Introduction: National estimates suggest that up to 80% of prison inmates meet diagnostic criteria for a substance use disorder. Because more substance abuse treatment while incarcerated is associated with better post-release outcomes, including a reduced risk of accidental overdose death, the stakes are high in developing novel predictors of substance abuse treatment completion in inmate populations.
Methods: Using electroencephalography (EEG), this study investigated stimulus-locked ERP components elicited by distractor stimuli in three tasks (VO-Distinct, VO-Repeated, Go/NoGo) as a predictor of treatment discontinuation in a sample of male and female prison inmates. We predicted that those who discontinued treatment early would exhibit a less positive P3a amplitude elicited by distractor stimuli.
Results: Our predictions regarding ERP components were partially supported. Those who discontinued treatment early exhibited a less positive P3a amplitude and a less positive PC4 in the VO-D task. In the VO-R task, however, those who discontinued treatment early exhibited a more negative N200 amplitude rather than the hypothesized less positive P3a amplitude. The discontinuation group also displayed less positive PC4 amplitude. Surprisingly, there were no time-domain or principle component differences among the groups in the Go/NoGo task. Support Vector Machine (SVM) models of the three tasks accurately classified individuals who discontinued treatment with the best model accurately classifying 75% of inmates. PCA techniques were more sensitive in differentiating groups than the classic time-domain windowed approach.
Conclusions: Our pattern of findings are consistent with the context-updating theory of P300 and may help identify subtypes of ultrahigh-risk substance abusers who need specialized treatment programs.
Keywords: Event‐related potentials; pattern classifier; principal component analysis; prison inmate; substance abuse treatment; support vector machine.