Predicting type of psychiatric disorder from Strengths and Difficulties Questionnaire (SDQ) scores in child mental health clinics in London and Dhaka.

Abstract

A computerised algorithm was developed to predict child psychiatric diagnoses on the basis of the symptom and impact scores derived from Strengths and Difficulties Questionnaires (SDQs) completed by parents, teachers and young people. The predictive algorithm generates "unlikely", "possible" or "probable" ratings for four broad categories of disorder, namely conduct disorders, emotional disorders, hyperactivity disorders, and any psychiatric disorder. The algorithm was applied to patients attending child mental health clinics in Britain (N = 101) and Bangladesh (N = 89). The level of chance-corrected agreement between SDQ prediction and an independent clinical diagnosis was substantial and highly significant (Kendall's tau b between 0.49 and 0.73; p < 0.001). A "probable" SDQ prediction for any given disorder correctly identified 81-91% of the children who definitely had that clinical diagnosis. There were more false positives than false negatives, i.e. the SDQ categories were over-inclusive. The algorithm appears to be sufficiently accurate and robust to be of practical value in planning the assessment of new referrals to a child mental health service.