Transdiagnostic Risk Calculator for the Automatic Detection of Individuals at Risk and the Prediction of Psychosis: Second Replication in an Independent National Health Service Trust

Abstract

Background: The benefits of indicated primary prevention among individuals at Clinical High Risk for Psychosis (CHR-P) are limited by the difficulty in detecting these individuals. To overcome this problem, a transdiagnostic, clinically based, individualized risk calculator has recently been developed and subjected to a first external validation in 2 different catchment areas of the South London and Maudsley (SLaM) NHS Trust.

Methods: Second external validation of real world, real-time electronic clinical register-based cohort study. All individuals who received a first ICD-10 index diagnosis of nonorganic and nonpsychotic mental disorder within the Camden and Islington (C&I) NHS Trust between 2009 and 2016 were included. The model previously validated included age, gender, ethnicity, age by gender, and ICD-10 index diagnosis to predict the development of any ICD-10 nonorganic psychosis. The model's performance was measured using Harrell's C-index.

Results: This study included a total of 13702 patients with an average age of 40 (range 16-99), 52% were female, and most were of white ethnicity (64%). There were no CHR-P or child/adolescent services in the C&I Trust. The C&I and SLaM Trust samples also differed significantly in terms of age, gender, ethnicity, and distribution of index diagnosis. Despite these significant differences, the original model retained an acceptable predictive performance (Harrell's C of 0.73), which is comparable to that of CHR-P tools currently recommended for clinical use.

Conclusions: This risk calculator may pragmatically support an improved transdiagnostic detection of at-risk individuals and psychosis prediction even in NHS Trusts in the United Kingdom where CHR-P services are not provided.

Keywords: psychosis; risk calculator; schizophrenia; transdiagnostic; validation.

Fig. 1.

Current detection strategies for individuals at risk of psychosis in secondary mental health care in South London, UK. The local early detection service (OASIS), which is embedded in the South London and Maudsley (SLaM) NHS Trust, runs an ongoing outreach to promote referrals on suspicion of psychosis risk. This strategy is highly inefficient and misses 95% of patients who are at risk and who will develop a first episode of psychosis over the ensuing 4 years.

Fig. 2.

Potential clinical use of the individualized, clinically based, transdiagnostic risk calculator in secondary mental health care. For any new patient accessing the local NHS Trust (South London and Maudsley, UK) clinicians will enter the predictors on the electronic case register, as part of their clinical routine. The calculator, embedded in the electronic system, would then use the predictors to estimate the individual risk of developing psychosis over time. This information would then be shared with clinicians through automated alerts, inform their decision making and promote appropriate referrals to the local early detection clinic (OASIS).

Fig. 3.

Cumulative incidence (Kaplan–Meier failure function) for risk of development of psychotic disorders in the Camden and Islington NHS Trust, UK. There were a total of 490 events (transition to psychosis). There were 212 events in the first 365 days, 123 events in the interval 365–730 days, 63 events in the interval 730–1095 days, 44 events in the interval 1095–1460 days, 28 events in the interval 1460–1825 days, 14 events in the interval 1825–2190 days, 6 events the interval 2190–2555 days, and no events in the interval 2555–2851 (end of follow-up). The last transition to psychosis was observed at 2466 days, when 13212 individuals were still at risk. The point estimates for cumulative incidence of psychosis were at: 1 year, 1.61; at 2 year, 2.76; 3 year, 3.53; 4 year, 4.36; 5 year, 5.19; and 6 year 5.88 (95% CI: 5.27–6.57).