Back Groundbreaking machine learning algorithm accurately predicts mental health crises

Groundbreaking machine learning algorithm accurately predicts mental health crises

The study was carried out by Koa Health together with Birmingham and Solihull Mental Health NHS Foundation Trust using anonymized data collected over seven years. Roger Garriga, a PhD student in the Department of Information and Communication Technologies at UPF, participated in the research. The results obtained were able to correctly predict more than half of the seizures 28 days in advance, which was clinically valuable in 64% of the cases.
17.05.2022

Imatge inicial

Koa Health, a provider of digital mental healthcare solutions, and Birmingham and Solihull Mental Health NHS Foundation Trust, today announced the results of a groundbreaking machine learning algorithm that has accurately predicted mental health crises in a real world NHS setting. The results of the study suggest that such technology could allow for early intervention and improve patient outcomes.

Koa Health developed a machine learning model that uses anonymised electronic health records to continuously monitor patients for risk of a mental health crisis. The model has been found to correctly predict more than half of the crises 28 days in advance, without a considerable false positive rate. A following six-month study evaluated Koa Health’s algorithm’s use in clinical practice. In as many as 64% of cases, the model proved clinically valuable, either in terms of managing caseloads or mitigating the risk of crisis.

Roger Garriga, a PhD student in the Department of Information and Communication Technologies (DTIC) at UPF, is the first author of the research, published in the journal Nature Medicine. Garriga explains that "we have analyzed the electronic medical records of more than 17,000 mental health patients to develop a machine learning model that predicts the next mental health crises in the next four weeks. Our algorithm was designed to provide a weekly updated risk score for all patients with a history of relapse to support clinical decisions with a sufficient window of time for clinicians to implement preventive measures." He adds that "this research shows how collaboration between data scientists and healthcare professionals can lead to preventive care, having a major impact on patients' lives."

"We have analyzed the electronic medical records of more than 17,000 mental health patients to develop a machine learning model that predicts the next mental health crises in the next four weeks"

Results of the study come during Mental Health Awareness Month and follow the UK’s Mental Health Awareness Week which acknowledged the nearly one billion people worldwide who live with a mental disorder. With the global mental health emergency considerably exacerbated by the COVID-19 pandemic, healthcare systems face a growing demand for mental health services and seek solutions to allow for preventative measures. Koa Health is exploring the incorporation of crisis prediction into its comprehensive mental health service as part of its aim to use technology to radically improve access to high quality support for mental health. The next step is to validate the model in other health systems, and to undertake the larger scale studies that will be required to obtain regulatory approval.

Dr. Oliver Harrison, CEO Koa Health says, “Working as an NHS psychiatrist, I saw too many people arrive in our services through the Accident and Emergency Department.  By the time we saw them, they were usually very sick, and had often experienced problems with their relationships, finances, employment and housing.  Putting the pieces back together was almost always expensive, time-consuming, and painful for the people involved.  We were delighted to be chosen as a Wave 1 NHS Test Bed to partner with Birmingham and Solihull NHS Mental Health Foundation Trust.  We shared a vision of moving from reactive care - driven by crisis - to proactive care - driven by prediction and prevention.  We have delivered outstanding results in this trial and look forward to continuing the development work.  We believe this is the future of mental healthcare in the UK and worldwide.”

Dr. Aleksandar Matic, Koa Health’s R&D Director says, ”Prevention is widely seen as the holy grail of treating mental health problems. Using AI to predict mental health crises can enable timely and proactive instead of reactive interventions to mitigate or prevent mental health crises and truly help both patients and healthcare systems. We were so excited when the results from the clinical implementation of our model suggested that our predictions may have indeed prevented more serious symptoms. Providing the right intervention at the right time sits at the core of Koa’s approach.”

Reference word: Garriga, R., Mas, J., Abraha, S., Nolan, J., Harrison, O., Tadros, G. and Matic, A. (May 2022). "Machine learning model to predict mental health crises from electronic health records". Nature Mediicine

https://doi.org/10.1038/s41591-022-01811-5

Multimedia

Categories:

SDG - Sustainable Development Goals:

03. Good health and well-being
Els ODS a la UPF

Contact

For more information

News published by:

Communication Office