referrals into services for offenders with intellectual disabilities: variables predicting community...

Copyright © 2010 John Wiley & Sons, Ltd 20: 39–50 (2010)DOI: 10.1002/cbm

Referrals into services for offenders with intellectual disabilities: Variables predicting community or secure provision

Criminal Behaviour and Mental Health20: 39–50 (2010)Published online in Wiley InterScience (www.interscience.wiley.com) DOI: 10.1002/cbm.755

DEREK CARSON1, WILLIAM R. LINDSAY1,5, GREGORY O’BRIEN2, ANTHONY J. HOLLAND3, JOHN L. TAYLOR2, JESSICA R. WHEELER3, CLAIRE MIDDLETON2, KAREN PRICE2, LESLEY STEPTOE1 AND SUSAN JOHNSTON4, 1University of Abertay, Dundee, UK; 2Northumbria University and Northumberland, Tyne and Wear NHS Trust, Newcastle, UK; 3University of Cambridge, Cambridge, UK; 4Rampton Hospital, Retford, UK; 5Castlebeck, Darlington, UK.

ABSTRACTBackground There is a need for research to promote an understanding among service developers on why people with intellectual disabilities (ID) are referred to offender services in order for them to receive appropriate assessment and treatment. Previous studies investigating referrals into forensic ID services have concentrated on referral sources and administrative variables such as legal status.Aims To construct a predictive model for choice of service referral based on a com-prehensive range of information about the clientele.Method We conducted a case record study of 336 people referred to community ser-vices and 141 to secure provision. We gathered information on referral source, demo-graphics, diagnosis, index behaviour, prior problem behaviours and history of abuse.Results Comparisons revealed 19 candidate variables which were then entered into multivariate logistic regression. The resulting model retained six variables: community living at time of referral, physical aggression, being charged, referral from tertiary health care, diverse problem behaviour and IQ < 50, which correctly predicted the referral pathway for 85.7% of cases.Conclusions An index act of physical aggression and a history of diversity of problem behaviours as predictors against the likelihood of community service referral suggest that professionals have similar concerns about people with ID as they do about their more average offending peers; however, the more severe levels of ID mitigated in favour of community referral, regardless. Offenders with ID tend to be referred within levels of service rather than between them, for example, form tertiary services into generic community services. Copyright © 2010 John Wiley & Sons, Ltd.

Carson et al.


40

Introduction

As part of the move towards deinstitutionalisation for people with intellectual disabilities (ID) and their use of ordinary community services, they have, where necessary, been encouraged to answer any criminal charges, with appropriate support and adaptation of procedures. Lund (1990), in a follow-up study of 91 offenders with ID on Statutory Care Orders in Denmark, found the incidence of sex offending doubled when comparing sentencing in 1973 and 1983. He sug-gested that this rise may have followed from people with ID who are no longer being subject to indeterminate detention in hospitals, and concluded that those more likely to be living in the community would also be those with more ability, and thus capacity, for offending, also with a greater likelihood of being subject to normal legal processes should they offend.

Noting the same trend in a study of 247 consecutive referrals to a community forensic ID service over a 13-year period, Lindsay et al. (2006) found that, in later years, a signifi cantly higher percentage of referrals came from criminal justice services rather than community or other services. Those referred in later years were also signifi cantly younger for all types of offence, probably refl ecting a trend in the later study years that ‘younger offenders with ID would not automatically be diverted to institutions, but might rather become involved with criminal justice agencies and the courts’ (p. 125).

These effects have undoubtedly infl uenced the profi le of services, to allow for their provision to offenders with ID at both local and national level. In turn, this has had an impact on the pathways into services for offenders with lower levels of function. In the 1960s and 1970s, it was likely that a large number of individu-als who had allegedly committed offences would have been diverted at an early stage from the criminal justice system into health-care institutions. Previously, large institutions for people with ID had a number of locked wards for such refer-rals. Because these institutions no longer exist, a range of pathways into services has developed, including entry into the criminal justice system and continuation through trial; some diversion from the court and criminal justice system; entry into statutory services; referral to private secure accommodation for individuals with ID and/or severe challenging behaviour; and, for some less serious offences, diversion into generic community learning disability services. There has been little research to date, however, indicating which personal, service or offence characteristics determine an individual’s pathway into services. It might be assumed that it would be the severity of the presenting offending or challenging behaviour, but Holland et al. (2002) and Sturmey et al. (2004) noted that it might also be related to the extent and expertise of local service provision. Much earlier, Walker and McCabe (1973) had noted a high prevalence of sex offenders and arsonists in secure ID services.

Hogue et al. (2006) investigated variables predicting admission to maximum security hospital for offenders with ID, but all of the variables reported by this

Referrals into services for offenders with intellectual disabilities


41

research group were either administrative or offence related. Therefore, they reported previous secure service admissions, nature of order for detention, Mental Health Act (MHA) 1983 category, and a range of offence-related variables such as the nature of the offence and relationship to the victim. Therefore, it is unsur-prising that the main variables contributing to the prediction model were admin-istrative and offence related. Already being in a secure hospital, MHA classifi cation of personality disorder, being a restricted patient under the MHA, receiving compulsory treatment under the MHA and having been treated under conditions of probation or civil detention were all highly correlated with entering high security, and made signifi cant contributions to the predictive model.

Our study investigated the effect of a wider range of variables in determining the service to which an individual is referred, in a large cohort of offenders with ID referred to a range of community and secure settings. In particular, we inves-tigated factors predicting referral to community or secure provision. Because of the large number of variables investigated, we have reported on the bivariate differences between groups elsewhere (Lindsay et al., 2009). We divided the cohort into: those referred to generic community services which had experience of receiving offenders; specialist forensic ID community services; and low/medium and high secure hospital services for offenders. The preliminary results are sum-marised in the methods below where the groups of variables are also described.

Method

Location of the study

We studied 24 separate services with a long history of accepting people with ID who had offended or shown similar behaviours in community or secure settings. They were in the East of England, the East Coast of Scotland and the North East of England, and, as such, included two high security hospitals with specialist units for people with ID. Data collection lasted for 1 year, 2002, in each service except the high security hospitals; there, individuals were referred at a slower rate, so the study was conducted over 2 years.

Materials and procedure

Four research assistants (RAs) were allocated to each of the main study sites and liaised with local clinical teams to extract information from the clinical fi les. The RAs used a standard form to gather all information. The form was guided by a manual (copies available from [email protected]). Data were col-lected on demographics, ethnicity, level of learning disability, possible medical diagnoses, psychiatric diagnoses, abuse experienced in childhood, living circum-stances, employment/occupation, referring agent, level of service referred to, dis-tance between usual residence and referred service, legal status on day of referral,

Carson et al.


42

index behaviour, age at time of index behaviour, charges, legal status on day of index behaviour, previous problematic behaviour and previous offending.

All the RAs had at least graduate qualifi cations in psychology. They under-took a dedicated 1-week training course in order to ensure the consistency and reliability of data collection across sites. Three anonymised cases were chosen by the grantholders for their diversity and complexity. One of the cases was ran-domly selected and was presented to the four RAs. As a group, the RAs discussed how they would approach the search of the case notes and how they would rate each item. If there were any ambiguity or disagreement, the group referred to the manual and worked through the issues until there was agreement on the decision they would take. Where this was not possible, the manual was returned to the clinicians and more specifi c guidance on coding was added. The second case was independently completed by the RAs and their responses were compared. Once again, when there was disagreement in their coding, the RAs discussed the rea-soning that led to their decision. After this second run through, the group were satisfi ed that the manual contained suffi cient detail to allow reliable coding from the case note information. The third case was used to test reliability. Each RA, again independently, completed the questionnaire based on this fi nal set of case notes. Reliability was calculated by comparing the responses of all four RAs to each question. Eighty-fi ve per cent of items were rated the same way by all four RAs.

The sample

Four hundred seventy-seven people were referred to the services during the study period. Full details of the basic analysis of the resulting information are available in Lindsay et al. (2009). In brief, 336 people were referred to community services while 141 were referred to secure services, 74% of them men. Average age on the day of referral was 33.0 years [standard deviation (SD) = 12.4 years; range 16–82 years]. Level of ID was stratifi ed: 7% had severe ID, 8% moderate, 51% mild, 17% borderline and 5% had average intellectual ability (IQ above 80). Forty-six per cent of referrals presented with at least one co-morbid adult psychiatric disorder. Nearly a third (31%) had some form of legal status on the day of the index act. Legal status included any court disposal or being under any section of the mental health legislation. Actual numbers in each category were small, so they were combined into one categorical variable of having legal status or not.

There were signifi cant differences between referrals to community generic services, specialist community forensic ID services, low/medium secure provision and high secure provision (chi-squared comparisons or Fisher’s exact probability where numbers in any cell fell below fi ve). There were signifi cantly more women, individuals with severe or profound ID, and signifi cantly fewer people under formal legal status in the community services than in the secure ones. In terms of the index behaviour, physical and verbal aggression and property damage were



43

all more common in the low/medium security and community generic groups; cruelty or neglect of children was confi ned to the community groups. Contact sex offences were most prevalent in the high security hospital group and were least so in the community generic group. There were no signifi cant differences between groups in relation to non-contact sex offences, fi re setting, theft, road traffi c offences or substance abuse.

Prior problem behaviours followed a different pattern from the index acts. People in both secure provision groups were signifi cantly more likely to have been physically aggressive than people in either community group, but their rates in verbal aggression were only higher than the community forensic group. The low/medium security group had higher rates of property damage incidents than all other groups. A history of any sexual offence was signifi cantly more common in the high security than the community generic group. The secure provision groups generally had higher levels of previous fi re starting, theft and substance abuse than the community groups. All three forensic specialist groups had sig-nifi cantly higher levels of previous offences than the community generic group; the secure provision groups were signifi cantly younger than the others for age at time of fi rst recorded problem. The community forensic and high secure groups had a higher rate of charges brought against them than the low/medium secure or community generic groups.

The only differences in psychiatric diagnoses between groups were that the low/medium security group had a signifi cantly higher rate of psychotic illness than the community specialist forensic group and the high security hospital group had a higher rate of personality disorder. The secure hospital groups had higher rates of non-accidental injury and ‘any abuse’ prior to the age of 15 than the community groups. For the source of referral, differences were as expected, with the community groups having a higher rate of referral from community sources, such as social services, the secure hospital groups having a higher rate of referral from other tertiary health-care sources, and the specialist forensic ser-vices having higher referral rates from court and other offender services than non-forensic groups.

Data analysis: Selection of predictor variables

From the preliminary fi ndings just described, a number of candidate variables were drawn up as associated with group membership at the bivariate level. For the purposes of the regression analysis, several were recoded into dichotomous variables (e.g. the 12 different co-morbid psychiatric disorders rated were com-bined into the single predictor variable: any co-morbid psychiatric disorder in adulthood/no such disorder). These dichotomous variables were then re-analysed to check the remaining differences between community referrals and secure referrals, using separate chi-squared for independence tests. Three variables were continuous and were also included: age at the time of the index behaviour; age

Carson et al.


44

at the fi rst recorded problematic behaviour; and the number of different types of problematic behaviour in the offending histories. The effects of the continuous variables were analysed using independent t-tests of difference between the two main groups – community and secure referrals – without Bonferroni or other type of correction being made. A multivariate binary logistic regression was then conducted with the membership of the community or secure group as the depen-dent variable, and all those dichotomous and continuous variables that were identifi ed as being individually associated with membership of the two groups as the independent variables.

Ethical approval

Ethical approval was granted for England under a Section 60 exemption of the Health and Social Care Act 2001, and for Scotland under Caldicott Guardian approval, meaning that it was not necessary to locate each person for individual consent to access their case record, but strict observance of individual confi den-tiality in any reporting was observed.

Results

Nineteen candidate variables were found to be associated, positively or negatively, with community-based referrals, as shown in Table 1. The strength of the associa-tion between the variables was categorised as weak, moderate, strong or very strong, based on the interpretation of the Cramer’s V statistic.

Binary logistic regression predicting referral to community and secure services

The fi nal regression model is displayed in Table 2. A total of 406 cases had full data sets and the full model signifi cantly predicted community or secure referrals (Omnibus chi-squared = 224.46, df = 6, p < 0.001). The Hosmer–Lemeshaw test confi rmed that the model was a good fi t (p = 0.38). The model accounted for between 42.5% and 60% of the variance in community or secure referrals, with 93.3% (sensitivity) of the community referrals and 68.5% (specifi city) of the secure referrals successfully predicted. Overall, 85.7% of the referrals were accu-rately predicted by this model (see Table 3).

Both outliers and infl uential cases can have detrimental effect on a fi nal model. To establish if either were a problem, a second model was run with this data set, excluding outliers (standardised residuals greater than 3.0) and infl uen-tial cases (Cook’s statistic greater than 1.0). Only six cases were potentially problematic according to these criteria, and a second model produced 87.3% of valid cases. Given that this correct prediction rate was within 2% of the fi rst model, the fi rst model was interpreted. A further check to identify potential



45

problems of multicollinearity between the 19 predictive variables was undertaken by checking the tolerance statistics and variance infl ation factors (VIF) following a multiple regression analysis (see Pallant, 2005). All tolerance statistics were above 0.5 and all VIFs less than 2.3 indicating that the 19 variables were unlikely to be correlated with one another to the extent that it would be problematic for this analysis.

According to the model (Table 2), six of our variables reliably predicted com-munity-based referrals. It was nearly 13 times more likely that a community referral would have been made if the individual was living in the community at the time, and just over three times more likely if the person had a moderate, severe or profound level of learning disability. By contrast, there was an inverse relationship between being referred to a community-based service and being charged as a result of the index behaviour (4.5 times less likely), showing index physical aggression (just under half as likely), greater diversity of problematic

Table 1: The proportion of referrals to community and secure services on the 16 dichotomous predictor variables showing the percentage of each cohort, chi-squared, the signifi cance of the difference and the effect size based on the interpretation of the Cramer’s V statistic

Variable Community(SD)

Secure χ2 (df 1) ort-test (whereindicated)

P Effect size(Cramer’sV)

Female 30.1 15.6 10.85 <0.01 WeakAdult psychiatric

disorder38.1 65.2 29.47 0.001 Moderate

Abuse as child 29.5 46.8 13.2 0.001 WeakAbuse as adult 8.3 17.0 7.72 <0.01 WeakModerate to severe ID 21.7 5.7 16.11 <0.001 ModerateIn a relationship at IB 13.1 5.7 5.63 0.05 WeakCommunity living at IB 91.6 35.9 158.86 <0.0001 Very strongLegal status at IB 16.5 74.0 135.84 <0.0001 Very strongPhysical aggression at IB 43.5 60.3 11.27 <0.001 WeakCruelty/neglect 8.3 0.0 12.48 <0.001 WeakContact sex offence 12.2 19.9 4.73 <0.05 WeakFire starting 2.7 7.8 6.5 <0.05 WeakCharged 28.8 46.1 13.65 <0.001 WeakPrevious history of IB 72.9 90.0 16.09 <0.001 WeakPreviously charged 28.8 58.2 35.5 <0.001 ModerateTertiary health referral 13.1 52.5 81.335 <0.0001 StrongMean age at IB 33.9 (13.0) 31.0 (10.5) t (473) = 2.34 <0.05Mean age at fi rst IB 16.6 (11.9) 12.4 (7.4) t (324) = 3.39 <0.01Mean different types

of IB2.6 (1.8) 4.6 (2.1) t (475) = 10.91 0.001

ID = intellectual disabilities, IB = index behaviour, SD = standard deviation.

Carson et al.


46

behaviour history (1.6 times less likely) or being referred by a tertiary (secure) healthcare service (three times less likely).

Discussion

At fi rst reading, the results of our study may appear only to present a systematic confi rmation of what we feel we already know. It will not come as a surprise that presenting physical aggression or diversity of past problem behaviour increases

Table 2: Summary statistics for binary logistic regression. Reference group is community referrals. IB equals index behaviour

Variable B Wald p (df 1) EXP (B) 95% CI for EXP (B)

Lower Upper

Community living on day of IB 2.55 44.50 <0.001 12.79 6.05 27.04IB includes physical aggression −0.78 6.04 0.014 0.46 0.24 0.85IB – at least charged 0.149 18.58 <0.001 0.22 0.11 0.44Tertiary health service referral −1.12 9.82 0.002 0.33 0.16 0.66Diversity of problem behaviour −0.52 39.03 <0.001 0.60 0.51 0.70Moderate to severe ID 1.22 5.38 0.02 3.39 1.21 9.49

Test χ2 df POverall model evaluation

omnibus χ2224.46 6 0.001

Goodness of fi t test Hosmer & Lemeshow

8.57 8 0.38

Note: Regression model using Statistical Package for the Social Sciences Version 14. Cox and Snell R2 = 0.42. Nagelkerke R2 = 0.600.IB = index behaviour, 95% CI = 95% confi dence interval, EXP(B) = odds ratio for predictors.

Table 3: The observed and predicted frequencies for community and secure referrals by logistic regression with cut-off of 0.50

Observed Predicted Percent Correct

Community Secure

Community (282) 263 19 93.3Secure (124) 39 85 68.5Overall percent correct 85.7

Note: Sensitivity = 263/(263 + 19) = 93.3%. Specifi city = 85/(39 + 85) = 68.5%. False positives = 39/(263 + 39) = 12.9%. False negatives = 19/(19 + 85) = 18.3%.



47

the likelihood of referral to secure services. Interpretation is, however, less intui-tively straightforward if we consider the variables that have not emerged as pre-dictors. Insofar as people with ID offend, fi re raising and sexual offending have been particularly associated with them (e.g. Prins 1980; Cullen 1993; Day 1993). Both of these, although showing signifi cant differences between those referred to the community and those referred to secure provision, dropped out of the regression model, suggesting that their association is mediated through one of the other variables.

Studies on prediction with offender samples are generally concerned with future offending or, in the case of mental health studies, institutional violence, but nonetheless provided a useful basis for the methods used here (e.g. Harris et al., 1993). Here, they reviewed a number of candidate variables in discharged offenders who had and had not re-offended. Based on the variables which dif-ferentiated signifi cantly between those two groups, they constructed a model for predicting violent incidents. Although we applied that research approach here to service entry, the risk prediction fi ndings are worth refl ecting on, because one might expect referral to secure services to follow from the presence of established risk factors.

The Harris group found that previous aggression in childhood, family instabil-ity in childhood and the nature of the index incident all contributed signifi cantly to the predictive model, and went on to construct a risk assessment tool accord-ingly (Quinsey et al., 2005). More recent studies have shown these risk assess-ments to be signifi cant, valid predictors of future incidents in populations of offenders with ID (MacMillan et al., 2004; Quinsey et al., 2004; Gray et al., 2007; Lindsay et al., 2008). Others have emphasised the diversity of the types of offend-ing as a risk factor for future violence. For example, it is a major item in the Psychopathy Checklist – Revised (PCL-R; Hare, 1991), the PCL-R being an effective predictor of recidivism in a number of studies of offenders generally (e.g. Hemphill et al., 1998; Walters, 2003) and offenders with ID (Morrissey, et al., 2007). Thus far, referral to services in our cohort of people with ID did seem appropriate – with those with a violent index offence and/or a diverse problem behaviour history being preferentially referred to secure services, and this is perhaps reassuring. What may seem more surprising is that other adverse histori-cal factors, such as, more specifi cally, aggression in childhood or unstable family backgrounds, did not emerge as predictors of the nature of referral, even though these too have been identifi ed as risk factors for recidivism and our measures in this study were wide enough to have picked them up.

The other signifi cant variables in the prediction of the direction of referrals seemed more administrative. First, if the index behaviour led to a criminal charge, even if charges were subsequently dropped, then the probability of being referred to community services was reduced. This may just refl ect a wish to respond more cautiously in these cases, although at least in respect of numbers of charges historically, other risk assessment tools have indicated that charges

Carson et al.


48

may be a further useful indicator of risk of recidivism (e.g. Static-99, Hanson and Thornton, 1999). A second more administrative variable relates to tested IQ levels. In our study, having moderate to severe learning disability (i.e. tested IQ of 50 or more) predicted an increased probability of being referred to community services (see also Lindsay et al., 2009). This may be partly because an individual with this level of intellectual ability would be considered to lack mens rea, so specialist forensic services would be seen as less relevant. Staff may also have more confi dence in managing people from this group in a challenging behaviour service. It is of potential interest to add here that it is also clear from the data that some forensic ID services accepted a minority of participants with an IQ above the diagnostic range for ID, with an average tested IQ of 80 or above. It is not possible to speculate on the reasons here, but it is widely acknowledged that tested IQ is a guide to need for specialist ID services, and perhaps appropri-ately not used rigidly at these borderline levels. Nevertheless, this might be a useful area for future study. The third more administrative issue was the tendency to refer within levels of service. Community living predicted community service referral, and tertiary service residency made it less likely.

There are several limitations in a study such as this, the fi rst of which is the possible limitations presented by the variables studied. There may be some vari-ables which are extremely potent in predicting the level of restriction an offender with ID is referred to, but which we have not included, even though we recorded a far wider range of variables than in previous studies, such as Hogue et al. (2006). A further limitation of this study is common to all case note studies. They are all dependent on the accuracy and completeness of information recorded by others. In our study, the RAs were experienced in applied psychology and received extensive training on reviewing case note information to a high standard of rating reliability, but we could not infl uence the quality of the original informa-tion recording. Furthermore, although we rated a complete annual cohort of referral to community services and specialist forensic low and medium security services and a 2-year cohort of high security referrals, numbers in some data cat-egories were very small indeed. Combining these into the larger collectives may have led to some masking of potentially useful indicators. In other words, in order to gain power in relation to certain categories, we may have lost sensitivity. All of these limitations should promote caution in interpretation of the results and further research inquiry.

Acknowledgements

This study was funded by the UK Department of Health National Forensic Health Research and Development Programme – Research Grant no. MRD/12/45.

Note. Copies of the proforma for data collection are available from [email protected]



49

References

Cullen C (1993) The treatment of people with learning disabilities who offend. In Howells K, Hollin CR (eds) Clinical Approaches to the Mentally Disordered Offender. Chichester: Wiley pp. 145–164.

Day K (1993) Crime and mental retardation: A review. In Howells K, Hollin CR (eds) Clinical Approaches to the Mentally Disordered Offender. Chichester: Wiley pp. 111–144.

Gray NS, Fitzgerald S, Taylor J, MacCulloch MJ, Snowden RJ (2007) Predicting future reconvic-tion in offenders with intellectual disabilities: The predictive effi cacy of VRAG, PCL-SV and the HCR-20. Psychological Assessment 19: 474–479.

Hanson RK, Thornton D (1999) Static-99: Improving Actuarial Risk Assessments for Sex Offenders. (User report 1999–02). Ottawa: Department of the Solicitor General of Canada.

Hare RD (1991) The Hare Psychopathy Checklist – Revised. Toronto, Ontario: Multi Health Systems.

Harris GT, Rice ME, Quinsey VL (1993) Violent recidivism of mentally disordered offenders: The development of a statistical prediction instrument. Criminal Justice & Behaviour 20: 315–335.

Hemphill JF, Hare RD, Wong S (1998) Psychopathy and recidivism: A review. Legal & Criminological Psychology 3: 141–172.

Hogue TE, Steptoe L, Taylor JL, Lindsay WR, Mooney P, Pinkney L, Johnston S, Smith AHW, O’Brien G (2006) A comparison of offenders with intellectual disability across three levels of security. Criminal Behaviour & Mental Health 16: 13–28.

Holland T, Clare ICH, Mukhopadhyay T (2002) Prevalence of ‘criminal offending’ by men and women with intellectual disability and characteristics of ‘offenders’: Implications for research and service development. Journal of Intellectual Disability Research 46: (Suppl. 1), S6–S20.

Lindsay WR, Hogue T, Taylor JL, Steptoe L, Mooney P, Johnston S, O’Brien G, Smith, AHW (2008) Risk assessment in offenders with intellectual disabilities: A comparison across three levels of security. International Journal of Offender Therapy & Comparative Criminology 52: 90–111.

Lindsay WR, Steele L, Smith AHW, Quinn K, Allan R (2006). A community forensic intellectual disability service: Twelve year follow-up of referrals, analysis of referral patterns and assessment of harm reduction. Legal & Criminological Psychology 11: 113–130.

Lindsay WR, O’Brien G, Carson D, Holland AJ, Taylor JT, Wheeler JR, Middleton C, Price K, Steptoe L, Johnston S (2009) Pathways into services for offenders with intellectual disability: Childhood experiences, diagnostic information and offence variables. Criminal Justice and Behaviour (in press).

Lund J (1990) Mentally retarded criminal offenders in Denmark. British Journal of Psychiatry 156: 726–731.

MacMillan D, Hastings R, Caldwell J (2004) Clinical and actuarial prediction of physical violence in a forensic intellectual disability hospital: A longitudinal study. Journal of Applied Research in Intellectual Disabilities 17: 255–266.

Morrissey C, Mooney P, Hogue T, Lindsay WR, Taylor JL (2007) Predictive validity of psychopathy in offenders with intellectual disabilities in a high security hospital: Treatment progress. Journal of Intellectual & Developmental Disabilities 32: 125–133.

Pallant J (2005) SPSS Survival Manual: A Step by Step Guide to Data Analysis. 2nd edition. Maidenhead, Berkshire, UK: Open University Press.

Prins H (1980) Offenders, Deviants or Patients? The Study of Socio-Forensic Problems. London: Tavistock Publications.

Quinsey VL, Book A, Skilling TA (2004) A follow-up of deinstitutionalised men with intellectual disabilities and histories of antisocial behaviour. Journal of Applied Research in Intellectual Disabilities 17: 243–254.

Carson et al.


50

Quinsey VL, Harris GT, Rice ME, Cormier CA (2005) Violent Offenders, Appraising and Managing Risk. 2nd edition. Washington DC: American Psychological Association.

Sturmey P, Taylor JL, Lindsay WR (2004) Research and development. In Lindsay WR, Taylor JL, Sturmey P (eds) Offenders with Developmental Disabilities. Chichester: John Wiley pp. 327–350.

Walker N, McCabe S (1973) Crime and insanity in England. Edinburgh: University Press.Walters GD (2003) Predicting institutional adjustment and recidivism with the Psychopathy

Checklist factor scores: A meta-analysis. Law & Human Behaviour 27: 541–558.

Address correspondence to: Bill Lindsay, Consultant Psychologist, Dundee. Email: [email protected]

referrals into services for offenders with intellectual disabilities: variables predicting community...

Documents