Page 41 - Contributed Paper Session (CPS) - Volume 7

CPS2027 Olayan A. et al.

Developing and validating risk prediction model
for re-offending of individuals with a severe
mental illness (Psychosis)
1
1
Olayan Albalawi ; Handan Wand ; Tony Butler
1,2
1 The Kirby Institute, University of New South Wales, Sydney, Australia.
2 Department of Statistics, Science Faculty, Tabuk University, Saudi Arabia.

Abstract
Objective- To develop and validate simplified risk score models for predicting
the risk of re-offending among those individuals who were diagnosed with a
severe mental illness (psychosis) prior to the first offence between 2001 and
2012 in New South Wales.
Methods- A cohort of 7,743 individuals were diagnosed with a severe mental
illness (Psychosis) prior to the first offence in New South Wales from 2001 to
2012. Individuals were randomly assigned to either a development (67%) or
an internal validation dataset (33%). The primary outcome was a reoffending
status. Cox regression models were used to create a risk prediction algorithm
from the development dataset. It was internally validated using standard
statistical measures.
Results- In the risk prediction model, six factors were identified as significant
of re-offending: age at the first offence, Indigenous status, type of a severe
mental illness, contact with mental health service after the first offence, the
outcome of the first offence and the type of the first offence. A score of ≥ 10
was selected as the optimum cut-point with 72% (43%) and 89% (19%)
sensitivity (specificity) for development and validation datasets, respectively.
Conclusion - A new risk score was predictive of re-offending for those
diagnosed with a severe mental illness and could help in local care and clinical
research setting.

Keywords
Risk Prediction; offending; Severe mental illness

1. Introduction
Risk prediction models are frequently used in clinical and public health
settings in order to identify those who are at risk of a disease of interest.[1-4]
Besides unstructured clinical assessments by mental health experts, more than
100 structured tools have been developed and routinely used in clinical and
justice system settings to predict the probability of future offending.[5] The
majority of these tools have been primarily designed to predict the likelihood
of future criminal behaviour based on evaluations of large numbers of
cognitive and antisocial behaviours.[6-8] For example, the Comprehensive
Health Assessment Tool (CHAT)[9] was developed as a standardised,
29 | I S I W S C 2 0 1 9