Violence-related deaths among people released from incarceration: protocol for a systematic review.

INTRODUCTION: Young people and adults released from incarceration have a risk of dying from violence that far exceeds that in the general population. Despite this, evidence regarding the incidence, elevated risk and predictive factors for violence-related deaths after release have not yet been synthesised. This information is important to inform the development of evidence-based approaches to effectively prevent deaths from violence in this population. This systematic review will synthesise the literature examining the crude mortality rates (CMRs), standardised mortality ratios (SMRs) and predictive factors for violence-related deaths among people released from incarceration. METHODS AND ANALYSIS: We searched key electronic health, social science and criminology databases (MEDLINE, PubMed, PsycINFO, Scopus, Web of Science, CINCH, Criminal Justice Abstracts) for peer-reviewed cohort studies published in English on 14th September 2020. Our primary outcome of interest is violence-related deaths occurring in the community following release from incarceration. We will not restrict study eligibility by year of publication or age of participants. The Methodological Standard for Epidemiological Research (MASTER) scale will be used to assess the quality of included studies. If there are sufficient studies and homogeneity between studies, we will conduct meta-analyses to calculate pooled estimates of CMRs, SMRs or predictive factors for violence-related deaths. If there is a sufficient number of included studies, meta-regression will be conducted to examine the influence of subgroups and methodological factors on the CMRs, SMRs or predictive factors. If the studies do not report sufficient data, or if there is substantial heterogeneity, findings will be presented in a narrative form. ETHICS AND DISSEMINATION: This review is exempt from ethics approval as it will synthesise findings from published studies that have already obtained ethics approval. Our findings will be disseminated through a peer-reviewed journal article, and national and international conference and seminar presentations. TRIAL REGISTRATION DETAILS: This study is registered with PROSPERO (CRD42020209422). © Author(s) (or their employer(s)) 2021. Re-use permitted under CC BY-NC. No commercial re-use. See rights and permissions. Published by BMJ.

Findings from this systematic review will support the development of evidence-informed approaches to prevent violence-related deaths after release from incarceration.

The systematic review will be reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA), and is registered with PROSPERO (CRD42020209422).

The systematic review uses a comprehensive search strategy that was developed in consultation with a research institute librarian and comprises seven health, social science and criminology databases.

To ensure the systematic review contains the highest standard of evidence possible for the outcomes, only published peer-reviewed cohort studies will be included.

Only studies published in English will be included in the systematic review.

Introduction

Despite growing acknowledgement that incarceration is a costly and largely ineffective response to crime,1–3 the number of people incarcerated globally has steadily increased at a rate that exceeds population growth and crime rates.4 Although the prison population in many countries increased rapidly prior to 2000, from 2000 to 2018 the global prison population grew by 24%, with the largest increase in Oceania (87%), the Americas (41%) and Asia (38%).5 It is currently estimated that on any given day there are approximately 11 million adults4 and 410 000 children and adolescents (aged <18 years) imprisoned worldwide.6 The vast majority of these people will be released from custody, most after a relatively short period of incarceration.4

Young people (aged <25 years) and adults released from incarceration are more likely to die from preventable causes than the general population. Previous systematic reviews have shown that adults released from prison are at increased risk of all-cause7 8 and external-cause deaths,8 drug-related deaths9 and suicide.10 Additionally, the Mortality After Release from Incarceration Consortium (MARIC) has recently been established to examine mortality among people who have experienced incarceration internationally using individual participant data meta-analysis methods.11 However, so far only a protocol of the Consortium has been published.11

To our knowledge, no reviews have examined deaths among young people released from youth detention in detail. Two reviews that focussed on the health of young people in detention briefly considered suicide among young people after release, and found that these young people have an elevated risk of suicide in comparison with their community counterparts.12 13 Cohort studies from high-income countries, such as the USA14 15 and Australia,16 have found that young people released from youth detention are at increased risk of all-cause and external-cause deaths. Given the large and increasing number of adults and young people being incarcerated globally, there is a strong imperative to reduce excess mortality among these populations.

Incarceration disproportionately impacts people from disadvantaged communities, such as those who are socio-economically deprived, from marginalised ethnic groups, or experiencing unemployment or homelessness.7–10 Incarceration may compound this existing disadvantage by further disrupting social support systems, employment, education and housing security.17 18 Experiencing these social and economic factors can lead to further social exclusion, poor health and an increased risk of death.19 Consistent with this, the prevalence of complex, co-occurring health conditions (eg, violence and abuse, mental health and substance use disorders, neurodevelopmental disabilities, chronic diseases and blood borne viruses) is higher among people who experience incarceration than in the general population.13 20 As such, addressing the excess morbidity and mortality of people released from incarceration is an important health equity issue and will likely reduce health inequalities at the population level.

Young people14 and adults8 21 released from incarceration have a risk of dying from violence that far exceeds that in the general population. Zlodre and Fazel8 considered violence-related deaths in their 2012 systematic review of all-cause and external-cause mortality, and found that people released from incarceration were between 3 and 10 times more likely to die from violence compared with the general population. However, this study did not calculate pooled cause-specific estimates of crude mortality rates (CMRs) or standardised mortality ratios (SMRs) for violence-related death, possibly due to the limited number of included studies that examined this cause (n=5/18). A preliminary search conducted while developing the search strategy for this review identified several new studies in this area14 22–30 that have been published since Zlodre and Fazel’s review was conducted almost a decade ago.

To our knowledge, no reviews have been conducted that have examined the predictors of violence-related death after release, or have examined violence-related death among young people released from incarceration. While Zlodre and Fazel’s8 review included some studies examining young people released from youth justice detention,16 31 this setting was not an explicit part of the search strategy, and as such some potentially relevant studies may have been missed. A comprehensive review explicitly searching for violence-related deaths among young people released from youth justice detention is needed, given emerging evidence that these young people have a risk of dying from violence that far exceeds that of their counterparts in the general population.14

To our knowledge, the evidence on violence-related deaths among young people and adults released from incarceration has not been synthesised at a global level. As such, the global epidemiology of violence-related death after release from custody remains poorly understood. This knowledge is an important prerequisite to the development of evidence-informed approaches to prevent violence-related deaths in these populations. By conducting a systematic review, we aim to: (1) synthesise the peer-reviewed literature examining the CMRs, SMRs and predictive factors for violence-related deaths among people released from incarceration; and (2) calculate pooled estimates of these measures.

Methods and analysis

This protocol is reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis for Protocols (PRISMA-P)32 and is registered with PROSPERO (CRD42020209422).

Eligibility criteria

Participants

We will include cohort studies examining people of any age who have been released from a youth justice detention facility or adult correctional institution (including both prisons and jails in the USA). Studies of people incarcerated outside of the criminal justice system (eg, in immigration detention) will be excluded.

Outcome measure

Our primary outcome will be violence-related death occurring in the community following release from incarceration. Consistent with the World Health Organization’s (WHO) definition of interpersonal violence, we will define violence-related deaths as those resulting from the intentional use of physical force or power between two or more people that is not specifically intended to further any political, economic or social objectives of any group or cause.33 Consistent with international literature,31 34 deaths from legal intervention (ie, the killing of a person by a law-enforcement officer acting in the course of duty35) will be included in our definition of violence-related deaths. We will not restrict study eligibility by the timeframe in which violence-related deaths are examined after release from incarceration (eg, within the first year after release). Studies examining deaths exclusively during a period of incarceration, or examining deaths both during and after incarceration (but which do not disaggregate by incarceration status at the time of death), will be excluded.

Study design

We will include published, peer-reviewed cohort studies (prospective and retrospective) of people released from incarceration. Previous reviews of this literature will not be included, as some of the included studies may not meet our eligibility criteria. However, the reference lists of these reviews will be searched to identify additional relevant studies which were not identified by the original search strategy. Study eligibility will not be restricted by year of publication. Only studies published in English will be included.

Information sources and search strategy

We searched key electronic health, social science and criminology databases (MEDLINE, PubMed, PsycINFO, Scopus, Web of Science, CINCH, Criminal Justice Abstracts) for peer-reviewed cohort studies published in English using variants and combinations of search terms relating to incarceration, death and violence on 14th September 2020. The search strategy was developed in consultation with a librarian at the Murdoch Children’s Research Institute in Melbourne, Australia. The MEDLINE search strategy is outlined in table 1. We used a version of Ovid’s recommended observational study search filter36 adapted for each database to identify cohort studies. Reference lists of all included studies will be screened to identify any additional relevant studies.

Table 1

MEDLINE search strategy

1.	(offender* or detain* or imprison* or prison* or custod* or incarcerat* or inmate* or criminal* or (secure adj facility) or (secure adj facilities) or ((youth or juvenile) adj (detention* or delinquen*)) or (detention* adj (centre* or center* or unit or units or facilit*)) or jail* or gaol* or (penal adj (centre* or center* or unit or units or facilit*)) or (correction* adj (institut* or centre* or center* or unit or units or facilit*))).tw, kf
2.	Prisons/
3.	Prisoners/
4.	Juvenile Delinquency/
5.	(former or postrelease* or post-release* or release* or ex).tw, kf
6.	(1 or 2 or 3 or 4) and 5
7.	(cohort* or follow-up or followup or observ* or longterm or long-term or longitudinal* or retrospective* or prospective*).tw, kf
8.	exp Cohort Studies/ or observational study/
9.	7 or 8
10.	(mortalit* or death* or dead or fatal* or lethal* or dying or die* or decease* or decedent*).tw, kf
11.	mortality/
12.	“cause of death”/
13.	fatal outcome/
14.	Death/
15.	mo.fs.
16.	10 or 11 or 12 or 13 or 14 or 15
17.	(violen* or assault* or legal-intervention* or murder* or manslaughter* or homicide* or gun? or firearm?).tw, kf
18.	exp Violence/
19.	homicide/
20.	firearms/
21.	17 or 18 or 19 or 20
22.	16 or 21
23.	6 and 9 and 22

Study selection

Studies identified through the database search will be imported into EndNote X8.237 and duplicates will be removed. The remaining citations will be uploaded into Covidence38 for screening. Titles and abstracts of potentially eligible studies will be reviewed by the lead researcher (MW), with 10% screened by a second researcher. Any uncertainty related to study inclusion will be resolved through discussion with a third research team member. The inter-rater reliability for the title and abstract screening will be calculated using Cohen’s kappa statistic.39 Studies will be coded as either 0=Does not meet eligibility criteria, or 1=Meets eligibility criteria or the full-text article needs to be screened to confirm eligibility. After 10% of the citations have been screened, the eligibility criteria will be reassessed through discussion and consensus between all researchers to ensure that they are relevant to the studies that have been identified. If applicable, the updated eligibility criteria will be used for the remaining screening.

After title and abstract screening is completed, all remaining full-text articles will be independently screened by MW and a second researcher. Again, any conflicts related to study inclusion will be resolved through discussion with a third research team member. Where multiple studies use the same data set, only the study with the longest follow-up period will be included.

Data extraction

Data extraction will be conducted by MW using a pre-specified Excel form developed by the researchers. A second researcher will check the data extraction and amend any errors. The following information will be extracted from the included studies: author and year of publication, geographical location of the study, year/s of study, study design (eg, prospective, retrospective), sample selection techniques (eg, population vs selected samples such as people classified as ‘serious violent offenders’, people who inject drugs), cohort size, length of follow-up, whether the time in subsequent periods of incarceration during follow-up was removed from analysis (ie, interval truncation), attrition, number and proportion of violence-related deaths in the cohort, median/mean age of people in the cohort overall and among those who died from violence, proportion of men and women in the cohort overall and among those who died from violence, proportion of people from a marginalised ethnic group in the cohort overall and among those who died from violence, definition of violence, record source (eg, death, court or police records), CMRs, SMRs and reference population, factors predictive of violence-related death and the measure of association (eg, risk ratio, hazard ratio) and type of incarceration experienced (eg, youth justice detention, prison, jail).

Risk of bias assessment

The Methodological Standard for Epidemiological Research (MASTER) scale40 will be used to assess the quality of included studies. The quality and risk of bias of each study will be assessed by MW, with any uncertainty being resolved through discussion and consensus with the research team. The MASTER scale has some advantages over other quality assessment scales as it has clearly defined domains of bias based in theory and empirical evidence that score studies’ propensity towards bias, independent of the direction or magnitude of bias, relative to the highest scoring study in the review, as opposed to using quality scores as an absolute measure of bias.40 This scale ranks studies based on the number of safeguards against bias present in the study, with a higher number of safeguards indicating a lower probability of bias.40 Quality and risk of bias assessment will be conducted at the study level.

Data synthesis

We will provide a descriptive overview of the included studies, including the study year/s and geographical location, length of follow-up, type of incarceration and definition of violence.

If there is a sufficient number of studies which report the CMRs, SMRs or predictive factors for violence-related deaths after release from incarceration, we will conduct meta-analyses. If the CMRs are not reported in an included study, we will calculate them using the method outlined in Zlodre and Fazel,8 using the number of deaths and total person-years at risk or using median duration of follow-up to estimate total person-years. Heterogeneity between studies will be assessed using the I statistic, with an I value of ≥50% considered indicative of substantial heterogeneity.41

If there is a sufficient number of included studies, meta-regression will be conducted to examine the influence of subgroups and methodological factors on the CMRs, SMRs or predictive factors.42 The following factors may be considered: sex, marginalised ethnic group status (as defined by the included study; for example, Indigenous people/non-Indigenous people), type of incarceration facility, study design (eg, prospective, retrospective), length of follow-up, geographical location of the study, sample selection techniques (eg, population vs selected samples) and whether the time in subsequent periods of incarceration during follow-up was removed from analysis (ie, interval truncation). To examine the effect of study quality on the outcomes, sensitivity analyses will be conducted whereby we restrict analyses to studies rated as high-quality (defined as scoring above the median using the MASTER scale).

If meta-analyses are not possible due to a lack of sufficient data on the CMRs, SMRs or predictive factors in the included studies, or if there is substantial heterogeneity, a narrative synthesis will be conducted. The narrative synthesis will follow existing guidelines,43 which contain four main elements: developing a theoretical model (if relevant), preliminary synthesis of the included studies’ findings, exploring relationships in the data and assessing the robustness of the synthesis.

Patient and public involvement

There was no patient and public involvement in the design of this study.

Ethics and dissemination

This systematic review is exempt from ethics approval as it will be carried out on published studies that have already obtained ethics approval. Our findings will be disseminated through a peer-reviewed journal article, and national and international conference and seminar presentations. We will also provide a plain language summary of our findings to relevant national and international bodies for dissemination.