Combining cohorts of prospectively collected and linked data to improve health after release from prison

Abstract

BackgroundWhile people who experience incarceration have remarkably poor health profiles, undertaking research to inform evidence-based responses is complicated by difficulties of recruiting people in prison; high rates of socioeconomic marginalisation, study attrition; and legislative and financial barriers to linked data research. There are many advantages to pooling data from multiple studies involving people who experience incarceration, including greater statistical power and geographic and participant heterogeneity. However, there are also challenges that need to be addressed. MethodsWe combined four prospective cohort studies of adults released from prisons in four Australian states. Pre-release interviews, and validated screening assessments, were linked to primary care, medicine dispensing, hospital, alcohol and other drug treatment services, ambulatory mental health, ambulance, corrective services, and death records. Data were harmonised by team members reviewing variable definitions and categories within subject domains. ResultsThe combined cohort consists of 4,232 adults, including 1,544 Indigenous people and 905 women, with a median age of 31 years. Data linkage will enable a median of 9.3 years of prospective follow-up after release from incarceration. Differences in data structures, coding systems between and within datasets, changes over time, and grouping of records belonging to the same event were also addressed. ConclusionThis combined multi-site cohort study is an example of a complex, policy-oriented data linkage project. It was developed to underpin evidence-based, culturally appropriate interventions, health policy and service development for people who were incarcerated. This presentation will discuss the processes and pitfalls experienced while building a multi-sectoral, multi-jurisdictional data linkage project.