European Journal of Epidemiology

, Volume 22, Issue 7, pp 417–428

Factors associated with mortality in a cohort of Australian prisoners

Authors

    • Centre for Health Research in Criminal Justice, Justice Health
    • School of Public Health and Community MedicineUniversity of New South Wales
  • Matthew G. Law
    • National Centre in HIV Epidemiology and Clinical ResearchUniversity of New South Wales
  • Tony G. Butler
    • School of Public Health and Community MedicineUniversity of New South Wales
    • National Drug Research Institute
  • Simon P. Corben
    • New South Wales Department of Corrective Services
  • Michael H. Levy
    • School of Public Health and Community MedicineUniversity of Sydney
    • Corrections Health, ACT Health
  • John M. Kaldor
    • National Centre in HIV Epidemiology and Clinical ResearchUniversity of New South Wales
  • Luke Grant
    • New South Wales Department of Corrective Services
Mortality

DOI: 10.1007/s10654-007-9134-1

Cite this article as:
Kariminia, A., Law, M.G., Butler, T.G. et al. Eur J Epidemiol (2007) 22: 417. doi:10.1007/s10654-007-9134-1

Abstract

We examined factors associated with increased mortality in a cohort of 85,203 adults with a history of imprisonment in New South Wales, Australia, between 1988 and 2002. Information on death was collected through linkage to the Australian National Death Index. The influence of demographic and criminological factors on the standardised mortality ratio (SMR) for all-cause mortality, and deaths due to drug overdose and suicide was examined using negative binomial regression models. The number of deaths identified was 5,137 (4,714 men, 423 women, 303 in custody). The overall SMR was 3.7 (3.6–3.8) in men and 7.8 (7.1–8.5) in women. SMRs raised for deaths due to drug overdose (men: 12.8, women: 50.3) and suicide (men: 4.8, women: 12.2). The high SMR was associated with hospitalisation for mental illness, multiple imprisonments, and early stage of follow-up independently of causes of death. Being released from prison increased the SMRs for all-cause and drug-related mortality, but not suicide. For women, significant trends for decreasing risk with increasing age were noted. Minority groups, in particular men, had a lower risk of death than white people. In men a sex or drug offence was associated with a lower risk and a property or violence offence was related to higher mortality. Our results reinforce how disadvantaged prisoners are, measured by mortality as the most fundamental scale of human wellbeing. Certain demographic and imprisonment characteristics are indicators of high mortality among this population. The underlying causes of some of these characteristics such as mental illness or multiple imprisonments are potentially treatable and preventable. Prison health services need to develop interventions targeting high-risk groups to avoid this situation.

Introduction

Studies conducted in different countries have documented that prisoners and ex-prisoners are at increased risk of death compared with the general population [17]. This excess mortality has been attributed mainly to an increased risk of unnatural death, particularly drug overdose and suicide.

Most studies conducted on prisoner populations provide an epidemiological description of mortality rates; very few have analysed possible factors associated with the risks, apart from suicide. However, the findings of the research in this area for suicide appear to be inconsistent [8]. Some studies suggest that suicide victims are young [9, 10], while others have found older age to be correlated with prison suicide [11], or reported that there is no relationship [12]. American and Australian studies have shown an over-representation of white [10, 12], non-Indigenous inmates [13], but a review by LIoyd [14] did not identify ethnicity as a prominent factor in prison suicide. Most studies have focused on men and those, which had included women, had limited numbers for analysis of gender difference. Nevertheless, prison suicide is considered a male-specific phenomenon [8, 15]. Studies suggest that violent or sexual offenders [11, 16], repeat offenders [17, 18], and those serving long sentences [19, 20] are over-represented in the suicide statistics. However, a few studies contradict this pattern, finding that non-violent [21], first time offenders [20], and offenders serving shorter sentences [22] are at increased risk of suicide. Several authors found that prison suicide is associated with psychological disorders [11, 18, 23], alcohol and drug misuse [10, 18], a history of self-harm [10, 11], isolation [11, 24], and prison overcrowding [25].

More recently, a limited number of researchers have examined factors associated with the risk of death in ex-prisoners. In one study on male prisoners in Finland [1], natural death was associated with alcohol dependence, a history of treatment in hospital for somatic diseases, and poor basic education. Suicide in the community was found to be associated with psychosis. Two research on drug-related mortality among newly released prisoners in England and Wales [26, 27] identified a history of drug abuse, a sex offence in men and a property offence in women, a short sentence, unemployment, being white, and being single as significant predictors of death following release.

In a previous study [28] we demonstrated that people with a history of full-time imprisonment in New South Wales (NSW), Australia had substantially higher risk of death than the general population. We extend this work to identify risk factors associated with increased mortality in our cohort, with a special focus on mortality related to drug overdose and suicide.

Methods

Study cohort

A retrospective cohort study was conducted, including 85,203 adult males and females who had been received into full-time custody in NSW between 1 January 1988 and 31 December 2002. The cohort was identified by the NSW Department of Corrective Services (DCS) from their computerised Offender Integrated Management System. For each individual we received information on demographic factors (names, date of birth, sex, country of birth), and imprisonment history (the start and end date of each incarceration, offence type), and hospitalisation for mental illness while in custody.

Mortality data source

Deaths were identified by record linkage to the Australian National Death Index (NDI), a register held by the Australian Institute of Health and Welfare (AIHW). This register provides complete population mortality data and is based on the deceased person’s death certificate received from Registrar of Births Deaths and Marriages in each Australian State and Territory. At the time we conducted the linkage, the NDI database contained information on all deaths reported up to 31 December 2002. Linkage was done by the AIHW using last name, family name, middle name, sex, date of birth (or estimated year of birth), and date of last contact with the prison system, and has been described elsewhere [29]. Only highest-level matches (that is, matches on gender, family name, first name, middle name and date of birth accurate within 1 year) were accepted; ICD-9 (deaths before 1997) and ICD-10 (deaths since 1997) classification code for underlying cause of death were extracted from NDI for all the accepted matches. We have previously found that our linkage with the NDI has a sensitivity of 88.4% and specificity of 99.7% for identifying deaths. For different causes of death, sensitivity was between 92.3% and 100% and specificity was between 98.1% and 100% [29].

Definition of predictor variables

We examined the influence of the following factors on mortality: age, country of birth, admission to the prison psychiatric hospital, most serious offence, location of death, number of previous imprisonments, cumulative time spent in adult prisons and length of follow-up.

For those individuals with multiple incarcerations, the most serious offence across all episodes of imprisonment was defined as the offence for which the longest sentence was imposed. For those with sentences of equal length, and for those where no sentence had been imposed (remand only prisoners), we used the most serious offence, based on the Australian National Classification of Offences [30]. For this study, offences were categorised as: property, drug, violent, sexual, and other (mostly relating to judicial procedures). The variable ‘location of death’ was coded as a dichotomous variable indicating whether the person died in prison or in the community following release.

Statistical analysis

For each member of the cohort, person-years at risk by age group, sex, and calendar year were calculated, starting from the date of first imprisonment from 1 January 1988, until the date of death or the end date of the study. Age and calendar year were treated as time dependent variables. Expected mortality in the cohort was calculated by multiplying the age, sex, and calendar-year specific person years at risk in the cohort by the corresponding mortality rate for the general population of NSW. We then calculated standardised mortality ratios (SMRs), reflecting the risk of all-cause mortality, drug-related mortality and suicide in the cohort in comparison with the corresponding risks in the general population, as the ratio of the number of observed deaths to the number of expected deaths.

Associations between explanatory variables and SMR were investigated using negative binomial regression models. Negative binomial regression is an extension of Poisson regression that allows for overdispersion (where variance is larger than mean) in the data, and hence ensures that standard errors and p values are correct [31]. When there is overdispersion, estimates from the Poisson regression are inefficient with standard errors that are biased downward [31]. The significance of overdispersion was tested using the likelihood-ratio test of α (parameter that determines the degree of dispersion in the predicted count). Analyses were repeated using Poisson regression models, and gave essentially similar results (data not shown).

We considered demographic and criminological characteristics that were statistically significant (P < 0.05) in univariate analysis to be potential risk factors. Multivariate models were then developed using a stepwise backwards selection procedure. The significance of each variable was assessed with the likelihood ratio test. The number of imprisonments, total time in prison and length of follow-up were treated as time-dependent covariates. Separate analyses were carried out for men and women. We used STATA release 8.0 for all statistical analyses [32].

Ethical approval

Approval for the study was obtained from the Human Research Ethics Committees of the NSW Department of Corrective Services, NSW Justice Health, the Australian Institute of Health and Welfare, and the University of NSW.

Results

The cohort consisted of 76,383 men and 8,820 women of whom, over two thirds were born in Australia (Table 1). The median age at entry into prison was 27.2 years (range = 18–86) for men and 27.3 years (range = 18–83) for women. Around 40% of the cohort aged less than 25 years. Approximately 47% of men and 45% of women had experienced at least two episodes of imprisonments. Multiple imprisonments were substantially more common in Aboriginal Australians, with 50.6% of them experiencing three or more terms compared to 30.7% of non-Aboriginal prisoners born in Australia and 17.6% of those born in other countries (P < 0.0001). Twenty-four percent of the cohort’s most serious offence was categorised as violent. The median follow-up time was 7.7 years (range = 1 day–15 years).
Table 1

Demographic and imprisonment history of the study cohort (n = 85,203)

Characteristic

Men (n = 76,383)

Women (n = 8,820)

Age at study entry in years, median (range)

27.2 (18–86)

27.3 (18–83)

Country of birth/Ethnicity

   Australian

  

     Aboriginal

7,980 (10.4)

1,373 (15.6%)

     Non-Aboriginal

43,544 (57%)

4,850 (55%)

   Other English speaking countries

14,866 (19.5%)

1,871 (21.2%)

   Asian

3,287 (4.3%)

315 (3.6%)

   Other non-English speaking countries

6,703 (8.8%)

411 (4.7%)

   Unknown

3 (0%)

0 (0%)

   Psychiatric hospital admission in prison

3,919 (5.1%)

108 (1.2%)

Most serious offence*

   Property

22,472 (29.4%)

3,542 (40.2%)

   Drug related

7,668 (10.0%)

969 (11.0%)

   Violent

18,514 (24.2%)

1,604 (18.2%)

   Sexual

3,854 (5.0%)

23 (0.3%)

   Other

23,875 (31.3%)

2,682 (30.4%)

Number of incarcerations, median (range)

2.5 (1–78)

2.5 (1–38)

Follow-up time in years, median (range)

7.7 (1 day–15 years)

7.7 (1 day–15 years)

* Refers to all incarcerations for repeat offenders and last incarceration for one off offenders

A total of 5,137 deaths (4,714 men, 423 women) were identified during the study period. Three hundred and three deaths occurred in custody, of which 130 (43%) happened in the first term of imprisonment. The median age at death was 36.6 years for men and 32.7 years for women. The overall SMR for men was 3.7 (95% CI: 3.6–3.8) and for women was 7.8 (95% CI: 7.1–8.5). The SMRs were elevated for drug-related deaths (men: 12.8, 95% CI: 12.2–13.5; women: 50.3, 95% CI: 43.7–57.8) and suicide (men: 4.8, 95% CI: 4.4–5.1; women: 12.2, 95% CI: 9.2–16.2).

Risk factors

All-cause mortality

The SMR for all-cause mortality showed a significant decreasing trend with age in women (Table 2). The relative risk dropped from 2.08 in the 18–19 years age group to 0.26 in those aged 45 years and over. In men a significant decline in the excess mortality was observed only among those 45 years of age and older. Mortality varied by race and ethnicity. Asians and those born in other non-English speaking countries had a lower risk of death than non-Aboriginal prisoners born in Australia (reference group). In women, a significant decrease, however, was observed only in the latter group (RR = 0.36; 95% CI: 0.19–0.69). The risk of death was higher for Aboriginal prisoners than non-Aboriginal prisoners born in Australia, but the risk became similar after adjusting for other factors.
Table 2

Association of demographic and imprisonment history with all-cause mortality among the study cohort, New South Wales, Australia, 1988–2002

 

All-cause mortality (Men)

All-cause mortality (Women)

 

Observed deaths

Expected deaths

Univariate RR (95% CI)

Multivariate* RR (95% CI)

Observed deaths

Expected deaths

Univariate RR (95% CI)

Multivariate* RR (95% CI)

Age (years)

    18–19

63

10.58

1.17 (0.91–1.50)

1.12 (0.85–1.46)

9

0.32

2.11 (1.08–4.13)

2.08 (1.03–4.19)

    20–24

497

101.41

0.96 (0.87–1.06)

0.93 (0.83–1.05)

77

3.12

1.83 (1.39–2.40)

1.80 (1.35–2.40)

    25–34

1,563

306.30

1.00

1.00

156

11.54

1.00

1.00

    35–44

1,261

257.43

0.96 (0.89–1.03)

1.03 (0.94–1.12)

116

14.14

0.61 (0.48–0.77)

0.64 (0.50–0.82)

    ≥45

1,330

592.96

0.44 (0.41–0.47)

0.55 (0.50–0.60)

65

25.42

0.19 (0.14–0.25)

0.26 (0.19–0.35)

Country of birth

    Non-Aboriginal Australian

2,983

733.20

1.00

1.00

254

30.60

1.00

1.00

    Aboriginal Australian

485

100.57

1.19 (1.08–1.30)

0.90 (0.81–01)

73

5.80

1.52 (1.17–1.97)

1.03 (0.79–1.34)

    Other English speaking

810

232.49

0.86 (0.79–0.93)

0.98 (0.90–1.08)

82

10.89

0.91 (0.71–116)

1.08 (0.84–1.38)

    Asian

105

38.33

0.67 (0.55–0.82)

0.76 (0.62–0.93)

4

1.32

0.36 (0.14–0.98)

0.41 (0.15–1.11)

    Other non-English speaking

331

164.03

0.50 (0.44–0.56)

0.62 (0.55–0.70)

10

5.92

0.20 (0.11–0.38)

0.36 (0.19–0.69)

Location of death

    Prison

295

175.80

1.00

1.00

8

3.83

1.00

1.00

    Community

4,419

1,092.87

2.41 (2.14–2.71)

2.95 (2.51–3.47)

415

50.70

3.92 (1.95–7.89)

4.72 (2.22–10.01)

Psychiatric hospital admission in prison

    Yes

467

85.25

1.53 (1.39–1.68)

1.46 (1.32–1.62)

17

1.06

2.11 (1.30–3.43)

1.81 (1.10–2.95)

    No

4,247

1,183.43

1.00

1.00

406

53.47

1.00

1.00

Most serious offence

    Violent

1,130

266.70

1.00

1.00

79

9.01

1.00

 

    Sexual

246

140.27

0.41 (0.36–0.48)

0.65 (0.56–0.76)

1

0.19

0.61 (0.09–4.41)

 

    Drug

357

153.50

0.55 (0.49–0.62)

0.71 (0.62–0.81)

35

7.51

0.53 (0.36–0.79)

 

    Property

1,547

331.56

1.10 (1.02–1.19)

1.01 (0.92–1.11)

173

21.98

0.90 (0.69–1.17)

 

    Other

1,434

376.64

0.90 (0.83–0.97)

0.96 (0.87–1.06)

135

15.85

0.97 (0.74–1.28)

 

No. of previous imprisonments

    No previous imprisonment

129

90.72

0.46 (0.39–0.55)

1.12 (0.88–1.42)

1

1.88

0.09 (0.01–0.65)

0.29 (0.04–2.37)

    Once

2,363

770.23

1.00

1.00

211

36.21

1.00

1.00

    2–3 times

1,375

288.99

1.55 (1.45–1.66)

1.55 (1.43–1.69)

115

11.09

1.78 (1.42–2.23)

1.59 (1.26–2.02)

    4 or more

847

118.74

2.33 (2.15–2.52)

2.27 (2.05–2.52)

96

5.35

3.08 (2.42–3.92)

2.69 (2.06–3.52)

Length of imprisonmenta

    No previous imprisonment

129

90.72

0.40 (0.33–0.48)

 

1

1.88

0.07 (0.01–0.50)

 

    <2 weeks

949

264.79

1.00

 

133

17.50

1.00

 

    2–26 weeks

1,657

420.90

1.10 (1.01–1.19)

 

170

19.61

1.14 (0.91–1.43)

 

    27–52 weeks

653

174.60

1.04 (0.94–1.15)

 

59

6.57

1.18 (0.87–1.60)

 

    >1 year

1,326

317.67

1.16 (1.07–1.27)

 

60

8.97

0.88 (0.65–1.19)

 

Length of follow-up

    <1 year

588

134.14

1.21 (1.11–1.32)

1.70 (1.52–1.89)

64

5.56

1.72 (1.31–2.26)

1.71 (1.24–2.35)

    1–3 years

938

252.35

1.03 (0.96–1.11)

1.18 (1.08–1.29)

100

10.32

1.45 (1.15–1.82)

1.25 (0.97–1.60)

    >3 years

3,188

882.18

1.00

1.00

259

38.65

1.00

1.00

* Includes variables associated with deaths at the P < 0.05 level in the univariate analysis

a Cumulative length of time spent in prison

All-cause mortality was higher following release from prison, particularly among women (Table 2). Men and women who had been admitted to the prison psychiatric hospital for a mental health problem had higher mortality. Among men, sex offenders and those with a drug related crime had the lowest all-cause mortality. The relative risk for SMR was 0.65 (0.56–0.76) for sex offenders and was 0.71 (0.62–0.81) for drug offenders compared with those with a violent crime (reference group). Among women, excess mortality was not significantly related to the type of offence after controlling for other risk factors. There was clearly an increasing trend in mortality with increasing number of imprisonments. Being imprisoned four or more times more than doubled the risk of all-cause mortality in men (RR = 2.27, 95% CI: 2.05–2.52) and in women (RR = 2.69, 95% CI: 2.06–3.52). The first year of follow-up was associated with the highest excess mortality in both men and women.

Drug-related mortality

In men, drug-related SMRs were elevated in those under 20 years of age and those aged 35 years and older, while in women, it showed a decreasing trend with age: the relative risk dropped from 3.83 in the youngest age group to 0.37 in those aged 45 years and older (Table 3). Drug-related mortality showed substantial differences by ethnicity and Aboriginal status. Men of Aboriginal heritage, Asians and those from non-English speaking countries had lower risk of death than non-Aboriginal prisoners born in Australia. For women, drug-related mortality was lower in ethnic minorities and Aborigines. However, we could not find a significant difference in the first group, probably due to the small number of women.
Table 3

Association of demographic and imprisonment history with drug-related mortality* among the study cohort, New South Wales, Australia, 1988–2002

 

Drug-related mortality (Men)

Drug-related mortality (Women)

 

Observed death

Expected death

Univariate RR (95% CI)

Multivariatea RR (95% CI)

Observed death

Expected death

Univariate RR (95% CI)

Multivariatea RR (95% CI)

Age

    18–19

20

0.84

2.05 (1.31–3.20)

1.68 (1.06–2.67)

8

0.03

4.76 (2.31–9.83)

3.83 (1.78–8.26)

    20–24

219

16.43

1.15 (0.99–1.34)

1.05 (0.89–1.24)

46

0.50

1.90 (1.33–2.72)

1.64 (1.13–2.39)

    25–34

690

59.47

1.00

1.00

88

1.83

1.00

1.00

    35–44

458

32.26

1.22 (1.09–1.38)

1.34 (1.18–1.52)

50

1.14

0.91 (0.64–1.29)

1.02 (0.72–1.46)

    ≥45

90

6.32

1.23 (0.99–1.53)

1.54 (1.22–1.93)

5

0.41

0.26 (0.10–0.63)

0.37 (0.15–0.91)

Country of birth

    Non-Aboriginal Australian

1,031

69.76

1.00

1.00

129

2.28

1.00

1.00

    Aboriginal Australian

111

13.68

0.55 (0.45–0.67)

0.48 (0.39–0.59)

27

0.62

0.77 (0.51–1.17)

0.62 (0.41–0.94)

    Other English speaking

228

18.02

0.86 (0.74–0.99)

1.01 (0.87–1.18)

35

0.72

0.86 (0.59–125)

1.00 (0.69–1.46)

    Asian

31

4.14

0.51 (0.35–0.72)

0.59 (0.41–0.85)

3

0.10

0.52 (0.17–1.63)

0.59 (0.19–1.87)

    Other non-English speaking

76

9.73

0.53 (0.42–0.67)

0.56 (0.44–0.71)

3

0.20

0.26 (0.08–0.83)

0.35 (0.11–1.09)

Location of death

    Prison

46

16.92

1.00

1.00

1

0.34

1.00

 

    Community

1,431

98.41

5.35 (3.99–7.17)

5.36 (3.87–7.41)

196

3.58

18.40 (2.58–131.25)b

 

Psychiatric hospital admission in prison

    Yes

145

7.54

1.56 (1.31–1.85)

1.36 (1.13–1.63)

7

0.08

1.71 (0.80–3.63)

 

    No

1,332

107.79

1.00

1.00

190

3.83

1.00

 

Most serious offence

    Violent

313

27.21

1.00

1.00

32

0.67

1.00

 

    Sexual

26

4.96

0.46 (0.31–0.68)

0.49 (0.32–0.73)

1

0.01

1.78 (0.24–13.06)

 

    Drug

121

12.22

0.86 (0.70–1.06)

0.96 (0.77–1.19)

16

0.47

0.70 (0.39–1.28)

 

    Property

635

36.65

1.51 (1.32–1.72)

1.35 (1.17–1.56)

84

1.65

1.06 (0.71–1.59)

 

    Other

382

34.29

0.97 (0.83–1.12)

1.02 (0.87–1.20)

64

1.12

1.19 (0.78–1.83)

 

No. of previous imprisonment

    No previous imprisonment

7

6.44

0.11 (0.05–0.24)

0.43 (0.19–0.97)

0

0.13

NA

NA

    Once

596

61.86

1.00

1.00

82

2.28

1.00

1.00

    2–3 times

501

30.89

1.68 (1.50–1.90)

1.95 (1.70–2.25)

60

0.96

1.75 (1.25–2.44)

2.04 (1.44–2.88)

    4 or more

373

16.13

2.40 (2.11–2.73)

3.11 (2.61–3.71)

55

0.55

2.80 (1.99–3.94)

4.03 (2.75–5.91)

Length of imprisonmentc

    No previous imprisonmentd

7

6.44

0.11 (0.05–0.24)

 

0

0.13

NA

 

    <2 weeks

249

25.87

1.00

1.00

50

1.27

1.00

 

    2–26 weeks

500

39.64

1.31 (1.13–1.53)

1.20 (1.01–1.41)

87

1.50

1.47 (1.04–2.08)

 

    27–52 weeks

231

15.70

1.53 (1.28–1.83)

1.34 (1.10–1.65)

24

0.45

1.34 (0.82–2.18)

 

    >1 year

490

27.67

1.84 (1.58–2.14)

1.61 (1.31–1.98)

36

0.57

1.61 (1.05 –2.46)

 

Length of follow-up

    <1 year

215

13.03

1.33 (1.15–1.54)

3.28 (2.73–3.95)

40

0.52

1.89 (1.31–2.73)

2.72 (1.74–4.24)

    1–3 years

305

25.05

0.98 (0.86–1.12)

1.57 (1.35–1.82)

56

0.91

1.51 (1.09–2.10)

1.67 (1.16–2.39)

    >3 years

957

77.24

1.00

1.00

101

2.49

1.00

1.00

* ICD-9 304, 305.2–305.9, E950.0–E950.5, E962.0, E850–E858, E980.0–E980.5, 305.0; ICD–10 F11–F16, F19, F55, X60–X64, X85, X40–X44, Y10–Y14

a Includes variables associated with deaths at the P < 0.05 level in the univariate analysis

b Only one death in prison, therefore, this variable was not considered in the multivariate analysis

c Cumulative length of time spent in prison

d This category was dropped from the multivariate analysis in men due to collinearity

Drug-related deaths occurred more excessively following release from prison. We did not include the variable ‘location of death’ as a predictor for drug-related mortality in the multivariate model for women, as out of the 197 drug-related deaths, only one occurred while in prison (Table 3). Hospitalisation for a mental illness during a period of incarceration was associated with drug-related mortality in men. Drug-related mortality was more common in male property offenders (RR = 1.35; 95% CI: 1.17–1.56) and less so in sex offenders (RR = 0.49; 95% CI: 0.32–0.73) than those with a violent crime. Among women, excess mortality due to drug overdose was not related to the type of offence. Repeat offenders had higher risk. In both men and women, the excess risk reduced significantly after the third year of follow-up.

Suicide

Suicide risk was greatest among the following groups (Table 4): those admitted to the prison psychiatric hospital, violent offenders and repeat offenders. Aboriginal men and men from other non-English speaking countries had lower risk of dying from suicide. Suicides were almost three times more likely to have occurred within the first year of follow-up than after 3 years. Among women, the length of follow-up was the only variable which showed a significant association with suicide in the univariate analysis.
Table 4

Association of demographic and imprisonment history with suicide among the study cohort, New South Wales, Australia, 1988–2002

 

Suicide (Men)

Suicide (Women)

 

Death, observed

Death, expected

Univariate RR (95% CI)

Multivariate* RR (95% CI)

Death, observed

Death, expected

Univariate RR (95% CI)

Multivariate* RR (95% CI)

Age

    18–19

25

2.22

2.39 (1.59–3.59)

 

0

0.04

NA

 

    20–24

122

27.09

0.95 (0.78–1.17)

 

10

0.46

1.79 (0.84–3.79)

 

    25–44

351

74.40

1.00

 

21

1.74

1.00

 

    45–64

201

43.17

0.99 (0.83–1.17)

 

12

1.29

0.77 (0.38–1.57)

 

    ≥65

98

20.74

1.00 (0.80–1.25)

 

6

0.49

1.02 (0.41–2.53)

 

Country of birth

    Non-Aboriginal Australian

512

100.65

1.00

1.00

34

2.32

1.00

 

    Aboriginal Australian

57

19.35

0.58 (0.44–0.76)

0.52 (0.39–0.70)

4

0.62

0.44 (0.16–1.25)

 

    Other English speaking

160

26.58

1.18 (0.99–1.41)

1.26 (1.03–1.55)

8

0.75

0.73 (0.34–1.58)

 

    Asian

20

5.98

0.66 (0.42–1.03)

0.72 (0.46–1.14)

2

0.11

1.29 (0.31–5.36)

 

    Other non-English speaking

48

15.05

0.63 (0.47–0.84)

0.67 (0.49–0.91)

1

0.22

0.31 (0.04–2.27)

 

Location of death

    Yes

117

25.40

1.00

 

3

0.34

1.00

 

    No

680

142.21

1.04 (0.85–1.26)

 

46

3.67

1.40 (0.44–4.51)

 

Psychiatric hospital admission in prison

    Yes

116

10.96

2.43 (2.00–2.96)

2.30 (1.85–2.87)

2

0.08

2.00 (0.49–8.25)

 

    No

681

156.65

1.00

1.00

47

3.93

1.00

 

Most serious offence

    Violent

231

39.17

1.00

1.00

12

0.68

1.00

 

    Sexual

33

8.32

0.67 (0.47–0.97)

0.72 (0.49–1.06)

1

0.01

4.52 (0.59–34.78)

 

    Drug

54

17.98

0.51 (0.38–0.68)

0.58 (0.42–0.80)

3

0.50

0.34 (0.10–1.22)

 

    Property

249

52.56

0.80 (0.67–0.96)

0.80 (0.65–0.99)

17

1.67

0.58 (0.28–1.21)

 

    Other

230

49.59

0.79 (0.66–0.94)

0.81 (0.65–1.01)

16

1.15

0.79 (0.37–1.67)

 

No. of previous imprisonment

    No previous imprisonment

55

10.31

1.21 (0.92–1.61)

0.78 (0.57–1.07)

0

0.13

NA

 

    Once

402

91.49

1.00

 

26

2.36

1.00

 

    2–3 times

228

43.83

1.18 (1.01–1.39)

1.48 (1.21–1.81)

15

0.97

1.40 (0.74–2.65)

 

    4 or more

112

21.98

1.16 (0.94–1.43)

1.47 (1.13–1.93)

8

0.55

1.31 (0. 59–2.89)

 

Length of imprisonmenta

    No previous imprisonment

55

10.31

1.07 (0.79–1.45)

 

0

0.13

NA

 

    <2 weeks

185

37.26

1.00

 

18

1.30

1.00

 

    2–26 weeks

267

57.64

0.93 (0.77–1.13)

 

18

1.53

0.85 (0.44–1.63)

 

    27–52 weeks

94

22.84

0.83 (0.65–1.06)

 

10

0.47

1.54 (0.71–3.35)

 

    >1 year

196

39.56

1.00 (0.82–1.22)

 

3

0.59

0.36 (0.11–1.24)

 

Length of follow-up

    < 1 year

184

20.52

2.20 (1.85–2.61)

2.89 (2.29–3.64)

9

0.51

1.97 (0.91–4.26)

 

    1–3 years

167

37.61

1.09 (0.91–1.30)

1.25 (1.01–1.55)

17

0.91

2.11 (1.13–3.96)

 

    >3 years

446

109.48

1.00

1.00

23

2.59

1.00

 

* Includes variables associated with deaths at the P < 0.05 level in the univariate analysis

a Cumulative length of time spent in prison

Discussion

Principal findings

This study found that among adult men with a history of incarceration, increased mortality was predicted by hospitalisation for mental illness during imprisonment, being in the community rather than prison, increased number of imprisonment episodes, and being in the early stage of the follow-up. Reduced risk of mortality among men was predicted by Asian born or other non-English speaking background, and a most serious offence related to sex or drugs. Among women, increased mortality was observed in those aged under 25 years, those hospitalised for mental illness, being in the community rather than prison, repeat offenders and those women in the early phase of their follow-up. Mortality was generally lower in Asian women and women from other non-English speaking countries. In both men and women, Aboriginal status was associated with lower drug-related mortality and suicide.

In this study, younger age was an important determinant of mortality in women, indicating the extreme vulnerability of this group. This observation suggests that efforts to reduce death among women prisoners and ex-prisoners should focus on younger offenders, especially those under the age of 20 years.

The increased risk of mortality in those who were admitted to the prison psychiatric hospital during incarceration confirms previous research [1, 11] and reinforces the increased vulnerability of this group. We acknowledge that admission to the prison psychiatric hospital is not a precise measure of psychiatric ill health and is likely to occur in only the most serious cases. Nevertheless, the increased mortality observed in this group provides support to the view that they are one of the most at-risk groups within the prisoner population. Further work needs to be done to examine the specific diagnoses in this group and length of time to death following their release.

By far the strongest determinant of mortality in the cohort was being released from prison into the community. The post-release population are known to be at a greater risk of death from overdose due to reduced tolerance to certain substances following a period of abstinence or reduced intake while in prison [3, 33]. Compounding this are the myriad social factors following imprisonment including financial stresses, profound isolation, shame and stigmatisation, homelessness, and problems re-integrating into family structures. Indices of social isolation and dysfunction have been reported as predictors of overall mortality [34]. Across two decades of research, no study has specifically compared mortality in prison with that following release in the same population. However, studies on post release population consistently reported higher mortality [3, 4, 7] than those conducted on prisoners [2, 5, 35].

Our study is among the few to document the association between ethnicity and mortality. The reduced risk of mortality from drug overdose and suicide in racial/ethnic minority prisoners is in agreement with findings from other studies on prisoners [10, 27, 36] and other marginalised groups [37]. The reduced mortality may be due to better social adjustment and better community support in ethnic minorities relative to non-minority groups. An American researcher suggests that the differences among black, white, and Hispanic suicide rates can be explained by socio-cultural factors such as better preparations for prison life by blacks as opposed to that of whites [38]. Haycock disagrees this theory, and disputes that the factors that lead to inmate suicide are complex and personal and do not simply depend on socio-cultural backgrounds [39].

In Australia, the Royal Commission into Aboriginal Deaths in Custody (RCIADIC) initially found that between 1980 and 1989, Aboriginal Australians died in prison at a rate 13 times that of non-Aboriginal people, but these figures were subsequently shown to be almost entirely due to the over-representation of Aboriginal people in prison [40]. Thomson and McDonald [13], analysed the data from the same period investigated by the RCIADIC, and showed that Aboriginal inmates were less likely to commit suicide in prison, but much more likely to die from diseases of circulatory system. A recent study on released prisoners in Western Australia [41], reported greater risk of death in Aboriginal than non-Aboriginals. However, after adjusting for number of releases the risk became similar due to a high correlation between this variable and Indigenous status. We did not examine factors related to excess mortality from disease-related causes, but for Aboriginal men in our cohort, the SMRs for deaths from cardiovascular and digestive system diseases, for instance, were in order of four and two and for Aboriginal women in order of four and five times that of the respective general prisoner population (data not shown). Therefore, our finding suggests that it may be wrong to focus only on the health of Aboriginal or non-Aboriginal prisoners, rather appropriate allocation of resources based on need should be our concern.

In men, property offenders had the highest excess mortality from drugs overdose. This finding is consistent with the recognised association between property crime and substance abuse [42]. Dependent drug users are forced to resort to crime involving property to fund their purchase of illicit drugs. However, no information was available on drug dependence in our study to provide direct support for this inference. Prisoners committed or charged with a drug offence had a lower risk of drug-related mortality suggesting that while their most serious offence was connected to drugs, they were perhaps less likely to be the end-users and more likely involved in the supply of drugs. We found that the offence profile was not a significant predictor of mortality in women. This could be due to the smaller sample size in the female group or that women prisoners tend to be a more homogenous group than men, the majority of whom are incarcerated for drug related crimes.

Most deaths in the cohort occurred in the early stage of follow-up, especially within the first year, indicating that offenders are at greater risk of dying at the beginning of their criminal careers. The first year of follow-up stands out as being particularly dangerous for drug-overdose and suicide. The authors could only speculate whether this was due to coping difficulties and also the fact that people addicted to drugs are at greater risk of death in the early phases of their addiction [43].

Limitations

Our analysis is based on a linkage study of more than 85,000 male and female prisoners and ex-prisoners. Because our sample is large, we have been able to document the association of several different variables to the overall and cause-specific mortality. However, complete information on lifestyle features was unavailable. As a result, it is possible that unadjusted confounders such as marital and employment status contributed to some of the observed associations.

The ascertainment of vital status and causes of deaths were based on linkage to the national register. We may have underestimated the overall SMRs in our study cohort and misclassification of causes of death may have also occurred. However, based on applying the same linkage procedure to prisoners with known vital status, we have previously demonstrated that the sensitivity of the linkage procedure was 88.4% and the specificity was 99.7% [29]. Therefore, if we assume these rates of misclassification of death apply in this study, these figures for sensitivity and specificity would lead to underestimation of the overall SMRs by around 7% [44]. Additional deaths might change the predictive significance of the variables ‘country of birth’ and ‘offence profile’ for mortality in women. Similarly, assuming that any misclassification of deaths in the linkage process is random and the sensitivity and specificity of our linkage is independent of risk factors, the only bias will be to underestimate relative risks by less than 5%. The sensitivity of our linkage for suicide was 93.8% and specificity was 98.1%. Applying these figures to our study cohort, would suggest an overestimation of suicide by 4% in men and 10% in women, which is unlikely to impact on our conclusion concerning suicide (Table 4).

Summary

A number of demographic and criminologic factors predict the excess mortality among prisoners and ex-prisoners. Because some of the factors with a strong association with mortality found in this study, particularly multiple imprisonment or severe mental illness, seem to have the greatest potential for modification, it is reasonable to assume that there is considerable scope for reducing mortality in this population. Programmes aimed at the treatment of drug addiction and mental illness both within and beyond period of imprisonment should be considered. Future studies should examine how access to psychiatric and drug treatment, diversion from prisons, and social support after release for this vulnerable group reduce their risk of death.

Policy implications

Programmes and intervention to target specific high-risk prisoners and address addictions, mental and physical health, and underlying causes of recidivism are potentially life-saving interventions.

Acknowledgments

This study is supported by the Research Grant No. 222849 from the National Health and Medical Research Council of Australia. We are grateful to Mr Simon Eyland, the Director of research at the NSW Department of Corrective Services for making the necessary arrangement for extraction of the inmate data.

Copyright information

© Springer Science+Business Media B.V. 2007