ABSTRACT
Child maltreatment is associated with health risk behaviors carrying high personal and societal costs,
such as substance use and sexual risk behavior.1–4 Placement of children into out-of-home care aims to
mitigate these negative effects.5 Findings about rates of health risk behavior among youth in out-of-home care
are mixed,6–16 although studies show that young adults formerly in care have heightened levels of health risk
behavior.17–20 These mixed findings may be due to the fact that most studies test effects of placement across
childhood on later health risk behaviors. However, placement during adolescence may be a turning point that
redirects trajectories of health risk behavior. Adolescence is a period of rapid change and a typical period for
initiation of these behaviors,21 and adolescents have more difficulties adjusting to placement than do younger
children.22 To test whether adolescent placement acts as a turning point, it is important to know whether
placement initiates long-term changes in trajectories, short-term (time-specific) changes, or both. To design
interventions, it is also necessary to understand the mechanisms underlying placement effects. However,
because placement may occur at various points across time, identifying its direct and mediated effects requires
modeling out-of-home placement, mediators, confounders, and outcomes as time-varying variables.
Importantly, out-of-home placement is a nonrandom event, and many pre-placement variables likely
predict both placement and subsequent health risk behavior. To deal with this threat to causal inference, recent
research has utilized propensity score methods.23–27 However, these methods cannot accommodate time-
varying variables, precluding tests of placement as a turning point or tests of its time-specific mediators.
Accordingly, the proposed study uses two modern methods for causal inference, inverse probability of
treatment weighting (IPTW) and g-estimation, to test direct and mediated effects of placement on health risk
behaviors across time. These methods can account for biases from observed confounders and accommodate
time-varying variables.28,29 Aim 1 tests whether adolescents with out-of-home placement since the prior wave
show short-term, time-specific increases in substance use and initiation of unprotected sex, compared to those
who remain in-home. Aim 2 tests whether adolescent placement predicts longer-term elevations in adult health
risk behavior. Aim 3 tests time-varying mediators of the effect of placement on adult health risk behavior.
These aims will be tested in a national sample of 738 maltreated youth (age 11-14 at baseline) studied across
four waves to adulthood. The training plan brings together a distinguished mentoring team to provide
multidisciplinary training in child welfare system research, modern quantitative methods for causal inference,
developmental psychopathology of health risk behaviors, implications for intervention and policy, and
professional development. This will provide the applicant with the foundation to launch a research career
studying the effects of placement and child welfare involvement on adolescents' developmental outcomes.