Objective: The aim of this study was to present scoring rules for predicting DSM-IV mental illness in the previous 12 months using the Kessler Psychological Distress Scale in the Australian population.
Method: The method described in Kessler et al. was closely followed using data from the 2007 Australian Survey of Mental Health and Wellbeing. A series of 93 nested logistic regression models were generated and compared to estimate the predicted probabilities of mental illness for each survey respondent using scores on the Kessler Psychological Distress Scale. The best model was selected using information theoretic criteria. Concordance between the predicted probabilities of mental illness generated by the best models with DSM-IV defined mental illness, assessed using the Composite International Diagnostic Interview, was determined using receiver operating characteristic analysis.
Results: The best fitting models were found to contain the quadratic form of the Kessler Psychological Distress Scale (both 6 item and 10 item versions). Age was found to be significant in the model predicting mood, anxiety, and substance use with serious impairment using the 6 item version whilst age and gender was found to be significant in the model for the 10 item version. The concordance between the predicted probabilities of mood, anxiety, and substance use with serious impairment generated from the best models and DSM-IV mood, anxiety, and substance use with serious impairment was within an acceptable level for both versions. Results were similar when predicting DSM-IV mood, anxiety, and substance use without seriousness indicators and DSM-IV anxiety and depression. The performance of predicted probabilities was then examined in various sub-populations of the Australian population.
Conclusions: Using a logistic regression model, the Kessler Psychological Distress Scale can be used to generate predicted probabilities of mental illness with an acceptable level of agreement in Australian-based population studies where it is not feasible to conduct a comprehensive assessment.