Housing Instability and Change in Maternal Depression Risk: A Longitudinal Mixed Effects Approach

Background: Housing instability may trigger depression among at-risk mothers, with substantial consequences for family stability and child well-being. Both housing problems and depression are episodic and dynamic, hindering analyses that adequately capture change over time. The present study used longitudinal generalized linear mixed modeling to examine how housing instability impacts trajectories of depression risk both within and between mothers across time. Method: Data came from the Fragile Families and Child Well-being Study (FFCW), a longitudinal cohort study that followed 4,898 children born 1998-2000 in 20 large American cities. The analytic sample was limited to mothers with primary custody of their children for all five waves (baseline and 1-, 3-, 5-, and 9-year follow-ups; N = 2,724). The dependent variable was a dichotomous measure of depression assessed at each wave using the Composite International Diagnostic Interview-Short Form (CIDI-SF). A minimum-centered mother’s age variable measured time. The dichotomous time-varying predictor housing instability was measured at each wave; mothers were considered unstably housed if they had skipped rent or mortgage payments, doubled up, moved more than twice, been evicted, or spent at least one night homeless in the past year. Time-constant predictors included race and family history of depression. Univariate statistics described the sample, while a series of generalized linear mixed models of change tested within- and between-mother risks for depression over nine years. Results: Mothers were overwhelmingly nonwhite (77%) and relatively young (Mean = 25.26 years; SD = 6.03) at baseline. Rates of housing instability ranged from 32% to 41%, and rates of depression ranged from 15% to 20% across all waves. An unconditional means model yielded an intraclass correlation (ICC) of 0.30, indicating 30% of the variation in depression risk was explained by person-level (between-mother) differences and justifying the use of a mixed effects approach. An unconditional growth model indicated depression risk changed over time; the addition of time as a random effect allowed the rate of change to vary between mothers (b = -0.05, p < 0.001; σ² = 0.01). A mixed model including the full set of covariates improved fit as indicated by AIC reduction (7,952.9 compared to 9,075.3). This model showed family history elevated risk for depression by nearly five-fold (b = 1.60, p < 0.001), while housing instability more than doubled risk (b = 0.78, p < 0.001). Depression risk declined gradually over time, with each one-year increase in age reducing depression risk by 2% (b = -0.02, p < 0.001). The random effects portion of the model yielded variance components of 2.20 for initial depression risk, and 0.003 for annual change. Overall, results indicated family history and housing stability had large effects on depression risk, but trajectories of change over time were similar across mothers. Conclusions and Implications: Housing crises strongly predict depression risk. Mother- and child-serving agencies should screen for housing and mental health problems, as well as family mental health histories to facilitate prevention and supportive services. Housing assistance programs for at-risk families may mitigate depression risk over time.