A closer look at predicting psychosis in high-risk young people

University of Adelaide researchers have taken a closer look at how to predict transition to psychosis in young people who meet a specific set of subthreshold psychotic symptoms and syndromes, also referred to as ultra-high risk (UHR) criteria.

UHR research provides greater insight into risk and protective factors for psychotic disorders, mechanisms that drive the onset of these disorders, and the most effective treatments for delaying or preventing the onset of psychotic disorders and other adverse outcomes in young people at high risk.

Associate Professor Scott Clark, Head of the Discipline of Psychiatry, and Postdoctoral Research Fellow Dr. Simon Hartmann analyzed the data of 1,245 UHR adolescents from 1995 to 2020 to create a transition to psychosis model. Their findings are published in the journal World Psychiatry.

“Youth mental health is currently a major global concern with increasing numbers of young people struggling,” said Associate Professor Clark.

“Early intervention has shown to improve outcomes for young people with emerging mental health illnesses.

“As part of our research, we try to find data-driven methods to better identify those at risk of a mental health illness to assist with early intervention.

“We require large datasets for our research like the one we have aggregated in our study in collaboration with Orygen, Melbourne, and the Center of Youth Mental Health at the University of Melbourne.

“The dataset represents Orygen’s accumulated work in UHR research over the past three decades and our analysis is supported by the NHMRC Center of Research Excellence (CRE) in Prediction of Early Mental Disorder and Preventive Treatment (PRE-EMPT) led by Professor Barnaby Nelson.

“It is so far the largest dataset on young people at risk of a psychotic disorder and will facilitate research into the development of data-driven methods in this field.”

Dr. Hartmann said the study shows the accuracy of data-driven methods deteriorates over time, due to changes in treatment and clinical processes, as well as population changes.

“Our research identified potential pitfalls in existing methods to ensure a safe and successful translation into clinical practice. This will help to get this field a step closer to implementing data-driven methods in mental health care services,” he said. “We expected performance to deteriorate over time based on research from other fields but were surprised by the extent and its clinical implications.

“We were concerned that temporal biases in existing models that are on the verge of being implemented in clinical practice may not lead to optimal decisions and could potentially do harm.

“Our results show that higher disorganized speech and unusual thought content as measured by Comprehensive Assessment of At-Risk Mental States (CAARMS), higher negative symptom severity, lower social functioning, and a longer duration of symptoms prior to UHR service entry are predictive of transition to psychosis.”

Now Dr. Hartmann and Associate Professor Clark know there is a problem with the deterioration of the data, and they’re turning their attention to how to correct it.

“Potential findings of this next stage will form part of a future of learning health systems that aim to constantly update knowledge from every patient by harnessing the power of data and analytics, and feed newly gained knowledge back to clinicians, nurses, and other health care stakeholders,” said Dr. Hartmann.