Survival Analysis Assessing the Effects of Prison Programming on Recidivism

Background/Purpose: Incarcerated individuals have fewer employment opportunities, lower education levels, and higher victimization rates than the general populace. Prison programs are treatment and skill-building services that address a range of these educational, vocational, and psychosocial needs. Although programs are implemented in prisons, only a limited number of prisoners receive such services during their incarceration. Research into correctional programming can indicate the services that are effective in reducing recidivism rates, and should subsequently be expanded to serve a greater population. Such research can also indicate what programs are not working.

Methods: A retrospective cohort study was developed based on secondary data from the North Carolina Department of Public Safety (NCDPS). The sample (n = 42,590) included all individuals released from North Carolina prisons from 2000 to 2001, and from 2004 to 2005. The collection of recidivism information ended in June, 2009. Four models were assessed using Cox proportional hazards regression. The first model investigated the hazard function between nine-year recidivism rates, vocational programming (VP), educational programming (EP), and cognitive behavioral intervention (CBI). The second model investigated the hazard function between recidivism and substance use treatment (SUP) among individuals with identified substance use histories (n = 9,215). The third model investigated the hazard function between recidivism and mental health treatment (MP) for individuals identified as having a severe mental illness (n = 3,589). The final model investigated the hazard function between recidivism and sex offender treatment (SOP) for all individuals convicted of a sex crime (n = 1,481). Each model controlled for the effects of race, sex, history of violent offenses, age, and length of sentence. Effect size in a Cox regression is measured using hazard ratios. This ratio explains the relationship between an exposure variable (prison program) and survival time.

Results: The sample had a 46.2% occurrence of recidivism within nine years. When controlling for covariates related to recidivism, the completion of VP indicated an 8% lower hazard for recidivism (HR = .92, p = .001). When controlling for other variables in the model completion of EP led to 13% decrease in hazard (HR = .87, p < .001). Completion of CBI led to no significant change in the hazard ratio, when controlling for other variables in the model. After controlling for other variables, the completion of SUP, MP, and SOP were not associated with statistically significant decreases in the hazard of recidivism within the subpopulations to which each program type respectively applies.

Conclusions/Implications: The results of the Cox regression analysis suggest that both EP and VP reduce the hazard of long-term recidivism. Conversely those who completed focusing on mental health, SUP, SOP, and CBI programs did not seem to have significantly lower the recidivism hazard than those who did not receive treatment. These results suggest that programs which build marketable skills for socioeconomic advancement are more effective at reducing recidivism than programs that emphasize clinical goals. This study does not, however, provide evidence that clinical interventions are not also beneficial and necessary programs.