Psychosis risk calculator predicts remission and antidepressant use, not just conversion

A widely used psychosis risk calculator does more than forecast who will develop psychosis, according to a new study, but its usefulness varies sharply across different clinical outcomes. The research, published in the journal Schizophrenia, tested the North American Prodrome Longitudinal Study risk calculator, known as the NAPLS-2 RC, in a fully remote longitudinal study of young people at clinical high risk for psychosis.

Risk calculators like the NAPLS-2 RC were built to estimate the probability that an individual at clinical high risk, often abbreviated CHR, will convert to full psychosis within a defined period. Conversion, however, represents only a minority of outcomes in this population. Many people at high risk experience other meaningful clinical changes, including worsening symptoms, remission, new mood disorders, relapse of mood problems, or the start of psychiatric medication. The study team asked whether a tool tuned for conversion could also anticipate these broader outcomes.

The analysis drew on 369 participants in the CAPR study, a fully remote longitudinal sample, among whom 23 converted to psychosis during follow-up. The researchers evaluated how well the calculator distinguished between people who did and did not experience each outcome, using the area under the curve, or AUC, a standard measure of discrimination in which 0.5 reflects chance performance and 1.0 reflects perfect prediction.

For conversion to psychosis within two years, the calculator achieved an AUC of 0.66, which the authors characterize as fair discrimination. That level of performance is consistent with the tool’s original purpose and suggests it retains reasonable accuracy even when assessments are conducted entirely remotely rather than in person.

The calculator also performed at a similar fair level for two outcomes beyond psychosis. It showed an AUC of 0.63 for predicting remission, meaning it carried some signal about who would see their symptoms improve, and an AUC of 0.63 for antidepressant initiation, indicating modest ability to anticipate who would begin taking antidepressant medication during the study period.

Performance fell to chance levels for several other outcomes. The calculator showed an AUC of 0.51 for illness progression, 0.56 for new onset of mood disorders, 0.54 for mood disorder relapse, and 0.49 for antipsychotic initiation. Values in this range indicate that the tool did no better than random guessing for these outcomes, despite capturing meaningful variation in conversion risk.

The authors interpret this pattern as evidence that the NAPLS-2 risk calculator demonstrates relative specificity for psychosis-related outcomes. In other words, the variables it combines, which reflect the clinical and demographic profile associated with elevated psychosis risk, appear to track processes tied to psychosis itself rather than general psychiatric deterioration or medication decisions across the board.

The study’s authors suggest the calculator may be useful as a screening tool to guide further evaluation. Because it discriminates fairly well for conversion, remission, and antidepressant initiation, clinicians and researchers could potentially use it to flag individuals who warrant closer monitoring, while recognizing that it offers little guidance for outcomes such as illness progression or antipsychotic treatment.

Several limitations shape how these findings should be read. The number of converters was small, at 23 within a sample of 369, which limits the precision of the conversion estimate. The sample came from a single fully remote study, and discrimination statistics in the fair range still leave substantial room for error at the individual level. The findings describe statistical associations measured in this cohort and do not establish that the calculator causes or determines any clinical trajectory. Nonetheless, the results extend the conversation about psychosis risk tools beyond a narrow focus on conversion, mapping where a well-validated calculator succeeds and where it reaches the limits of its predictive reach.

Subject of Research: Validation of the NAPLS-2 psychosis risk calculator for predicting conversion and broader clinical outcomes in clinical high risk individuals

Article Title: Beyond conversion: extending the NAPLS-2 risk calculator to predict clinical outcomes in CHR within the CAPR study

Article References: Parmacek, S. N., Pratt, D. N., Pokorny, V. J., Ellman, L. M., Gold, J. M., Silverstein, S. M., Strauss, G. P., Walker, E. F., Waltz, J. A., Powers, A. R., Woods, S. W., Schiffman, J., Corlett, P. R., & Mittal, V. A. (2026). Beyond conversion: extending the NAPLS-2 risk calculator to predict clinical outcomes in CHR within the CAPR study. Schizophrenia. https://doi.org/10.1038/s41537-026-00822-2

Image Credits: AI Generated

DOI: 10.1038/s41537-026-00822-2

Keywords: psychosis, clinical high risk, NAPLS-2 risk calculator, conversion to psychosis, remission, antidepressant initiation, illness progression, mood disorders, antipsychotic medication, CAPR study, risk prediction, schizophrenia