Ranapurwala et al. Respond

We appreciate the opportunity to respond to the comments of Sewitch et al. The purpose of our study was to illuminate the high opioid overdose death (OOD) risk in the former North Carolina inmate population by comparing OOD rates among all released inmates in the state with rates in the state’s general population.^1–3 Sewitch et al. suggest that prior substance use disorder is a potential confounder for which we should have adjusted.⁴ However, the purpose of our article was simply to call attention to the high rates of OOD in this large population of formerly incarcerated individuals, not to explore causal relationships. Therefore, we controlled only for covariates that might confound comparisons of rates (age, gender, race, and calendar year of prison release).^1–3 Relationships between determinants of incarceration and OODs are complex and were outside the scope of our descriptive investigation.

We also appreciate the comments by Sewitch et al. regarding potential information bias. We believe that the net potential for information bias was low given that death (thankfully) remains a rare event in this population. We agree that OOD risks increased with time, and thus we adjusted for calendar year in calculating standardized mortality ratios. We clearly defined the follow-up time as being at risk until death, reincarceration, or the end of the study. Thus, complete follow-up simply means following up until one of those three events occurs (see Figure 1 in our article).

Sewitch et al. correctly point out that former inmates who moved out of the state would have contributed person-time to the denominator without contributing OODs to the numerator. This bias would have underestimated the OOD rate among former inmates, and thus the true rates among formerly incarcerated individuals may be even higher than the extremely high rates that we observed in this study.

Finally, we considered the option of a multi-opioid category but believed that the opioid-specific and overall categories we included in our study would be more informative from a prevention standpoint. We caution readers that the sum of death rates from heroin, methadone, other opioids, and other synthetic narcotics will not be equal to total OOD rates because deaths attributable to more than one type of drug will appear in more than one category. The total number of OODs was clearly noted in Table 1 of our article.

Once again, we are grateful to Sewitch et al. for providing the opportunity to expound on these important methodological aspects of our article.