	48
Hispanic	27	25	27
Other	4	3	4
Foreign-born**	15	10	15
Education
< high school	35	35	34
High school	31	31	30
Any college	35	34	35
Birth covered by Medicaid insurance	65	69	64
Employed during 2 y before the birth	81	83	80
Relationship status**
Married	23	18	24
Cohabiting	37	35	37
Neither married nor cohabiting	40	47	39
No. of children	1.1 (1.3)	1.2 (1.3)	1.1 (1.4)
Prenatal diagnosed mental illness**	11	22	10
Prenatal physical health condition	20	21	20
Prenatal illicit drug use*	10	14	10
Child characteristics
Severe health condition	2	2	2
Male	52	52	52
Multiple birth	2	2	2
Age when outcomes were measured, mo	35.7 (2.4)	35.8 (2.4)	35.7 (2.4)
Paternal characteristics
Higher education category than mother	24	23	24
Depression score (CES-D)	1.24 (1.2)	1.25 (1.1)	1.24 (1.2)
Suboptimal physical health	41	44	41
Did not complete baseline interview	18	19	18
Prenatal housing situation
Housing problems	2	3	2
Lived with adult (not including child’s father)	33	34	33
Maternal grandparents’ mental illness status
Grandmother had history of depressive symptoms**	26	46	23
Grandfather had history of depressive symptoms**	12	21	11
Data on grandmother’s depressive symptoms missing*	7	11	7
Data on grandfather’s depressive symptoms missing**	28	35	27

Note. All row characteristics other than child’s age were measured at or before the birth of the focal child. For fathers who did not complete baseline interviews, suboptimal physical health was set equal to 1 and scores on the Center for Epidemiologic Studies Depression Scale (CES-D) were set equal to the mean score for cases in which fathers’ data were available. Mothers’ reports of fathers’ education were used when fathers’ reports were not available.

*P < .05; **P < .01 (significant differences between mothers who experienced depression during the postpartum year and those who did not, based on 2-tailed t test for comparisons of means for binary variables and χ² test for categorical variables).

We also included variables related to the focal child: sex, poor health, multiple birth, and age (in months) at the time of the 3-year interview. In the case of poor infant health, we used a measure based primarily on information derived from the newborn’s medical record and designed to capture serious conditions considered random by the medical community (for details, see Table A, available as a supplement to the online version of this article at http://www.ajph.org). This measure has been used previously to study the effects of unexpected infant health conditions on housing circumstances²⁶ and homelessness.⁷

In addition, we included information about the educational level of the infant’s father (relative to the mother), his mental and physical health status, and whether or not he completed a baseline interview. The father’s risk of depression was based on his score on a short version of the Center for Epidemiologic Studies Depression Scale, which was administered during his baseline interview. We used the father’s self-reports to code his overall health status (good, fair, or poor vs excellent or very good) at baseline (more information about coding of paternal variables is provided in the notes for Table 1).

We controlled for the mother’s housing conditions and living situation prior to the birth. The first measure captured housing problems on the basis of information included in the mother’s prenatal medical record. The abstractors were instructed to record any mention (in progress notes or elsewhere) of “homelessness or threatened eviction” and “inadequate heat, electricity, or running water or other poor housing/living conditions.” These categories were combined to create the baseline measure of housing problems. The second measure was whether the mother lived with an adult other than the baby’s father during the pregnancy.

To capture a mother’s family-related (genetic or environmental) predisposition for depression, we included measures characterizing the mental health of each of her parents. These measures were based on positive responses to the following question in the 3-year survey: “Did your biological mother/father ever have periods lasting 2 weeks or more when she/he was depressed, blue, or down in the dumps most of the time?” We also included indicators for missing data on these measures, setting the corresponding measures to zero for those cases.

Finally, we included indicators for the mother’s city of residence at baseline to control for state policies or other potentially confounding city-level factors. Cities with fewer than 100 observations were combined into a single category.

Statistical Analyses

First, we compared mothers who experienced and who did not experience depression during the postpartum year. We used the 2-tailed t test (for comparisons of means) and the χ² test (for categorical variables) to assess differences between groups. Second, we estimated logistic regression models that controlled for the maternal, child, paternal, and family characteristics described in the previous section as well as the mother’s baseline city of residence. We used all 2974 observations for the models of homelessness and 2894 of those observations (for respondents who had not experienced homelessness) for the models of risk of homelessness.

Finally, we estimated logistic regression models for subgroups of mothers, including those with no indication of a diagnosed mental illness in their prenatal medical record, those whose own mothers had no history of depressive symptoms, those who had no indication of a prenatal diagnosed mental illness and whose own mothers had no history of depressive symptoms, those with no housing problems at baseline (according to the medical record measure), and those with no indication of housing problems at baseline and no homelessness or risk of homelessness at the 1-year follow-up (according to measures corresponding to the 3-year measures).

Odds ratios (ORs) and 95% confidence intervals are presented for the logistic regression models. We estimated models with alternative specifications to assess sensitivity and patterns of estimates. Throughout we refer to values outside the 95% confidence intervals as statistically significant. Stata version 12.1 statistical software (StataCorp LP, College Station, TX) was used in conducting all analyses.

RESULTS

Thirteen percent (n = 375) of mothers experienced depression during the postpartum year. Although statistically different, the mean ages of women with and without depression during the postpartum year were similar. The median age was 24 years; more than three quarters of the women were younger than 30 years, and fewer than 3% were younger than 18 years (data not shown). Mothers who experienced depression were significantly more likely than those who did not to become homeless (6% vs 2%) and to be at risk for homelessness (conditional on not having become homeless; 14% vs 9%). They were less likely to be foreign-born. There were significant differences with respect to relationship status but not with respect to maternal race/ethnicity, education, or physical health or child or paternal characteristics.

Although women who experienced depression were more likely to have had housing problems at baseline, between-group differences were not statistically significant. Mothers who experienced depression were significantly more likely than those who did not to have a prior diagnosed mental illness (22% vs 10%) and to have both mothers and fathers with histories of depressive symptoms. They were also significantly more likely to have used illicit drugs during their pregnancy (14% vs 10%) (Table 1).

The logistic regression models included all of the characteristics outlined in Table 1 along with quadratic terms for age (estimates shown) and city indicators (estimates not shown). Depression during the postpartum year was associated with more than twice the odds of homelessness (OR = 2.29) and almost 1.5 times the odds of risk of homelessness (OR = 1.40) at 3 years (Table 2). Few covariates were statistically significant. Prenatal drug use was a strong predictor of homelessness even after controlling for prenatal mental illness, depression during the postpartum year, and grandparents’ history of depressive symptoms. Variables related to grandmother’s depression appeared to be more strongly associated with the outcomes than those related to grandfather’s depression. Baseline housing problems were strongly associated with subsequent homelessness, whereas doubling up prior to the birth was strongly associated with subsequently being at risk for homelessness at 3 years.

TABLE 2—

Effects of Depression During the Postpartum Year and Study Covariates on Homelessness and Risk of Homelessness at 3-Year Follow-Up for Those Who Gave Birth Between 1998 and 2000: Fragile Families and Child Wellbeing Study, United States

Characteristic	Homeless, OR (95% CI)	At Risk for Homelessness, OR (95% CI)
Depressed during postpartum year	2.29 (1.08, 4.85)	1.40 (1.12, 1.75)
Other maternal characteristics
Age	1.09 (0.81, 1.45)	0.75 (0.61, 0.92)
Age squared	1.00 (0.99, 1.00)	1.00 (1.00, 1.01)
Non-Hispanic Black	3.80 (1.43, 10.11)	0.73 (0.47, 1.14)
Hispanic	1.83 (0.68, 4.89)	0.84 (0.56, 1.27)
Other non-White	9.26 (2.38, 35.93)	1.30 (0.51, 3.27)
Foreign-born	0.38 (0.11, 1.31)	0.91 (0.39, 2.12)
High school graduate	1.06 (0.57, 1.98)	0.80 (0.57, 1.12)
Completed any college	0.60 (0.27, 1.31)	0.79 (0.55, 1.15)
Birth covered by Medicaid insurance	1.59 (0.92, 2.74)	1.29 (0.87, 1.91)
Employed	1.24 (0.61, 2.48)	0.73 (0.58, 0.92)
Married	0.72 (0.21, 2.47)	0.51 (0.27, 0.94)
Cohabiting	0.80 (0.48, 1.31)	0.74 (0.53, 1.05)
No. of children	1.04 (0.85, 1.28)	0.90 (0.78, 1.04)
Prenatal diagnosed mental illness	0.75 (0.33, 1.73)	0.99 (0.73, 1.33)
Prenatal physical health condition	1.02 (0.72, 1.44)	1.03 (0.74, 1.45)
Prenatal illicit drug use	2.80 (1.42, 5.53)	1.50 (1.09, 2.06)
Child characteristics
Severe health condition	2.73 (0.69, 10.80)	1.26 (0.59, 2.69)
Male	0.67 (0.37, 1.24)	1.10 (0.88, 1.38)
Multiple birth	1.57 (0.43, 5.74)	1.12 (0.34, 3.65)
Age when outcomes were measured	1.09 (1.01, 1.16)	0.99 (0.94, 1.04)
Paternal characteristics
Higher education category than mother	1.28 (0.75, 2.17)	0.75 (0.52, 1.10)
Depression score	1.00 (0.82, 1.21)	1.04 (0.96, 1.14)
Suboptimal physical health	1.31 (0.80, 2.15)	0.88 (0.54, 1.41)
Did not complete baseline interview	0.75 (0.35, 1.61)	1.06 (0.59, 1.93)
Prenatal housing situation
Housing problems	3.18 (1.31, 7.72)	0.62 (0.27, 1.41)
Lived with adult (not including child’s father)	1.09 (0.65, 1.82)	1.63 (1.25, 2.13)
Maternal grandparents’ mental illness status
Grandmother had history of depressive symptoms	2.71 (1.26, 5.80)	1.31 (1.04, 1.65)
Grandfather had history of depressive symptoms	0.76 (0.41, 1.42)	1.59 (1.01, 2.48)
Data on grandmother’s depressive symptoms missing	2.85 (1.34, 6.07)	2.02 (1.24, 3.30)
Data on grandfather’s depressive symptoms missing	0.66 (0.30, 1.45)	1.02 (0.76, 1.36)

Note. CI = confidence interval; OR = odds ratio. Models include indicators for the mother’s city of residence at baseline; cities with fewer than 100 observations are grouped together (estimates not shown). All variables other than the study outcomes (homelessness and risk of homelessness), depression during the postpartum year, and child’s age were measured at or before the birth of the focal child. The sample size was n = 2974.

On the basis of past literature pointing to one’s own and a family history of psychopathology as known risk factors for postpartum depression,²⁷ we estimated models for subsamples of mothers among whom depression during the postpartum year should have been a particularly unexpected event: mothers with no prenatal diagnosed mental illness, mothers whose own mothers had no history of depressive symptoms, and mothers who had no diagnosed prenatal mental illness and whose own mothers had no history of depressive symptoms. In addition, given longitudinal evidence suggesting that homelessness is a risk factor for maternal depression,²⁸ and thus that some of the depression experienced by women during the postpartum year could have been a consequence of ongoing housing instability, we estimated models for subsamples of mothers with no housing problems at baseline and mothers who had no housing problems at baseline and who were not homeless or at risk for homelessness at 1 year.

The results of these models are summarized in Table 3, along with corresponding full-sample results for comparative purposes. In all cases, we found that the associations between depression during the postpartum year and the outcomes assessed were on par with those for the overall sample. The estimates from the analyses focusing on women who had no housing problems at baseline and were not homeless or at risk for homelessness at 1 year should be interpreted with caution because some women who were homeless or at risk for homelessness owing to postpartum depression may not have been included. However, it is reassuring that our estimated effects of depression during the postpartum year were insensitive to this very strict specification.

TABLE 3—

Effects of Depression in the Postpartum Year on Homelessness and Risk of Homelessness at 3-Year Follow-Up for Those Who Gave Birth Between 1998 and 2000: Fragile Families and Child Wellbeing Study, United States

	Homeless		At Risk for Homelessness
Depression During Postpartum Year	No.	OR (95% CI)	No.	OR (95% CI)
Full sample	2974	2.61 (1.43, 4.78)	2894	1.51 (1.23, 1.87)
Mothers with no prenatal diagnosed mental illness	2639	2.67 (1.33, 5.37)	2575	1.53 (1.21, 1.93)
Mothers whose own mothers had no history of depressive symptoms	2209	2.65 (1.05, 6.69)	2162	1.48 (0.95, 2.04)
Mothers who had no prenatal diagnosed mental illness and whose mothers had no history of depressive symptoms	1995	3.04 (1.08, 8.58)	1956	1.52 (0.97, 2.40)
Mothers with no baseline housing problems	2912	2.49 (1.35, 4.59)	2841	1.47 (1.18, 1.82)
Mothers who had no baseline housing problems and were not homeless or at risk for homelessness at 1 y	2240	2.30 (1.04, 5.08)	2204	1.86 (1.14, 3.04)

Note. CI = confidence interval; OR = odds ratio. Owing to small sample sizes, covariates are limited to those classified as maternal characteristics, child characteristics, and paternal characteristics; city indicators and the indicator for multiple birth are not included.

We estimated supplementary models to assess sensitivity and patterns of estimates. Comparisons between unadjusted models and models that adjusted for all of the covariates other than grandparents’ depression revealed decreases in the odds ratios for depression related to homelessness (from 2.91 to 2.73) as well as to risk of homelessness (from 1.59 to 1.56). Adding the 4 measures of grandparents’ depression further reduced the odds ratios only modestly to 2.62 and 1.52, respectively, underscoring that we captured a largely unexpected mental health event. Models incorporating a broader validated measure of depression during the postpartum year, one that characterized respondents who reported experiencing symptoms for at least half of the day (instead of most of the day) for a period of at least 2 weeks in the preceding 12 months, produced slightly lower odds ratios for depression than those shown in Table 2 in the case of both homelessness and risk of homelessness.

Finally, we estimated models predicting prenatal housing problems and living with an adult other than the baby’s father at baseline as a function of maternal depression during the postpartum year. The logic was that if depression during the postpartum year was truly an unexpected event, it should not be associated with women’s prenatal housing situation. We found that this was the case, providing some evidence that we isolated causal effects of maternal depression on homelessness.

DISCUSSION

In a national sample of childbearing women in large US cities, we found robust associations between maternal depression during the postpartum year and both homelessness and risk of homelessness 2 to 3 years later, even among mothers who had no history of mental illness, whose own mothers had no history of depressive symptoms, and who had no previous housing problems. By exploiting longitudinal data, focusing on a mental illness with clear timing of onset, and producing results that are robust across specifications and subsamples, we have come closer than previous studies to establishing directionality and a plausibly causal connection. Of course, our findings do not preclude the possibility that homelessness has deleterious effects on mental health.

Although our findings are quite robust, they are subject to several limitations. For example, findings for predominantly unmarried adult mothers giving birth in urban areas may not generalize to all women giving birth in the United States. In addition, results related to maternal depression during the postpartum year may not generalize to women who are not parents or who have older children. The measure of risk of homelessness did not map perfectly to the relevant HEARTH statutes, and our measure of prenatal mental illness was limited to documented and diagnosed cases of any mental illness and therefore was not specific to depression. Mothers self-reported their own parents’ depressive symptoms. Other potential sources of bias included hospitals that did not make medical records available and individual-level attrition from the longitudinal study. We were not able to explore the potential buffering effects of treatment of depression with our data. Finally, despite our attempts to isolate causal effects, it is possible that the observed associations reflect factors not assessed.

Notwithstanding the limitations just described, this study establishes that maternal depression during the postpartum year is a significant risk factor for homelessness. Research on mental illness and homelessness has been plagued by a lack of representative data that capture episodes of mental illness, document transitions to homelessness, and allow for rigorous investigation of causal links. This study takes a significant step forward in addressing that gap. It contributes to the literature in that not only was a traditional measure of homelessness used but also a measure of risk of homelessness that, in the spirit of the HEARTH Act, included experiences of eviction, multiple moves, and doubling up. Another important contribution of this study is the focus on mothers with young children, a group that has been underrepresented in the literature on mental illness and homelessness and in the homelessness literature more generally.

Overall, our study provides robust evidence that maternal mental illness places families with young children at risk for homelessness, adds to the scant literature that elucidates directional and causal links between mental illness and homelessness, and contributes to a largely stagnant but important body of literature on family homelessness.