Introduction Worldwide more than 30 million people move through prisons each year. Record linkage studies have identified a markedly increased risk of death in ex-prisoners. In order to inform preventive interventions it is first necessary to understand who is most at risk, when and why. Unfortunately, limitations of existing studies have rendered synthesis and interpretation of this literature problematic.

Methods Systematic review of studies using record linkage to explore mortality in ex-prisoners. Based on analysis of >20 studies, we illustrate how methodological limitations and heterogeneity of design, analysis and reporting both hamper data synthesis and create potential for misinterpretation of findings. Using data from a recent Australian study involving 42 015 ex-prisoners and 2329 observed deaths, we quantify the variation in findings associated with various approaches.

Results For example, given the very different age distributions of prisoners and the general population, the all-cause SMR among the cohort was 1.4 (95% CI 1.4 to 1.5) using direct methods and 3.1 (95% CI 3.0 to 3.2) using indirect methods. When the period of observation was constrained to 12 months from any release, the indirect SMR increased to 5.3 (95% CI 4.9 to 5.7). Similarly, when analyses were based on the first occasion of release during the period of observation the CMR was 9.3 (95% CI 8.4 to 10.3) per 1000 py, whereas based on the most recent release the CMR rose to 16.9 (95% CI 15.6 to 18.2) per 1000 py.

Conclusion We conclude with a series of recommendations for future studies, and provide a checklist for optimising study design, analysis and reporting.