Using Event History and Survival Analysis to Examine Program Records: An Example from a Juvenile Justice Residential Program

Background: Event history analysis has gained popularity in social work research because of its ability to examine time-to-event data. Originating from biomedical research, where it is called “survival analysis”, it was traditionally used to examine time until the event of death. However, it has been widely adopted across various disciplines to examine any event of interest.  In the social work, studies using event history analysis have examined time until family reunification, recidivism, adoption, and marriage or divorce. Event history analysis is favored over logistic regression for studies where participants do not experience the event or drop out of the study over time (“censored” data), have different lengths of time in the study, or if there are variables of interest that change in value over time.

This presentation illustrates how an event history analysis was used to examine secondary data of program records at a residential program for adjudicated males ages 13-17.  This was conducted as part of a larger mixed-methods study where the main research question was, “What factors contribute to a youth’s time in a residential program leading up to transition?” Residential programs are a challenging area of research due to variations among settings and environments. Recently, authors have recommended that there needs to be more focus placed on which aspects of programs work for different kinds of youth, as there is currently a lack of consensus on best practices for residential settings.

Method: Program records over a 22-month period were examined (n=101 cases, 14,933 obs.). The event of interest was an “unplanned” transition out of the program. The data included daily program records of duration and level of engagement in group, individual, and family therapies. Additional variables examined were race/ethnicity, age, juvenile sex offense status, probation/parole status, and number of home visits. Kaplan-Meier (KM) estimates were first obtained to plot survival curves, examine any significant differences between groups, and select appropriate covariates for model building. Cox proportional hazards models were then used to analyze the data.

Results: Finding indicated that demographic variables influenced the hazard and timing of an event. Findings also indicated that youth with more home visits, higher levels of engagement in family and individual therapies, and more time spent in group therapies had a higher survival rate. An interaction effect of engagement and duration in individual therapies was also found to be significant. Findings were from Cox regression models that included both fixed and time-varying variables.

Implications: The implications of this are vast, in that social service programs can use event history methods to examine administrative records or other time-to-event data. In the example provided, the findings indicated that youth with more home visits and engagement in family therapy had higher survival rates which suggests that family involvement in residential treatment can support a planned transition out of the program. The findings also suggested that demographic variables played a significant role, which is an important practice and research consideration for the field of juvenile justice and residential programs.