SA49. Improving Detection of Individuals at Risk for Psychosis: Beyond Attenuated Positive Symptoms

Abstract

Abstract Background: Inventory-based methods of detecting individuals at risk of psychosis (e.g., the prodromal questionnaire [PQ]) have been investigated in a series of studies, but the preponderance of studies has relied on clinically referred samples. The present study sought to (1) determine how the PQ compares to the structured interview for psychosis risk syndromes (SIPS) in a nonclinical sample and (2) determine whether adding nonpsychosis questionnaires to the PQ improves detection of individuals at clinical high risk (CHR) for psychosis. Methods: In a nonclinical sample of undergraduates, a series of questionnaires were administered to 2836 participants, and then a subsample was administered the SIPS based on high and low endorsement of positive symptoms on the PQ, using cutoffs from previous clinical studies (51 PQlow and 70 PQhigh). First, an LCA was conducted using the PQ and 19 other clinical and risk factor-based measures (all of which have been associated with psychosis). Second, we examined whether LCA classes were associated with increased the odds of being classified as CHR for psychosis compared to the PQ alone. Third, we combined variables that appeared to best differentiate between LCA classes to determine whether prediction could be improved by adding these variables to the PQ. Results: LCA analyses suggested that a 2-class model best fit the data, with Class 2 elevated on variables frequently associated with psychosis. Being in Class 2 was associated with a 14.5 increased odds of being at CHR for psychosis compared with a 5.6 odds using the PQ cut-off alone. Adding variables to the PQ substantially improved the positive predictive values (PPVs) and specificities of identifying those at CHR for psychosis without any changes in sensitivities or negative predictive values. Specifically, the PQ cutoff alone was associated with an 18.57% PPV, while PPVs ranged from 25 to 28.89% when adding 3 permutations of additional variables. True positives remained similar among the groups; however, adding additional variables substantially reduced the false-positive rates on the SIPS. Conclusion: Findings from the present study suggest that screeners that include nonpsychosis clinical measures (e.g., depression) have the potential to substantially improve detection of who is at CHR for psychosis, which could have major implications for future prevention and intervention studies.