Theoretical models of crime underlying the deterrent approach to crime control often fail to account for the role of mental health in mediating deviance. Nor does this approach account for the role of system responses, unique to a post-apartheid context. There is paucity in the literature on the role of mental health on recidivism in South Africa. This study was therefore designed to determine the relationship between mental health and recidivism among incarcerated youth offenders in South Africa and the role of substance use. The Ecological Systems Theory and the Theory of Planned Behaviour provided the theoretical frameworks for this study. A quantitative exploratory research design was used. 280 youth offenders (male and female, aged 18-35) incarcerated at a Durban Maximum Correctional Service in KwaZulu-Natal that do not have a known psychiatric diagnosis were sampled, using multi-stage random sampling. A self-administered structured questionnaire was used to collect the data. Data were analyzed using SPSS. Several statistical analyses were used in this study to include, descriptive analysis, multiple regression, binary logistic regression, cluster analysis, factor analysis, and chi-square analysis. Descriptive analysis showed that the rate of recidivism (re-offending) among this offender population is 32.4% (n=82), while 67.6% (n=171) of the participants have not recidivated. Descriptive analysis showed that the prevalence of anxiety symptomatology is very high where all 100% (n=280) participants presented with anxiety symptomatology. The prevalence of depression is 99.6% (n=279). The prevalence of appetitive aggression (attraction to criminal behaviour) is 52.1% (n=146). The prevalence of substance use meriting the need for substance dependence assessment is 71.8% (n=201). The prevalence of pro-criminal attitude is 52.1% (n=146). The prevalence of perceived control to execute future criminal behaviour is 55.7% (n=156), and the prevalence of subjective norm influence to executing future criminal behaviour is 54.3% (n=128). Multiple regression analysis showed that anxiety and depression do not significantly predict recidivism (F (2,249) = .096, p>.05), anxiety, depression and appetitive aggression do not significantly predict recidivism (F (3,247) = .361, p>.05), anxiety, depression and substance use do not significantly predict recidivism (F (3,245) = .824, p>.05). Multiple regression of the Theory of Planned Behaviour variables i.e. attitude, subjective norms, perceived behavioural control and intention do not significantly predict recidivism (F (4,237) = .298, p>.05). Binary logistic regression showed that anxiety, depression, substance use and appetitive aggression do not significantly predict recidivism. Similarly, attitude, subjective norms, perceived behavioural control and intention, did not significantly predict recidivism. However, cluster analysis showed that the combination of anxiety and depression disorder, substance use and appetitive aggression do increase the likelihood of recidivism. In this study, Cluster 1, “low risk” constituted 42.6% of the population. This cluster has low levels of anxiety, low levels of depression, moderate appetitive aggression and somewhat low substance use. Cluster 2 “high risk”, constituted 26.5% of the population. This group presents with high anxiety and high depression symptomology, moderate appetitive aggression and high substance use. This cluster presents with a high mental health and high substance use crisis. Cluster 3 “high risk” constituted 17.0% of the sample. This cluster presents with high anxiety disorder, high depression symptomology, moderate high appetitive aggression and moderate high substance use. Similar to cluster 2, this group has a high mental health crisis, with a moderately higher attraction to criminality. Cluster 4 “low risk” consisted 9% of the population. Overall, similar to cluster 1, cluster 4, has low anxiety disorder, low depression disorder symptomology. This group has somewhat low appetitive aggression and moderate substance use. Overall, cluster analysis shows that appetitive aggression plays a key role in distinguishing risk for recidivism among the recidivist and non- recidivist clusters. However, cluster analysis using the variables (attitude, perceived behavioural control, subjective norms) failed to distinguish the likelihood of recidivism between the recidivist and non-recidivist clusters. This study promotes social change by highlighting the need for further research that examines combined factors that increase recidivism. Findings from this study are helpful to health and criminal justice organizations as they indicate the need for strengthening the provision of programs that address mental health screening, diagnosis and treatment as well as programs that address appetitive aggression issues in efforts to mitigate recidivism.