Can clinicians predict psychosis in an ultra high risk group?

Aims: Criteria for identifying people likely to develop a first psychotic episode are now used in many clinical services worldwide. In recent years within these services, there has been an increase in the practice of prescribing antipsychotic medication with the aim of reducing symptoms and preventing onset of full-blown disorder. This practice is based on clinical impression of an incipient psychosis, that is, a clinical judgment that a particular patient may soon progress to full-threshold disorder and may therefore benefit from antipsychotics. However, it is unclear how accurate this clinical impression is. If not accurate it could mean that individuals are receiving antipsychotics unnecessarily. In this study, we investigated the predictive validity of clinical impression of whether ultra high risk patients would develop frank psychosis. Methods: Experienced psychologists rated their clinical impression of incipient psychosis in 168 ultra high risk patients. Ratings were made upon entry to the PACE clinic, a clinical service for ultra high risk patients. Psychosis status over the subsequent 12-month period was established using the Comprehensive Assessment of At Risk Mental States or State medical records. Results: A total of 8.9% of the sample transitioned to psychosis over the 12-month follow-up period. There was a sensitivity of 0.80, specificity of 0.84, positive predictive value (PPV) of 0.32 and negative predictive value (NPV) of 0.98 for the prediction of psychosis using the clinical impression ratings. Conclusions: The results indicate that clinical impression is not sufficient for predicting psychosis outcome in ultra high risk cohorts and that ongoing rigorous research into predictors of outcome in such cohorts is required. The results also caution against the prescription of antipsychotic medication based on clinical impression of incipient psychosis. Future work should address the predictive validity of clinical impression with a larger sample and over a longer follow-up period.