T80. CARDIOMETABOLIC RISK PREDICTION ALGORITHMS AND THEIR APPLICABILITY FOR YOUNG PEOPLE WITH PSYCHOSIS: A SYSTEMATIC REVIEW AND ILLUSTRATIVE EXAMPLE USING ORIGINAL DATA FROM A POPULATION-BASED BIRTH COHORT

Abstract

BackgroundCardiometabolic risk prediction algorithms are used in clinical practice. Young people with psychosis are a high-risk group for developing cardiometabolic disorders, but it is unclear whether existing algorithms are suitable for this group.MethodsWe conducted a systematic review employing PRISMA criteria to identify studies reporting the development and/or validation of cardiometabolic risk prediction algorithms for general or psychiatric populations. A narrative synthesis was conducted to compare algorithms and consider their suitability for young people with psychosis. In addition, we used data from 3,470 young adults aged 18 years from the ALSPAC birth cohort to illustrate the impact of age on model performance of QDiabetes, an established algorithm.ResultsHaving screened 6,609 studies, we included 57 risk algorithms designed for type 2 diabetes, cardiovascular disease or stroke, all of which were developed/validated in relatively older participants. Three algorithms featured psychiatric predictors and could be used for young people with psychosis. However, in all of three, age was weighted to a much greater extent than other risk factors. Furthermore, using ALSPAC data, we report that QDiabetes significantly under-predicted cardiometabolic risk in young people. Increasing the sample age to 50, leaving all other predictors unchanged, improved algorithm calibration markedly.DiscussionExisting cardiometabolic risk prediction algorithms are heavily weighted on age and so under-predict risk in young people. A new or recalibrated algorithm is required for young people with psychosis that appropriately balances the weighting of relevant risk factors.