Modelling Depression, Anxiety, and Stress Among Inmates in North Central Nigeria: A Comparison of Baseline and Interaction-Based Multivariate Ordinal Regression Models

Abstract

The global correctional system faces a severe mental health crisis, with inmates suffering disproportionately from depression, anxiety, and stress (DAS). In low-resource settings like Nigeria, this crisis is acute. Traditional statistical models often inadequately capture the ordinal nature of standard mental health scales and the correlations between these cooccurring conditions. This study addresses these methodological gaps by developing and comparing two advanced models to assess DAS among inmates in North Central Nigeria. Using a cross-sectional design, data were collected from 830 inmates across six facilities with the DASS-42 questionnaire and a socio-demographic form. The baseline multivariate ordinal probit model was first fitted to jointly model the three correlated outcomes. To overcome its limitation of assuming constant predictor effects, a novel interaction-based multivariate ordinal model was developed incorporating theoretically-grounded interaction terms. The interaction-based model demonstrated a superior fit (AIC = 8791.96) over the baseline (AIC = 8811.23), revealing critical effect heterogeneities. For instance, the impact of marital status on depression differed by gender. Predictions from the superior model indicated alarming prevalence rates, with 46.9% of inmates likely experiencing extremely severe anxiety and 42.4% severe depression. Distinct joint DAS profiles were identified, highlighting significant co-morbidity. This study concludes that the interaction-based multivariate ordinal model provides a robust and detailed framework for understanding inmate mental health, enabling the precise identification of high-risk subgroups for targeted, efficient, and effective clinical interventions and resource allocation within correctional systems.