You Only Die Twice

We read with interest the study by Ranapurwala et al.,¹ which found that in the first year after release from prison, former inmates had nearly 11 times the risk of opioid overdose death compared with the general population and that the risk was highest in the first two weeks postrelease. However, we have concerns with the methodology and the conclusions drawn.

We took issue with the lack of a clear exposure definition. Differences in the baseline prevalence of substance abuse disorder between the study groups were not presented, even though, according to the US Bureau of Justice Statistics, incarcerated adults are 12 times more likely to meet substance dependence or abuse criteria than a demographically similar sample of the general population.² Had the prevalence of substance abuse disorder in the general population been comparable to that of the incarcerated population, would a similar long-term risk of opioid overdose death have been observed, or was something unique to the prison and postrelease experience at least partially responsible for the elevated risk? Although incarceration and release serve as natural intervention points for substance abuse disorder, a clear definition of the exposure would have illuminated which aspect of the prison and postrelease experience was considered the putative cause of opioid overdose death, which could give shape to future interventions.

Information bias is a major concern, especially pertaining to the assessment of study outcomes. Some former inmates entered the postincarceration population more than once, which allowed for multiple opportunities to experience opioid overdose death at two weeks and one year. Calculation of (cumulative) person-years in these former inmates was problematic because of the increasing risk of opioid overdose death over time. A more thorough definition of “complete follow-up” would have helped the reader to understand how reentry into the postincarceration population contributed to the calculation of standardized mortality ratios and person-years. Moreover, loss to follow-up or opioid overdose death in former inmates who moved out of state was not captured, resulting in an artificially inflated denominator and an underestimated numerator. Assigning multiple opioid deaths to more than one opioid overdose death category gave the impression that there were more opioid overdose deaths than people who died; a better solution would have been the creation of a multiple opioid death category.

Finally, the regression analysis may have violated the assumption of independence or may not have accounted for the change in the characteristics of the nearly 40% of former inmates who reentered the postincarceration population.