Data sources
This study used a retrospective cohort design that involved data linkage of three iterations of the New South Wales (NSW) Inmate Health Surveys conducted in 1996 [19], 2001 [20] and 2009 [21] to the NSW Bureau of Crime Statistics and Research’s Re-offending Database [22]. The surveys are described in detail elsewhere [19-21]. Briefly, the surveys recruited random samples of men and women prisoners stratified by age, sex, and Indigenous status from all NSW correctional centres. Participants were included if they were over the age of 18 years, spoke sufficient English to participate in the interview, and were able to provide informed consent. Face-to-face interviews were conducted in 1996 and 2001, and a combination of face-to-face and telephone interviews conducted in 2009. The surveys were wide ranging: covering mental and physical health as well as health risk behaviours; and included serological and urine screening for chronic and infectious diseases. The most serious offence for the custodial episode when the survey was conducted was recorded.
The Re-offending Database holds records of all finalised court appearances and full-time prison episodes for offenders in NSW from 1994 to present [22]. The offence histories were coded according to the Australian and New Zealand Standard Offence Classification [23].
Approval for use of the survey data was provided by the Justice Health and Forensic Mental Health Network (JH&FMHN) and for offending data by the NSW Bureau of Crime Statistics and Research. Ethics approval for the linkage was provided by the JH&FMHN Human Research Ethics Committee (G70/14).
Data linkage procedure
Data from the 2,327 men that participated in the surveys were combined into one dataset[i]. There were 213 men that participated in more than one of the surveys. When this occurred, the data from the earliest survey was used resulting in a cohort of 2,114 men. Deterministic linkage using a unique prisoner identifier or common identifiers (name, sex, and date of birth) matched the survey data to offending data extracted from the Re-offending Database. Offending data for 1,853 (87.7%) of the cohort were available from January 1994 to October 2014[ii].
Classification procedure
The cohort was classified into groups based on the type of crimes identified by the survey and offending data from the date of the first available offence until the date of participation in the survey. Participants were classified as sex offenders if any Sexual assault and related offence [23] was recorded in their histories before or at the time of the survey. Only contact sexual offences were included (aggravated sexual assault and non-aggravated sexual assault). Sex offenders were classified into subgroups based on victim age using the age of consent in NSW resulting in: child sex offenders (ChildSOs) that had sexual offences against only victims under the age of 16 years old; adult sex offenders (AdultSOs) that had only victims 16 years and above; and polymorphous sex offenders (PolySOs) that had both children and adult victims. Violent (non-sex) offenders included those with records of Homicide and related offences, Acts intended to cause injury, Abduction, harassment, and other offences against the person, and Robbery, extortion and related offences [23] before or at the time of the survey. Other (non-sex, non-violent) offenders were those with crimes that did not fall into either sexual or violent crime categories.
Measures
Demographic characteristics and summary health variables
Using the survey data, we examined basic demographic and descriptive characteristics of the sample including: age; Indigenous status (0=no, 1=yes ATSI); less than high school education (0=no, 1=yes); usual occupation as self-employed (0=no, 1=yes), employed six months prior to imprisonment (0=no, 1=yes); marital status as single (0=no, 1=yes); children (0=no, 1=yes); first time in prison (0=no, 1=yes); and number of previous imprisonments. Juvenile histories were examined and included: parental imprisonment (0=no, 1=yes); ever placed in care (0=no, 1=yes); juvenile detention (0=no, 1=yes); age at first juvenile detection; and number of juvenile detentions. Summary health variables were also examined and included: self-reported mental health issue(s)(0=no, 1=yes); self-reported sexually transmissible infection (0=no, 1=yes); number of self-report chronic health conditions (0=none, 1=one, 2=two, 3=three, 4=four or more); and overall physical and mental wellbeing in the past four weeks using the Short-Form Health Survey (SF-12) which is a reliable and valid 12 item measure [24].
Criminal careers
Using the survey and offending data, we examined the age at first crime, frequency, and variety/specialisation (i.e., for any crime, other (non-sexual, non-violent) crimes, violent (non-sexual) crimes, and sexual crimes). Variety or versatility of offending was calculated on a scale from 1 to 3 by examining the presence of the following crime types: sexual crimes; violent (non-sexual) crimes; and other (non-sexual, non-violent) crimes. Higher scores indicate higher offence variety. That is, men charged only for a sexual crime were scored with a 1, indicating no variety in the offending histories; while men with records of all three crime types received a 3, indicating the highest level of variety. Specialisation for each crime category (i.e., sexual; violent; and other) was calculated as a ratio of the number of charges in the respective crime category to the total number of charges in their offending histories, and then multiplied by 100 to reflect a percentage.
Statistical analysis
This study adopted an exploratory approach to data analysis to investigate differences between types of sex- and non-sex offenders in terms of demographic and descriptive characteristics and criminal career parameters. Pearson’s Chi-square tests were used for categorical variables. One-way Analysis of Variance tests and t-tests were used for continuous variables. The Scheffe test was used for post hoc comparisons because it is a conservative procedure that allows the examination of differences among groups, despite unequal group sizes. When the equality of variance assumption was not met, Tamhane’s T2 tests were used. An a priori alpha level of α = .05 for each statistical analysis was used. Data were analysed by using IBM® Statistical Package for the Social Sciences, version 24.
[i] There was a total of 2,831 participants in the three health surveys. For this study, the women (n = 504) who participated were excluded.
[ii] After the original linkage, data for a cohort of participants (n=506) was collected from the Re-Offending Database and were available until July 2016.