AI can predict involuntary admissions and pave way for prevention in psychiatric services

Can artificial intelligence aid in the treatment of mental illness?

A newly published study from Aarhus University and the Psychiatric Services in the Central Denmark Region suggests that the answer is “yes.”

Here, a research group has developed a machine learning algorithm that, by analyzing electronic health record data, has “learned” to identify patients at elevated risk of involuntary admission.

“This is a major step towards more targeted treatment in the psychiatric service. We believe this technology can improve our ability to help patients before they become so ill that involuntary admission becomes necessary,” says Professor Søren Dinesen Østergaard from the Department of Clinical Medicine at Aarhus University and the Psychiatric Services in the Central Denmark Region, who contributed to the study.

The findings are published in the journal Psychological Medicine.

A supplement to clinical assessment

The machine learning algorithm can, at the time of discharge from voluntary inpatient treatment, identify patients at high risk of involuntary admission within the following six months.

For every 100 patients the algorithm identifies as high risk, approximately 36 will be involuntarily admitted within the next six months. Conversely, for every 100 patients identified as low risk, about 97 will not be involuntarily admitted.

“The machine learning algorithm is not perfect, but it is accurate enough that we should consider whether it can be used as a decision-support tool. It is important to emphasize that the algorithm cannot replace clinical assessment but rather be used as a supplementary source of information, enabling more informed clinical decision-making,” says Østergaard.

“If the algorithm would identify a patient at high risk for involuntary admission at the time of discharge, we could, for example, plan a very close outpatient follow-up to detect and treat any deterioration in the patient’s condition as early as possible.”

Learning from thousands of patient cases

The study is based on electronic health record data from 50,634 voluntary inpatient treatments in the Psychiatric Services of the Central Denmark Region between 2013 and 2021.

The machine learning algorithm analyzed the relationship between approximately 1,800 variables from the electronic health records—including diagnoses, medication, prior involuntary measures, and clinical notes—and subsequent involuntary admission.

“This means that the machine learning algorithm has learned from past treatment of thousands of patients—to benefit future patients,” says Østergaard.

Early detection of physical illnesses

Predicting involuntary admission is just one example of how this technology can be used.

The research group’s findings also indicate that machine learning can be used to predict the development of cardiovascular disease and type 2 diabetes among patients receiving treatment in psychiatric services.

“The average life expectancy for people with severe mental illness is significantly shorter than that of the general population, with cardiovascular disease and type 2 diabetes contributing significantly to this excess mortality,” explains Østergaard.

“According to our research, machine learning may allow us to detect and treat these diseases earlier. In some cases, we might even be able to prevent them from developing.”

Machine learning requires big data

Machine learning relies on large datasets to ensure that the developed algorithms are sufficiently accurate.

In a newly launched project, the research group is investigating whether machine learning can predict cardiovascular disease and type 2 diabetes among hospital patients—regardless of the department they received treatment at—by analyzing electronic health record data from approximately 1.4 million adult patients from hospitals in the Central Denmark Region.

“Working with such large volumes of health record data comes with a great responsibility, which we take very seriously,” says Østergaard, emphasizing the significant potential of this research field.

“There is a huge amount of knowledge hidden deep within the health care data our society has generated over decades. Now, new technology can help bring this knowledge to the surface, where it can benefit individual patients.”