Dissertations and Theses
Date of Degree
8-14-2025
Document Type
Dissertation
Degree Name
Doctor of Philosophy (Ph.D.)
Department
Epidemiology and Biostatistics
Advisor(s)
Heidi E. Jones
Committee Members
Katarzyna E. Wyka
Zach Shahn
Kristin Turney
Subject Categories
Epidemiology | Public Health
Keywords
Social epidemiology, mental health, criminal legal system involvement, health inequities, causal inference, g-methods
Abstract
BACKGROUND: Interactions with the criminal legal system (CLS), such as through arrest or incarceration, are relatively common in the United States (US) and are increasingly recognized as social risk factors for poor mental health. Previous studies have reported a positive association between CLS involvement and poor mental health, with some theoretical frameworks, including the ecosocial theory and the stress process paradigm, supporting a plausible causal relationship. However, prior research has significant limitations in providing casual evidence given possible reverse causation in cross-sectional studies and insufficient consideration for the dynamics between CLS exposures, mental health status, and related confounders over time (i.e., potential bias from treatment-confounder feedback). Further, the relationship between CLS involvement and mental health may vary by social position but has been rarely explored using the intersectionality framework. This dissertation was undertaken to address the methodological shortcomings in prior research to examine the effects of CLS involvement on mental health and to further explore the intersectionality of the heterogeneity of the effects.
METHODS: This dissertation comprised three studies analyzing data from the National Longitudinal Study of Youth 1997 (NLSY97), an ongoing national cohort of youth in the US who were born between 1980 and 1984. In Aim 1, I implemented sequence analysis to analyze annual measures of CLS involvement status to identify longitudinal patterns of CLS involvement from 12 to 31 years of age. Bivariate associations between the identified patterns and mental health and related risk factors were examined. In Aim 2, I used the parametric g-formula to estimate the effects of CLS involvement on severe depressive symptoms and self-reported mental health problems (primary outcomes) as well as binge drinking (secondary outcome) among a subset of NLSY97 respondents followed every two years between 2000 and 2010. The effects were quantified by the differences in the outcome risks under the natural course of CLS involvement versus an intervention of removing CLS involvement (intervention 1: removing both arrests and incarceration; intervention 2: removing incarceration only). Aim 3 shared a similar methodology as Aim 2. I first conducted latent class analysis to identify intersectional social strata based on baseline indicators for social class, socioeconomic status, and other factors related to experiences of hardships and marginalization. I then conducted stratified parametric g-formula analysis to obtain risk differences resulting from the removal of arrest and incarceration within each of the identified social strata. The effect estimates between the most privileged stratum and each of the other disadvantaged strata were then compared.
RESULTS: In Aim 1, four distinct longitudinal patterns of CLS involvement were identified from sequence analysis among those ever involved, ranging from scattered to heavy involvement. Findings indicate unique mental health profiles by trajectory of CLS involvement. Individuals with moderate or heavy involvement showed higher burden of severe depressive symptoms, especially at adolescence but not at later adulthood, and early-age mental health risk factors than those not or less involved. Findings from Aim 2 indicate negative but modest effects of CLS involvement on mental health. Under a hypothetical intervention of removing arrests and incarcerations from the study population, the risks for severe depressive symptoms and mental health problems decreased by 0.4 (95% confidence interval [CI]: -0.8, -0.04) and 0.6 (95% CI: -1.0, -0.3) percentage points (pp), respectively. The intervention also lowered the risk for binge drinking behaviors by 0.6pp (95% CI: -1.2, -0.04). Such reductions were likely attributable to the removal of arrests given that incarceration was rare in the study population. Removal of incarceration alone did not affect the outcome risks. In Aim 3, latent class analysis identified one relatively privileged stratum comprising mostly White non-Hispanic people of high socioeconomic status (SES) residing in two-parent households and three relatively disadvantaged strata. The disadvantaged stratum of predominantly low-SES mother of color had a 1.9pp (95% CI: -3.3, -0.08) reduction in the risk of mental health problems, and the disadvantaged stratum comprising mostly low-SES Black non-Hispanic people residing in single-parent households had a 1.4pp (95% CI: -2.7, -0.3) reduction in the risk for binge drinking, but no impacts were seen in other strata or outcomes. These risk reductions, however, were not statistically different from the ones observed among the privileged stratum.
DISCUSSION: My findings demonstrate longitudinal variations in CLS involvement among American youth transitioning to early adulthood and indicate an association between longitudinal involvement pattern and early-age mental health-related characteristics, suggesting possible pre-existing mental health risks among people more severely involved with the CLS over the life course and nuanced dynamics between CLS exposures and mental health. The application of the parametric g-formula, which accounted for the dynamic relationships between CLS exposures, mental health status, and related confounders over time, suggest small harms to mental health from CLS involvement via primarily arrests, and some of which may be manifested through binge drinking behaviors. My findings support the hypothesis that there is a causal association between CLS involvement and poorer mental health, highlighting the importance of addressing health disparities experienced by the CLS-involved population through public health programming. This dissertation also offers insight into the heterogeneity of mental health impact from CLS involvement by demonstrating some variation by intersectional social strata and potential inequities among individuals occupying multiple disadvantaged statuses. The lack of statistical evidence about differential impacts of CLS involvement between privileged and more disadvantaged strata warrants further investigation from future studies. Overall, this dissertation underscores poorer mental health experienced by the CLS-involved population and, given the root of CLS involvement in structural racism, adds to the literature of racial equity.
Recommended Citation
Chan, Pui Ying, "Disentangling the effect of criminal legal system involvement on mental health in the United States" (2025). CUNY Academic Works.
https://academicworks.cuny.edu/sph_etds/119
