Dynamic risk factors are modifiable aspects of justice-involved people and their environment that can change over time and are associated with criminal behaviour. Because they are considered changeable, they are therefore typically the focus of rehabilitative treatment (Bonta & Andrews, 2024). Research shows that correctional rehabilitation that targets dynamic risk factors effectively reduces recidivism, and psychometric tools assessing these factors generally show improvements during treatment. However, there is a missing link; current research lacks consistent evidence linking changes on dynamic risk factors during treatment to recidivism. Relatedly, there are few studies assessing within-treatment changes in psychopathology and how they may affect recidivism, though they are commonly assessed alongside dynamic risk factors in correctional settings. This study, involving 164 people with violent offence convictions who completed the Tai Aroha community treatment programme, aimed to address this missing link. A battery of tests measured pre- and post-programme included dynamic risk factor measures, the Anger Disorders Scale (ADS), Psychological Inventory of Criminal Thinking Styles (PICTS), Criminal Attitudes to Violence Scale (CAVS), and the Millon Clinical Multiaxial Inventory (MCMI-III and IV) which is designed to assess personality and clinical psychopathology. Standardised residual change scores were used to account for variance in pre-treatment scores, and Cox regression hazard ratios to test prediction of general and violent recidivism. Most subscales for dynamic risk factor measures and the MCMI-III and IV showed substantial improvements during treatment, however, this did not translate to predicting the probability of recidivism. While several hazard ratios indicated effects in a positive direction, none were significant when adjusted for multiple comparisons, and controlling for actuarial risk did not improve predictive accuracy. There was some evidence of increased socially desirable responding which increased at the end of treatment, this included responses on the MCMI-III and IV consistent with the “Normal Quartet.” Interpretations of these findings and both theoretical and empirical implications are discussed.