New research shows that machine learning can help predict the onset of schizophrenia and bipolar disorder by analyzing routine clinical data from electronic health records.
The study, led by Aarhus University researcher Lasse Hansen, demonstrated that artificial intelligence (AI)-based tools are better at predicting schizophrenia than bipolar disorder, but within the next five years It has been demonstrated that both onsets can be predicted to a reasonable degree. Of accuracy.
“Schizophrenia and bipolar disorder are severe mental disorders that often impair the ability to live a normal life,” the author writes Jama Psychiatry.
“Despite the usual appearance of late adolescence or early adulthood, diagnosis is often delayed by several years. A timely, accurate diagnosis is important as delayed diagnosis prevents the initiation of targeted treatments. Furthermore, the longer the period of untreated illness, the worse the prognosis will be.
Hansen and his colleagues wanted to assess whether using AI would help speed up diagnosis of people at risk for these conditions.
This study used electronic health record data from everyone ages 15-60 who had at least two contacts with psychiatric services in the Central Denmark region for at least three months from early 2013 to late 2016. . Overall, this group included 24,449 people. They were between 24 and 42 years old, with 57% being female.
The researcher is xgboost Analyze the data. They first trained the model and then tested its effectiveness in different groups of participant data. The team found that the algorithm could accurately predict the onset of schizophrenia or bipolar disorder within five years.
The area under the Receiver Operator Curve (AUROC) test is a way of measuring how machine learning models can accurately convey differences between two groups. This algorithm was able to distinguish between 70% time and 64% of the test group from positive and negative cases in the training set.
When the risk of the two conditions was assessed separately, the AUROC score for schizophrenia prediction was superior to the bipolar score of 62% at 80%. Generally, an AUROC score of 70% or more is considered fair, but this depends on the particular test.
“These findings can be realized to detect progression to schizophrenia through machine learning based on routine clinical data, which could reduce diagnostic delays and duration of untreated illnesses. It suggests that “can be implemented in clinics.
