Predictive algorithms for psychosis found to be effective

Researchers have demonstrated that prospective screening of treatment-seeking patients using predictive algorithms can help determine individuals at increased risk of psychosis. Various prediction and prevention models exist across various disciplines of medicine such as cardiovascular disease, diabetes and cancer, and have led to reductions in morbidity and mortality. It is unclear; however, if preventive models can be applied to psychotic disorders. Cannon et al. sought to develop a multivariate risk prediction algorithm. They hypothesised that particular variables, such as genetic risk for schizophrenia, the severity of prodromal symptoms, the severity of non-specific symptoms, social and role functioning, and substance abuse, would contribute uniquely to the prediction of psychosis, which could then be developed into an algorithm providing higher positive predictive power (PPP) than prodromal symptoms alone. They conducted a longitudinal study with 30 month follow-up that used potential predictor variables found to be associated with risk of conversion to psychosis in previous studies using smaller samples. The current study included a total of 291 patients, all of whom were identified prospectively, were treatment-seeking and met the Structured Interview for Prodromal Syndromes (SIPS) criteria. The main outcome measure was the time to conversion to a fully psychotic form of mental illness. The study identified five key features at baseline that contributed uniquely to the prediction of psychosis. These were a genetic risk for schizophrenia with recent deterioration in functioning, higher levels of unusual thought content, higher levels of suspicion/paranoia, greater social impairment, and a history of substance abuse. Statistical analysis of these five predictive variables and their 26 possible combinations yielded the prediction algorithms. Of these, the algorithms that combined two or three of the variables resulted in large increases in positive predictive power compared with the prodromal criteria alone. Prediction algorithms incorporating combinations of three baseline variables (genetic risk for schizophrenia with recent functional decline, higher levels of unusual beliefs or suspiciousness, and greater social impairment) resulted in increases in PPP (74—81%) compared with SIPS criteria alone (35%). Prediction algorithms incorporating combinations of two baseline variables (genetic risk for schizophrenia with recent functional decline, and unusual thought content or impaired social functioning) had the highest PPP (69% and 61%, respectively), both of which were found to be substantially higher than that of the SIPS criteria (38% and 55%, respectively). It is possible to use predictive models for identifying individuals at risk for psychosis, and that the use of such prediction algorithms may enable more selective recruitment of patients into relevant treatment programs, the researchers concluded. However, they caution that the current results apply to treatment-seeking patients and thus cannot be translated to the general population. Reference...