• Open Access

Risk of death for young ex-prisoners in the year following release from adult prison

Authors

  • Kate van Dooren,

    1. The Queensland Centre for Intellectual and Developmental Disability (QCIDD), The University of Queensland
    Search for more papers by this author
  • Stuart A. Kinner,

    1. School of Population Health, The University of Melbourne, Victoria; Murdoch Childrens Research Institute, Victoria; School of Medicine, The University of Queensland; School of Public Health and Preventative Medicine, Monash University, Victoria
    Search for more papers by this author
  • Simon Forsyth

    1. School of Population Health, The University of Queensland
    Search for more papers by this author

Correspondence to: Dr Kate van Dooren, QCIDD, The Queensland Centre for Intellectual and Developmental Disability, The Mater Hospital, South Brisbane, Queensland 4101; e-mail: k.vandooren@uq.edu.au

Abstract

Background : In the community, all-cause mortality rates among those younger than 25 years are considerably lower than those of older adults and are largely attributable to risk-taking behaviours. However, given the unique health profiles of prisoners, this pattern may not be replicated among those leaving prison. We compared rates and patterns of mortality among young and older ex-prisoners in Queensland, Australia.

Methods : We linked the identities of 42,015 persons (n=14,920 aged <25 years) released from adult prisons in Queensland, Australia with the Australian National Death Index. Observations were censored at death or 365 days from release. We used Cox proportional hazards regression to explore associations between mortality and demographic and criminographic characteristics. We used indirect standardisation to compare rates of all-cause mortality for both age groups with those for the general population. We calculated proportion of deaths across specific causes for each age group and relative risks for each cause for young versus older ex-prisoners.

Results : Being young was protective against death from all causes (AHR=0.7, 95% CI 0.5–0.8); however, the elevation in risk of all-cause death relative to the general population was greater for those aged less than 25 years (SMR=6.5, 95% CI 5.3–8.1) than for older ex-prisoners (SMR=4.0, 95% CI 3.5–4.5). Almost all deaths in young ex-prisoners and the majority of those in older ex-prisoners were caused by injury or poisoning.

Conclusions : Young people are at markedly increased risk of death after release from prison and the majority of deaths are preventable.

Risk of death is greatly elevated among ex-prisoners compared with the general population.1–6 Although many deaths are drug-related7 or the result of suicide,8,9 little is known about risk and protective factors for death in this population. To inform targeted preventative interventions it is first necessary to identify those at greatest risk of death.2,7

Internationally, young people are defined as those aged less than 25 years.10 In most high-income countries, mortality rates among young people are relatively low compared with older adults. Mortality among this group is associated primarily with risk-taking rather than chronic disease.11 However, given the high prevalence of risky health-related behaviours among ex-prisoners,12 patterns of mortality seen in the community may not be replicated among those leaving prison. For example, in the general population, age-standardised mortality rates are greater for young males than for young females,11 but among those released from prison all-cause standardised mortality ratios are higher for females than for males,5,13,14 probably reflecting sex differences in the prevalence of risky substance use15,16 and mental illness17 in this population.

Although a number of studies have documented elevated mortality among young people leaving juvenile facilities compared with the age-matched general population,18,19 few have investigated mortality among young people following release from adult prison. Methodological differences among studies comparing mortality rates across age groups4,9,20–22 make it difficult to draw conclusions that are relevant across settings. Young people leaving adult prisons are a highly marginalised but largely forgotten group.23

This study used data from a large, retrospective data linkage study of ex-prisoner mortality in Australia to: 1) identify demographic and criminographic risk and protective factors for mortality, including age; 2) measure risk of death across specific causes for young and older ex-prisoners; and 3) compare mortality rates among young and older ex-prisoners with those of the age- and sex-matched general population.

Methods

We linked the identities of all persons with at least one recorded release from adult prison in Queensland, Australia, from 1 January 1994 to 31 December 2007 with the Australian National Death Index (NDI). All natural and unnatural deaths over the study timeframe were observed. Given the high level of recidivism among ex-prisoners in Australia,24 we used the index release (the first occasion of release during the study period) rather than most recent release, as using most recent release would bias the analysis towards deaths.7 Young ex-prisoners were defined as those aged less than 25 years at the time of index release. Queensland Corrective Services (QCS) prisoner data were linked with the NDI using probabilistic matching on prisoners’ names and any aliases, sex, date of birth, and date of last release from prison or most recent admission to a community order. Probabilistic linkage between Australian corrective services data and the NDI is sensitive and highly specific25 and the inclusion of aliases in the linkage process improves sensitivity without adversely affecting specificity.26

The study received ethics approval from The University of Queensland's Behavioural and Social Sciences Ethical Review Committee, QCS Research Committee and the Queensland Health Research and Ethics Advisory Unit.

Data sources

QCS administers all adult correctional facilities in Queensland and provided data on sentence details and basic demographic and criminographic information for all individuals released from prison during the study period. Variables included age, sex, Indigenous status, marital status, number of incarcerations, duration of index incarceration, subsequent incarcerations within 365 days of index release and history of a drug conviction.

The Australian Bureau of Statistics (ABS) supplied a count of deaths in Queensland by cause, stratified by age category and sex, for 1996, 2001 and 2006,27 as well as Queensland census population data.28 The latter were used to derive reference population crude mortality rates (CMRs) to calculate standardised mortality ratios (SMRs) for ex-prisoners, adjusted for age and sex.

The NDI records date and state of death, and International Classification of Disease 9th (ICD-9) and (from 1996) 10th Revision29 (ICD-10) codes for the underlying and contributing causes of death. Groupings of ICD-10 death codes are not mutually exclusive and one death may be included in multiple cause of death groups (e.g. ‘drug-related’ and ‘cardiovascular’).1,26 For the purposes of this study, ICD-9 and ICD-10 codes were categorised into groups reflecting broad or related causes of death, using a system proposed by Randall et al.30 (Complete coding can be found at http://hisdu.sph.uq.edu.au/marc/) Codes were not assigned to deaths (n=206) that had ongoing investigations or deaths that occurred during 2008, due to standard delays involved in death registration, coding and compilation by the ABS.31

We censored person time at 365 days after index release, consistent with evidence that risk of death is elevated during this period.9,20,32 For persons with subsequent incarcerations in the 365-day period, subsequent time in prison was deducted from time at risk and deaths in prison were excluded. Day of release was counted as a whole day in the community.

Analysis

To address Aim 1, we used Cox proportional hazards regression to explore associations between mortality and demographic and criminographic characteristics including: age less than 25 years; male sex; identifying as Indigenous; being unmarried; number of previous incarcerations; length of index incarceration; having a history of drug convictions; and being placed on a community supervision order following index release. Variables with a significant association (p<0.05) in univariate analyses were included in a multifactorial logistic regression model. Through a process of stepwise backwards elimination we removed variables that were not significant from the multifactorial model, so that all variables in the final model were significant at p<0.05.

Regarding Aim 2, we calculated relative risks for each cause of death category, to compare the risk of death being due to specific causes for young versus older ex-prisoners. We calculated the potential years of life lost33 for the remaining life expected at the age of death using the mean of the age and sex specific potential additional life expectancy for 1994 to 2006 published by the Australian Bureau of Statistics.27

For Aim 3, we calculated indirect SMRs with 95% confidence intervals (95% CI) standardised against Queensland's general population by five-year age group and sex. All analyses were performed using SAS Software (version 9.2).

Results

The study cohort consisted of 42,015 individuals who had been released from adult prisons in Queensland from 1994 to 2007. During the first year following index release, there were 38,769 person years (py) of observation excluding subsequent time in prison.

Young ex-prisoners represented 36% of the study cohort (n=14,920) and contributed 13,207 py of follow-up. A smaller proportion of young ex-prisoners (n=92 deaths, 0.6%) died within 365 days of index release, compared with older ex-prisoners (n=271 deaths, 1.0%). (Table 1).

Table 1. Characteristics of young (<25 years) and older (≥25 years) ex-prisoners at index release (n=42,015), by sex.
 Young at index release (<25 years)Older at index release (≥25 years)
 MaleFemaleMaleFemale
  1. a Person years include time in community and excludes time subsequently incarcerated, during the 365 days after index release

  2. b Incarcerations include prior to index release

Number of ex-prisoners13,349157123,6903405
Deaths within 365 days of index release (%)82 (0.61%)10 (0.64%)247 (1.04%)24 (0.70%)
Person years of observationa11,755145222,2973265
Median age at index release in years20.021.034.034.0
 (IQR)(19.0–22.0)(19.0–23.0)(29.0–42.0)(29.0–40.0)
Indigenous Australian (%)25.8%33.0%14.2%20.4%
Married (%)22.6%24.6%34.7%34.3%
Total number of incarcerations, meanb3.02.42.11.8
 (range)(1–50)(1–30)(1–45)(1–16)
Reincarcerated within 365 days of index release (%)34.527.417.115.7
Median length of index imprisonment in days63338954
 (IQR)(21–168)(11–89)(27–273)(15–149)
History of a drug conviction (%)32.9%33.2%25.7%25.5%
Mean years lost per death56.462.337.946.9
 (95% CI)(55.9–56.9)(60.9–63.6)(36.5–39.3)(43.6–50.3)

Risk factors for mortality

In multifactorial analyses, being young was protective against death in the 365 days following index release (AHR=0.7, 95% CI 0.5–0.8). Independent risk factors for mortality during this period included being unmarried (AHR=1.8, 95% CI 1.4–2.3) and longer sentence length during index incarceration (AHR=1.1 per year, 95% CI 1.0–1.1); history of a drug conviction was a significant protective factor (AHR=0.8, 95% CI 0.6–1.0). (Table 2).

Table 2. Unadjusted and adjusted hazard ratios for mortality within 365 days of index release.
 Univariate hazardMultifactorial hazard
 Hazard Ratio p valueAdjusted Hazard Ratio p value
<25 years at index release
 Yes vs. no0.7 (0.5–0.8)<0.0010.7 (0.5–0.8)<0.001
Male
 Yes vs. no1.3 (0.9–1.9)0.11  
Non-Indigenous Australian
 Yes vs. no1.2 (0.9–1.6)0.18  
Unmarried
 Yes vs. no1.7 (1.3–2.2)<0.0011.8 (1.4–2.3)<0.001
Number of previous imprisonments
 Per additional episode1.1 (1.0–1.2)0.13  
Longer index imprisonment
 Per year1.1 (1.1–1.5)<0.0011.1 (1.0–1.1)<0.001
History of drug conviction
 Yes vs. no0.8 (0.6–1.0)0.030.8 (0.6–1.0)0.04
Community supervision order post-release
 Yes vs. no1.2 (0.9–1.4)0.15  

Risk of death across specific causes

For both age groups, the majority of deaths were caused by injury/poisoning. Deaths among young ex-prisoners were mostly due to injury and poisoning (83% males; 100% females), primarily drug related causes (43% males; 50% females) and suicide (32% males; 30% females). As expected, fewer among the younger group died from natural causes compared with their older adult counterparts (Table 3).

Table 3. Leading underlying and contributing causes of death for young (<25 years) and older (≥25 years) ex-prisoners, within 365 days of index release, by sex.
Cause of deathYoung at index release (<25 years)Older at index release (≥25 years)Relative risk of deatha (95%CI)
 MaleFemaleMaleFemale 
  1. a Risk of death examines all prisoner releases and compares the absolute risk of death being due to a specific cause for all young releases versus all older releases

  2. * No valid observations in at least one age group

All causes82 (100%)10 (100%)247 (100%)24 (100%)0.6 (0.5–0.8)
Injury/poisoning68 (83%)10 (100%)136 (55%)13 (54%)1.0 (0.7–1.2)
Drug-related35 (43%)5 (50%)73 (30%)6 (25%)0.9 (0.6–1.3)
Suicide26 (32%)3 (30%)47 (19%)8 (33%)1.0 (0.6–1.5)
Opioid-related21 (26%)1 (10%)43 (17%)3 (13%)0.9 (0.5–1.4)
All alcohol-related7 (9%)1 (10%)20 (8%)2 (8%)0.7 (0.3–1.5)
Cardiovascular5 (6%)1 (10%)50 (20%)5 (21%)0.2 (0.1–0.5)
Any respiratory disease4 (5%)0 (0%)22 (9%)6 (25%)0.3 (0.1–0.7)
All liver-related diseases3 (4%)1 (10%)18 (7%)3 (13%)0.3 (0.1–1.0)
Viral hepatitis3 (4%)0 (0%)6 (2%)1 (4%)0.8 (0.2–3.0)
Chronic respiratory disease2 (2%)0 (0%)7 (3%)1 (4%)0.5 (0.1–2.1)
Influenza and pneumonia2 (2%)0 (0%)11 (4%)1 (4%)0.3 (0.1–1.4)
Violence3 (4%)1 (10%)10 (4%)1 (4%)0.7 (0.2–2.1)
Alcoholic liver disease0 (0%)0 (0%)5 (2%)1 (4%)*
Cancer-related0 (0%)0 (0%)29 (12%)1 (4%)*
Diabetes-related0 (0%)0 (0%)8 (3%)0 (0%)*
Digestive disorders0 (0%)1 (10%)4 (2%)0 (0%)0.5 (0.1–4.1)
Epilepsy0 (0%)0 (0%)1 (0%)0 (0%)*
HIV-related0 (0%)0 (0%)0 (0%)0 (0%)*
Non-alcoholic liver disease0 (0%)1 (10%)9 (4%)1 (4%)0.2 (0.0–1.4)

Comparison of mortality rates with the general population

The all-cause CMR for the entire cohort in the first year after index release was 9.4 deaths per 1,000 py (95% CI 8.5–10.4). The crude incidence of all-cause mortality was significantly lower among young ex-prisoners (CMR=6.8 per 1,000 py, 95% CI 5.5–8.7) than among older ex-prisoners (CMR=10.6 per 1,000 py, 95% CI 9.4–11.9), and young ex-prisoners had lower risk of all-cause death than their older counterparts (RR=0.6 95% CI 0.5–0.8). However, the all-cause SMR was significantly higher for young ex-prisoners (SMR=6.5, 95% CI 5.3–8.1) than for older ex-prisoners (SMR=4.0, 95% CI 3.5–4.5) (Table 4).

Table 4. Standardised mortality ratios for young (<25 years) and older (≥25 years) ex-prisoners within 365 days of index release, by sex.
Age groupPerson time (years)Observed deathsExpected deathsaCrude mortality rate per 1000 py
(95%CI)
Standardised mortality ratiob
(95%CI)
 MaleFemaleMaleFemaleMaleFemaleMaleFemaleAllMaleFemaleAll
  1. a Expected deaths calculated by standardising study population to Queensland population mortality rates

  2. b Standardised mortality ratios are indirectly calculated from the Queensland general population by five year age group and sex

All ex-prisoners
All ages34,0524,7173293478.94.29.7
(8.6–10.7)
7.2
(4.8–9.6)
9.4
(8.5–10.4)
4.2
(3.7–4.6)
8.1
(5.8–11.3)
4.4
(3.9–4.8)
<25 years11,0631,357741012.40.56.7
(5.3–8.4)
7.4
(4.0–13.7)
6.8
(5.5–8.7)
6.0
(4.8–7.5)
20.3
(10.9–37.8)
6.5
(5.3–8.1)
≥25 years22,9893,3602552466.63.711.1
(9.8–12.5)
7.1
(4.8–10.7)
10.6
(9.4–11.9)
3.8
(3.4–4.3)
6.5
(4.3–9.7)
4.0
(3.5–4.5)

Discussion

Consistent with evidence from community studies,27 we found that being ‘young’ was protective against all-cause death in the first year following release from adult prison. However, young ex-prisoners had markedly elevated mortality rates compared with the general population: in the first year after release from prison, young people in this study had more than six times greater risk of death than sex- and age-matched community peers. Furthermore, in our study the elevation in risk of mortality for young ex-prisoners was considerably greater than that for older ex-prisoners. The risk of death post-release was most elevated for young females, who experienced 20 times greater risk of death compared with age- and sex-matched peers in the general population. Among prisoners, females have particularly poor health,15–16,32,34,42 suggesting that transitional services for ex-prisoners should pay particular attention to this group. However, young women represent only a small proportion of the young ex-prisoner population (in our study, around one in ten) and focusing on this group alone will have limited impact on the incidence of death among young ex-prisoners overall.

Given that the vast majority of deaths among young people in this study were due to drug-related causes or suicide, our findings highlight the importance of transitional programs for young prisoners and ex-prisoners focusing on risky substance use and mental illness. Evidence suggests that, to be most effective, these programs should commence before release from prison, and continue following release.12,35–40 We found that a history of a drug conviction was protective against all-cause death; an unexpected finding that warrants further investigation using methods other than record linkage.41 Given the very high prevalence of mental illness in this population,17 programs designed to improve mental health literacy and adherence to psychiatric medications may also prove beneficial, although again rigorous evaluation studies are urgently needed. Recent efforts to identify and understand the risk factors for mortality in ex-prisoners have been severely constrained by the limitations of correctional records, highlighting the need for more comprehensive electronic records in correctional settings, as well as the application of novel research methods to identify modifiable risk factors for death in this population.42–43

Study limitations

Our analysis is based on a linkage study of more than 40,000 ex-prisoners and has good statistical power due to the number of deaths captured in the first year after release.7 However, data linkage has inherent limitations, including the difficulty in interpreting some variables and associations, and reliance on limited demographic and sociocultural variables. Our study contributes important, but limited, evidence about post-release mortality among young prisoners; further qualitative and quantitative investigations are required to better understand how to reduce mortality in this group. Longitudinal studies of individuals following their release from prison that include a richer set of potential exposures (e.g. socioeconomic status, educational achievement, risk behaviours, health service utilisation and measures of health and wellbeing) and qualitative investigation of the determinants of health and wellbeing would be valuable. Our data were also limited to correctional facilities in one Australian jurisdiction and, although our results may be relevant to other Australian jurisdictions, their relevance in the international context remains to be established. Finally, although mortality among ex-prisoners remains elevated for at least a decade following release from prison,41 we limited our follow-up to one year because our focus was deaths soon after release from custody. This also avoided the potential complexity of individuals defined as ‘young’ at release moving into the ‘older’ age category during follow-up.

Prisoners are a vulnerable group who typically enter prison characterised by social disadvantage, poor health and risky behaviours.44 Time away from the community may exacerbate these problems,45 but the extent to which release from prison is causally associated with elevated mortality remains unclear. Record linkage studies such as the one described here are ill-suited to examining questions of causality, although one recent study in New York found a dose-response relationship between time in prison and mortality,46 suggesting that the possibility of a causal association deserves closer examination.

Conclusions

Compared with the general population, young people experience markedly increased risk of death in the year following release from adult prison. This elevation in risk is greater than that experienced by older ex-prisoners. Among young ex-prisoners, the majority of deaths are due to preventable causes, particularly injury and poisoning, and suicide. Young ex-prisoners would likely benefit from age-specific throughcare programs that inform them about the risks of substance misuse and support them to address substance use, reduce substance-related harm and tackle mental health issues, after return to the community.

Ancillary