Background

The world prison population size was over 10.7 million in 2021 or 140 per 100,000 of population [1]. However, the prison population is estimated to be more than 11.5 million when we take into account statistical information about prisoners which is unavailable or unrecognised internationally or is missing from published national prison population sizes [1]. Prison population rates vary by country and region. For example, the USA has the highest prison population—over 2 million people, equivalent to a rate of 629 per 100,000 [1]. There are higher rates of mental and physical health problems in prison populations compared to the general population, and substance use disorders are common in people who are committed to prison [2, 3]. There is a risk of disruption to treatment and care and a deterioration in health when former prisoners transition from prison to living in the community [2]. Furthermore, negative health effects may be compounded by post-release experiences of former prisoners including loss of social support, enduring stigma, financial insecurity and difficulties obtaining stable housing [4].

There are concerns about the increased risk of mortality after release from prison and the contribution of drug-related causes to deaths in former prisoners [5,6,7]. A review in 2010 reported that 76% of deaths in the first 2 weeks after release and 59% of deaths within the first 3 months of release were due to drug-related causes [7]. There is a need to examine the range of potential factors that may contribute to the increased risk of drug-related deaths after release from prison, including decreased tolerance following relative abstinence in prison and the concurrent use of multiple drugs [8]. Observational studies investigating the risk of mortality after prison release often use large administrative datasets to link prison and death records. An updated review of the evidence in this area, including the extent of the literature, methodologies, findings and gaps in knowledge is warranted and a scoping review approach has been chosen to map key concepts and summarise evidence in this field. This scoping review updates and maps research evidence in the area of record linkage studies of drug-related deaths among former adult prisoners, and identifies and profiles at-risk former prisoners. The findings are discussed in terms of their contribution to potential interventions and to informing future research and policy. This review was undertaken as part of a work programme in the Administrative Data Research Centre, Northern Ireland and in response to concerns from public health, criminal justice, voluntary and community groups and wider society about prisoner health and well-being in Northern Ireland after release from prison.

Methods

We chose to conduct a scoping review because of the broader scope of our review that included a focus on how the research was conducted and differences in methodologies used among record-linkage studies in this research area. This broader scope was informed largely by the results of previous systematic reviews/meta-analysis regarding reported high levels of heterogeneity [5,6,7]. The methods used to conduct this scoping review have been published previously as a protocol [9] and a summary of the methodology is provided here. This review followed the first five stages of the framework for conducting scoping reviews by Arksey and O’Malley [10] and adhered to the guidance developed by the Joanna Briggs Institute (JBI) and the JBI Collaboration. For example, as recommended by the JBI, the population, concept and context (PCC) guide was incorporated into the scoping review title, research questions and inclusion criteria [11]. In addition, the Preferred Reporting Items for Systematic reviews and Meta-Analyses extension for Scoping Reviews (PRISMA-ScR) checklist and guidance was used to structure and report this review [12]. This scoping review was structured to meet the requirements of the PRISMA-ScR checklist. A completed PRISMA-ScR checklist (used to report this work) has been provided as supplementary material in this scoping review.

Stage 1: identifying the research question

The following questions were addressed by the scoping review:

  • 1. What is the scope of the literature on record linkage studies of drug-related deaths among former adult prisoners who have been released to the community?

  • 2. How is research conducted on this topic?

  • 3. What methodologies are used?

  • 4. What are the findings in relation to mortality?

  • 5. Where are the knowledge gaps on this topic?

Stage 2: identifying relevant studies

In order to summarise the most recent evidence, the start date of 2011 was chosen for this scoping review. Four bibliographic databases (MEDLINE, EMBASE, PsychINFO and Web of Science) were searched for studies from January 2011 to September 2021 using keywords and index headings (modified as required for each database). The search terms related to ‘mortality’, ‘drugs’ and ‘ex-prisoner’ (and their variants). The review focused on drug-related deaths and as such, the search strategy included a broad range of terms including substance-related disorders, drug overdose and drug misuse. The full list is found in appendices 123, and 4. The search strategy for MEDLINE was developed by JAC and MD with assistance from the Subject Librarian for the School of Medicine, Dentistry and Biomedical Sciences in Queen’s University Belfast, and was published with the review protocol [9]. JAC and MD developed search strategies for EMBASE, PsychINFO and Web of Science, and all search strategies used in this review have been provided as supplementary material (appendices 123, and 4). The reference lists of included studies were screened by JAC to identify any additional publications. Due to the absence of resources for translation, all search strategies were limited to publications in the English language. There was no geographical restrictions on studies.

Stage 3: study selection

All bibliographic database searches were performed by JAC on 15th September 2021 and the results were combined in Endnote Reference Manager where duplicate publications were subsequently removed. JAC and IO independently screened all titles and abstracts using the pre-defined inclusion criteria and excluded any non-eligible publications. Publications were screened using criteria defined in Table 1. Publications were screened in full if an abstract was not available and/or there was uncertainty over inclusion. Subsequently, JAC and IO independently screened full publications for inclusion and any discrepancies between JAC and IO regarding eligibility were resolved in a discussion between JAC and MD during which a unanimous decision was made. No authors of publications were contacted during this process.

Table 1 Modifications made to inclusion and exclusion criteria as part of review

Stage 4: charting the data

A draft charting form was piloted as part of the protocol development stage. As part of the review process, the charting form was retested by JAC and EP (the final data charting form used is provided in appendix 5). JAC independently extracted information from all included publications using this data charting form. The accuracy and consistency of the recorded information was checked by using a second reviewer (EP) to independently extract information for a proportion of the included publications (n = 14) and resolving any discrepancies via discussion by team members. In addition, JAC and MD met weekly and discussed the studies in the review particularly for their fit with the pre-specified inclusion criteria and the charting procedure.

Stage 5: collating, summarising and reporting the results

Information was extracted from the charting forms and entered into Microsoft Excel sheets for data management and analysis. Data in the Microsoft Excel sheets were subsequently cleaned and extracted information was summarised. The data were analysed and presented in a format that was designed to answer the scoping review questions and organised according to the main conceptual categories including methodology, key findings and gaps in the research. All descriptive tables and figures for this review were prepared using these data contained in the Microsoft Excel sheets. We reported mortality outcomes following release from prison in terms of, for example, crude mortality rates (CMRs) and standardised mortality ratios (SMRs). Where possible, age, sex/gender and race/ethnicity, time period examined after prison release and information on specific drugs were reported in relation to drug-related mortality. SMRs for drug-related deaths after release from incarceration were pooled statistically, where possible. The log SMR was determined as well as the Standard Error (SE) of the log SMR from the published SMR and confidence intervals (CIs). In meta-analysis, the consistency of effects across studies should be assessed [13]. The random-effects DerSimonian-Laird model was used. In STATA version 16.1 [StataCorp, College Station, Texas, USA], the meta command was used to compute effect sizes and summarise data and produce forest plots. The heterogeneity was measured using the I2 squared statistic and testing using a formal chi-squared test for heterogeneity. Meta-analyses are not a usual feature of the methodology of scoping reviews [10, 11], however, exploratory meta-analyses were conducted following this scoping review to deepen the level of critical analysis by, for example, assessing in a quantitative way, the consistency of effects. Meta-analyses were performed posteriori and were not planned in the study protocol [9].

Patient and public involvement

Our empirical study of prisoner post-release mortality and this scoping review were initiated in response to concerns about the increasing number of drug-related deaths generally from the UK Chief Medical Officers (CMOs) including the CMO for Northern Ireland. We continue to consult with, and involve, key prison health care staff including the Clinical Director of Healthcare in Prisons in Northern Ireland in our ongoing programme of prison health research (co-author of this paper).

Results

Study selection

The search strategy identified a combined total of 4397 publications across four bibliographic databases. Using the Endnote duplicate tool, 717 duplicate publications were removed. A total of 3680 publications were screened by title and abstract; 109 publications were deemed to meet eligibility criteria and full-text publications were screened by two authors (reviewer 1 fully screened 105 publications and reviewer 2 fully screened 36 publications i.e. there was some overlap of screened publications). Authors noted that some remaining duplicate articles were among the publications excluded at this stage (9 remaining duplicates removed). There was agreement between reviewers to include 23 publications and exclude 49 publications. There was disagreement or uncertainty between reviewers about 28 publications and these publications were fully screened by reviewer 3 and resolved via discussion with reviewer 1; 25 of these 28 publications were included. A total of 48 publications were included at this stage and the reference lists of included publications were screened, resulting in the addition of one further publication. Four publications were excluded during the data extraction stage after discussion between reviewer 1 and reviewer 3. The reasons for exclusion of publications were as follows: summary of another paper included in review [14], no drug-related deaths [15], not people released from prison [16] and an ambiguity over whether a study included individuals who had been recently released from, or admitted to, jail, prison or a detention facility [17]. Following this review process, a total of 45 publications were included. A flow diagram for each stage is presented in Fig. 1.

Fig. 1
figure 1

Flow diagram showing the number of publications at each stage of the review process

Study characteristics and methods for included studies are shown in Tables 2 and 3, respectively.

Research questions

The data were analysed in a format that was designed to answer the review questions, as presented below.

  • 1. What is the scope of the literature on record linkage studies of drug-related deaths among former adult prisoners who have been released to the community?

    The included studies (n = 45) were published across 25 different journals. The five most common journals that published studies in this area were Addiction (n = 12), Drug and Alcohol Dependence (n = 7), American Journal of Public Health (n = 2), JAMA Psychiatry (n = 2) and Public Health Reports (n = 2) (Table 2). The remaining included studies (n = 20) were published in 20 different journals. The geographical distribution of the included studies, by location of the custody setting, shows a total of 9 countries/regions (appendix 6). The most common locations were the USA (n = 24) and Australia (n = 7). Other locations were Canada, Denmark, Norway, Sweden, Taiwan and the UK. One publication included both USA and Australia by way of comparing cohorts [18]. The search strategy included January 2011 to September 2021 in order to summarise the most recent evidence (the distribution of publications across this time period is presented in appendix 6).

    This scoping review focused on studies of mortality risk during the time period after release from incarceration—the number of years and months following release from incarceration were provided in 60% of studies (n = 27) (Table 3). Information about incarceration dates was provided in half of those studies without release dates (n = 9/18). The earliest reported period of release was 1988 to 2002 (in a study by Kinner et al. 2011) and the study with the most recent year analysed data from 2018 [19]. The studies with the longest release period covered a total of 16 years, which included releases between 2000 and 2015 [20, 21] and the study with the shortest time period was a follow-up of all prisoners released on a specified date in July 2007 [22]. For studies which provided information about incarceration dates (rather than specified release dates), two [23, 24] used a single incarceration date (specified in June 1991) as the index date for follow-up, whereas all other studies used either a single year or range of years. Although some of the key questions in this scoping review were around methodology, the extent to which included studies reported key characteristics varied. For example, in all studies, sex or gender was reported in some format throughout various sections of the paper, whereas age and race or ethnicity were less well documented. Approximately 31% of included studies did not report the age of their study population (n = 14) and 31% did not report race or ethnicity in any format (n = 14) (Table 2 and appendix 7).

  • 2. How is research conducted on this topic?

    The most commonly reported study designs were retrospective cohort studies (n = 16), prospective cohort study studies (n = 5) and nested case–control studies (n = 3) (Table 2). Several included studies did not state the study design (n = 7). The type of data used by included studies to investigate prison release and mortality are shown in Table 3. Prison data was often obtained from national prison or criminal registries, department of correction/correctional services or records, or records from single prison systems, for example, individuals released from one county jail. Mortality data used by included studies to determine drug-related death was often obtained from the national death index or national death registries or regional (for example, USA State) death records.

    Study parameters such as number of people released, number of releases (as an individual may have been committed and released more than once during the study period) or person-years of follow-up are reported in Table 2, and included studies differed in size. The study with the largest number of people reported that 229,274 were released over a 16-year time period (between 2000 and 2015) – data from this retrospective cohort study of people released from prison was analysed and presented in two separate papers [25, 26].

    The review found that the terminology that was used in the included studies to report death outcomes varied; the most commonly used terms were overdose deaths, opioid overdose deaths, opioid-related overdose deaths, drug-related deaths and similar, less frequently used variants including death from drug-related infections, drug toxicity and contributing substance use–related cause of death. Approximately 64% of the published studies (n = 29) used the codes from the International Classification of Diseases (ICD) to describe cause of mortality (Table 3). Data linkage (or similar meaning terms) was an inclusion criterion in this scoping review. The methods used for data linkage included probabilistic linkage/matching/score (n = 17), deterministic linkage (n = 2), deterministic and probabilistic linkage (n = 2), personal identifiers or unique identification linkage in methods (n = 11), and combinations of name, date of birth, sex or gender and race or ethnicity (n = 5). In several studies (n = 8) linkage methods were not stated (Table 3).

    Only four studies reported the use of a quality assessment checklist or technique. All four of these studies used the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) checklist [20, 27,28,29]. The STROBE guideline provides a checklist of items about the planning and conduct of epidemiological observational studies and best practice requires researchers and authors to include a completed checklist in their reports and papers. Only one study, Chang et al. 2015, provided a copy of the STROBE statement [27].

  • 3. What methodologies are used?

    The included studies examined various time periods after release from prison and it was common for studies to examine more than one time period (n = 19) (Table 3). Commonly investigated time periods included the first two weeks after release (n = 11), the first month (including studies examining intervals up to one month e.g. 1–2 weeks and 3–4 weeks) (n = 14) and the first year after release (including studies examining intervals up to one year e.g. up to 4 weeks, after 4 weeks up to 6 months, after 6 months up to 1 year, all follow-up to 1 year) (n = 16). During follow-up any re-committals would reduce the at-risk period for mortality as the individual would be in custody rather than in the community, and 60% of included studies took into consideration person-time at risk in the community time and during any subsequent re-incarcerations (n = 27). The methods used for dealing with repeated incarcerations included person-time being excluded at re-incarceration i.e. person-time was calculated from the day of release from prison until re-incarceration. Another approach used excluded time during a subsequent incarceration, whereas the time between the next release and death, another incarceration, or the end of the study was included. Other methods used the most recent prison release date/index release was that closest to death or calculated person-time following every release during follow-up until death, re-incarceration or the end of study follow-up. In one study, time periods of 4 weeks and 1 year from the date of first release were used regardless of reimprisonment within these time frames, therefore this method did not exclude time whilst in custody [30]. Another study, coded each re-incarceration after the index release date as an ‘additional post-release booking’ to determine any effect on survival [25].

  • 4. What are the findings in relation to mortality?

    A summary of the drug-related mortality outcomes reported in the included studies is provided in appendix 8. Studies reporting SMR by characteristics and time after release are shown in Tables 4 and 5, respectively. CMRs reported by time after release are shown in Table 6. The pooled SMRs across the included studies, grouped by time periods examined after release, are shown in Table 7. The pooled drug-related SMR was 6.99 (95% CI 4.13–11.83; I2 = 99.14%) for any time after release (5 studies), 27.07 (95% CI 13.32–55.02; I2 = 93.99%) for the first two weeks (4 studies), 10.17 (95% CI 3.74–27.66; I2 = 83.83%) for the first 3–4 weeks (3 studies) and 15.58 (95% CI 7.05–34.40; I2 = 97.99%) for the first 1 year after release (3 studies) (Table 7). In all studies the SMR was significantly above 1, but in some, this was much higher than others. These results suggest differences in each study. There was a high level of heterogeneity and this must be considered when interpreting the pooled estimates as it may reflect substantial inter-study differences in study design, setting or population. CMRs were not pooled for specific time periods due to a low number of studies reporting these findings. Forest plots are provided in appendix 9. A summary of variables investigated in included studies is provided in appendix 10.

  • 5. Where are the knowledge gaps on this topic?

    Our review suggests that knowledge gaps in this topic revolve around methodological differences in study design and limitations in the capacity to synthesise the evidence. Only a limited number of the 45 eligible studies were suitable for inclusion in the pooled analyses for SMRs – there is a need for increased consistency in the use of observational study methodology about mortality among former prisoners. More rigorous reporting of characteristics of former prisoners would allow subgroup analyses to profile those people most at-risk after prison release. For example, reporting characteristics of former prisoners, in terms of age, married or single, health etc. would give a fuller presentation of the results. Our review captured studies from USA, Australia, Canada, Denmark, Norway, Sweden, Taiwan and the UK and pointed to a distinct lack of studies undertaken in low and middle income (LMIC) countries. Clearly, therefore, there is a need for studies to be conducted of this population in LMIC countries in order to understand the extent of global drug-related mortality among people following release from prison.

Table 2 Key characteristics of each included study
Table 3 Key methodological features of each included (record-linkage study)
Table 4 Standardized mortality ratios (SMRs) by characteristics
Table 5 Standardized mortality ratios (SMRs) by time after release
Table 6 Crude mortality rates (CMRs) reported by time after release
Table 7 Pooled standardized mortality ratios (SMRs) across the included studies, grouped by time periods examined after release

Discussion

This scoping review maps and summarises research evidence from record linkage studies about drug-related deaths among former adult prisoners and the extent to which drug-related causes contribute to post-release prisoner mortality. The research questions in this review focused on the scope of the literature, methodologies used in observational data-linkage studies and the most recent findings in relation to mortality (published between 2011 and 2021). This scoping review found an increased risk of drug-related death after release from prison, particularly in the first two weeks after release, although the drug-related mortality risk remained elevated for the first year among former prisoners. However, despite this review identifying 45 relevant publications, only a limited number of studies were included in the pooled analyses for SMRs due to differences in study design (for example, time periods examined after release) and methodologies used which has significantly limited evidence synthesis. In addition, we found high levels of heterogeneity in our pooled analyses meaning that our interpretation of the pooled estimates is more hesitant.

The findings of our scoping review are supported by previous literature mapping this topic. A recent scoping review by Mital et al. described the relationship between incarceration history (custody in a jail or prison facility) and opioid overdose in North America, including 18 studies published between 2001 and 2019, with the scoping review methodology following guidance by Levac et al. [63, 64]. The review reported four important findings; (1) an increased risk of opioid overdose among formerly incarcerated people; (2) an increased risk of opioid overdose was associated with some demographic, substance use, and incarceration-related characteristics (including substance use disorders and mental health issues); (3) incarceration history was identified as a risk factor for opioid overdose among individuals who inject opioids and (4) opioid overdose was suggested as the leading cause of death in people who have been formerly incarcerated [63].

The results of this review in terms of an increased mortality risk after prison release concurs with the findings previously published in systematic reviews and meta-analyses. It is concerning that post-release mortality risk is high. Collectively, the reviews appear to indicate that post-release mortality has persisted over time. For example, a previous systematic review pooled SMRs from studies which used record linkage methods to examine deaths in ex-prisoners between 1998 and 2011, reporting SMRs for drug-related death of 32.2 (95% CI 22.8–45.4) for < 1 year, 26.2 (95% CI 6.4–107.3) for ≥ 1 year and 27.3 (95% CI 9.8–76.0) for any time after release [6]. A separate systematic review of publications between 1980 and 2011 explored the literature on studies of mortality in released prisoners using linkage of prisoner and mortality databases, and reported all-cause SMR, ranging from 1.0 to 9.4 in males and from 2.6 to 41.3 in females [5]. Furthermore, similar to our findings, where the drug-related death risk was highest in the first two weeks after release; a meta-analysis of mortality during the 12 weeks after prison release reported an increased risk of drug-related mortality during the first 2 weeks after prison release compared to the subsequent 10 weeks (however, the mortality risk was elevated during the first 4 weeks) [7].

Kinner et al., Merrall et al. and Zlodre and Fazel, all reported high levels of heterogeneity, for example between countries [7], in study design [5, 6], and in analysis and findings of publications [6]. In our scoping review, differences in the study design, methodologies and findings of included studies limited the degree to which studies could be synthesised meaningfully. The included studies examined various time periods after release from prison and this limited the number of studies included in the pooled analyses in this review. Differences were also found in study design, i.e. retrospective cohort, prospective cohort and nested case–control study designs, but differences were also found in methodologies, for example in the approaches used for determining the time at-risk during follow-up. During re-incarceration, re-committals would reduce the at-risk period for drug-related mortality as the individual would be in custody rather than in the community. Other differences included various types of data used by included studies to determine mortality (for example, national and regional death records) and prison release (for example, national prison registries and single prison records). The type and geographical distribution of death records used in the study would likely have affected the number of missed deaths, for example if mortality records covered one country and the death occurred outside of this border. The study size differed in the included publications and the size of the prison population(s) and location(s) of prison(s) would affect the generalisability of the study findings. The terminology used to describe or define drug-related deaths differed between studies, with some studies using ICD codes and definitions. The definition used to describe drug-related deaths may have an effect on the findings, for example combining multiple ICD codes for drug-related deaths in the definition would be more inclusive compared to very specific definitions.

The reporting of characteristics of individuals varied between included studies. Gender/sex was reported in all studies, but age and race/ethnicity were only reported in one-third of papers, making it difficult to contextualise the findings. Approximately 9% of included studies stated the use of a quality assessment checklist or technique and in only one study was a copy of the STROBE statement provided as an appendix. Adequate reporting of research facilitates the assessment of published studies and following recommended guidelines in the reporting of research allows rigour and transparency in the process. In summary, this review suggests a need for a more consistent methodology and rigorous reporting of observational studies about mortality among former prisoners.

Study strengths and limitations

Although meta-analyses are not consistent with the methodology of scoping reviews [10, 11], this scoping review included exploratory meta-analyses. We conducted a scoping review (rather than a systematic review) because we wanted to scope and search broadly and at the same time deepen the level of critical analysis where there was an opportunity to do so. For example, the review included a focus on how record linkage research was conducted, and on differences in methodologies that were used among record-linkage studies in this research area. This broader focus stemmed largely from the results of previous systematic reviews/meta-analysis that reported high levels of heterogeneity [5,6,7]. Our scoping review summarised the methodologies and findings narratively, and the accompanying meta-analyses added to this narrative by explaining high levels of heterogeneity in our pooled analyses, and showing differences across studies. This review recommends that a more consistent approach to methodology and reporting is followed in the future. This scoping review has several strengths. This is the first scoping review of record linkage studies about drug-related deaths among former adult prisoners. The methods followed the first five stages of the framework for conducting scoping reviews by Arksey and O’Malley [10] and adhered to the guidance developed by the Joanna Briggs Institute (JBI) and the JBI Collaboration. The methods for this scoping review were previously published in a protocol allowing transparency and forward planning [9]. Modifications from the original protocol have been stated in this review, and in a deviation from that stated in the protocol, data were independently extracted by one reviewer, with a proportion of papers checked by using a second reviewer due to time constraints. Using this approach allowed a check of the accuracy and consistency of the recorded information. There are some limitations to this review, the search strategy was limited to publications available in English due to resources for translation and the review did not include a search of the grey literature which may limit the interpretation of the findings.

Future research and policy

This scoping review focused on former prisoners. However reviews on other prisoner groups, such as prisoners on remand or probation, would be of benefit. Prisoners have higher rates of mental and physical health problems compared to the general population, and substance use disorders are common in people who are committed to prison. Research on mental and physical health conditions, substance use disorders, and physical and mental ill health comorbidity in people released from prison could help profile risk after release. As part of this scoping review process, authors identified one randomised controlled trial in Australia and one randomized controlled pilot trial after prison release in England, but these publications were excluded at the full screening stage [65, 66]. The NALoxone InVEstigation (N-ALIVE) pilot trial tested feasibility measures for randomized provision of naloxone-on-release to eligible prisoners and demonstrated the feasibility of recruiting prisons and consenting of prisoners [65]. A randomised controlled trial of a service brokerage intervention for adult former prisoners involved an intervention group receiving a personalised booklet with their health status and appropriate community health services, and telephone contact for each week in the first month after release to assess any health needs and health service utilisation (control arm received usual care) [66]. A separate review of trials in former prisoners after release would provide evidence to help guide the development of future research in this area.

This review was undertaken in response to concerns from public health, criminal justice, voluntary and community groups and wider society about prisoner health and well-being in Northern Ireland after release from prison. Our findings suggest the need for formalised joined-up working and interagency collaboration regarding the way in which people released from prison are supported, and an ongoing review and consideration of interventions and service responses designed to reduce drug-related deaths among this group, including novel service responses such as overdose centres, transition clinics and drug consumption rooms [67, 68]. It is clear from the available evidence that the transition from prison to community is an at-risk period and there is need for sustained joined-up service responses and support that help people released from prison to negotiate this transition.

Conclusions

This scoping review found an increased risk of drug-related death after release from prison, particularly in the first two weeks after release, although the drug-related mortality risk remained elevated for the first year among former prisoners. Our results are of concern as we show that post-release mortality risk is still high despite similar findings having been reported in the literature more than a decade ago. This scoping review has detailed examples of differences in study design and methodology in included studies which has significantly limited evidence synthesis. This review suggests a need for a more consistent methodology and rigorous reporting of observational studies about mortality among former prisoners.