Abstract
Individuals at clinical high risk for psychosis (CHR-P) experience attenuated positive psychotic symptoms and impaired level of overall functioning. Currently, methods available for detecting individuals at risk are sub-optimal with early detection services for CHR-P individuals identifying only a small minority (5-12%) of first episode psychosis (FEP) cases prior to illness onset. Moreover, once CHR-P subjects have been identified, available methods of predicting their clinical outcomes have limited accuracy. This thesis aims to improve the detection of individuals at high risk of developing psychosis (Part A), and the estimation of psychosis risk in this population (Part B). It will also review gaps in current knowledge and provide directions for future research (Part C).Part A of this thesis aimed to improve the detection of individuals at risk of developing psychosis. In Chapter 2, I performed the largest replication study of a risk prediction model in psychiatry. As part of this, I validated the discrimination performance of an individualised transdiagnostic psychosis risk calculator that leverages electronic health record (EHR) data, previously developed in our group, in an international US dataset comprised of 2.4 million patients. This was the first replication of the transdiagnostic risk calculator outside of the UK. This validation indicated that the transdiagnostic risk calculator retained significantly better discrimination performance than chance (Harrell’s C = 0.68), highlighting its clinical transportability and potential for automated screening for individuals at risk for psychosis at scale in international settings. In Chapter 3, I assessed the feasibility of implementing the transdiagnostic risk calculator in a clinical EHR system. This represents one of the first clinical applications of a risk prediction model in a mental health setting. The primary barrier to successful use of prediction models in clinical care is clinician endorsement. As such, I investigated the real-world feasibility of screening all individuals accessing secondary mental healthcare at the South London and Maudsley NHS Foundation Trust (SLaM) with clinician adherence being the primary outcome. Clinician adherence to the calculator was high with 78% responding to recommendations of the transdiagnostic risk calculator, emphasising the feasibility of implementing the risk calculator to improve detection of individuals at risk of developing psychosis. Through these studies, I have advanced knowledge and the potential of real-world, automated, systematic detection of individuals at risk for psychosis.
Part B aimed to improve the prognostication of outcomes in CHR-P individuals. Chapter 5 presents the identification of robust, non-genetic factors that modulate transition risk in CHR- P individuals through the synthesis of the available evidence. The results showed that no factors met the criteria for the highest classification of evidence (class I, convincing). However, attenuated positive psychotic symptoms and global functioning were associated with highly suggestive evidence (class II) and negative psychotic symptoms with suggestive evidence (class III). The remaining factors were associated with either weak (class IV) evidence or were non-significant. Chapter 6 outlines the development, digital implementation and piloting of a novel multivariate assessment for non-genetic risk and protective factors for psychosis, termed the Psychosis Polyrisk Score (PPS). This assessment can be completed online, in less than 15 minutes, to provide an individualised estimate of exposure to non-genetic risk and protective factors for psychosis “en masse”. A simulated general population dataset was also used to show the range and distribution of scores. Pilot data found that individuals referred for a CHR-P assessment had higher PPS scores compared to healthy controls, highlighting the feasibility of its use in real world settings and potential for refining prognostication of outcomes, complementing the CHR-P assessment. Together, these studies advance the potential for refining prognostic estimates of clinical outcomes for CHR-P individuals.
Finally, in Part C, I discussed the collective implications of the findings from Part A and Part B, reviewed their limitations and considered how to overcome these in future work. I then discussed ways to take work from this thesis forward in order to further improve the identification of people at high risk and the prognostication of their outcomes.
| Date of Award | 1 Jan 2022 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Paolo Fusar-Poli (Supervisor) & Philip McGuire (Supervisor) |