Researchers from Aarhus University unveil an AI algorithm designed to predict schizophrenia and bipolar disorder, potentially revolutionizing early diagnosis and personalized treatment strategies in mental health care.
Schizophrenia and bipolar disorder are debilitating mental health conditions that often emerge in early adulthood. While effective treatment options exist, they heavily rely on prompt and accurate diagnosis — a challenge that has perplexed the medical community for years. Now, research from Aarhus University and Aarhus University Hospital – Psychiatry, published in the journal JAMA Psychiatry, may provide the answer, courtesy of artificial intelligence.
“It’s a difficult clinical challenge to solve, but we have given it a try, and the results of this study show that we are on the right track,” Søren Dinesen Østergaard, a professor in the Department of Clinical Medicine at Aarhus University who led the research team, said in a news release.
The study leverages electronic health record data from 24,449 patients initially treated for less severe mental conditions like anxiety and depression.
From this dataset, the researchers developed a machine-learning algorithm poised to identify the likelihood of patients being diagnosed with schizophrenia or bipolar disorder within the next five years.
“If the algorithm indicates a high likelihood of developing schizophrenia or bipolar disorder within the next five years, healthcare staff can focus their examination on symptoms associated with these disorders — potentially leading to earlier diagnosis and the initiation of targeted treatment,” added Østergaard.
To achieve this, the algorithm parsed over 1,000 variables extracted from the health records, including diagnostic information, medication histories and text from clinical notes.
The results were promising but highlighted areas for improvement. For instance, for every 100 patients flagged as high risk, approximately 13 were diagnosed with the mental disorders within five years. In contrast, for those labeled low risk, about 95 did not receive such a diagnosis.
“This level of accuracy is probably not sufficient for the first version of the algorithm to be used in clinical practice, but we have a good idea of how to improve it. The key appears to be a more sophisticated analysis of the text in the clinical notes,” Østergaard added.
Interestingly, the most predictive components turned out to be specific terms in clinical notes. Words referring to symptoms like social withdrawal and auditory hallucinations were significant indicators, aligning with what health care providers already recognize as signs of severe mental illnesses.
“The 10 factors that contribute the most to the predictions all come from the clinical notes. These include words describing symptoms such as social withdrawal and auditory hallucinations, as well as words describing admissions to psychiatric hospitals — clear indicators of severe mental illness. This makes perfect clinical sense,” Østergaard added.
The research team plans to enhance their algorithm using advanced language models that understand entire sentences, akin to those driving technologies like ChatGPT. Such improvements could make the diagnostic tool robust enough for clinical usage, offering a transformative approach to mental health care.
“We are optimistic that this technology can make our predictions of schizophrenia and bipolar disorder precise enough for future versions of the algorithm to support clinical practice. This is an opportunity we will definitely pursue,” Østergaard concluded.