Predicting Clinical Outcomes for Patients with Mental Illness using Electronic Health Records:

Can machine learning enhance treatment and prognosis in mental healthcare?

Psychiatric services in Denmark are under significant strain due to rising demand and limited resources. At the same time, treatment of mental disorders often remains suboptimal. Machine learning may offer a path forward by enabling more precise, data-driven approaches. Denmark's homogeneous and highly digitised healthcare system provides a unique opportunity to lead innovation in this space.

At the PSYCOP group, we explore the feasibility of developing machine learning algorithms trained on electronic health records designed for real-world clinical implementation. Our projects aim to predict adverse outcomes – such as the use of coercive measures – with the goal of enabling timely interventions. Accurate predictions can help clinicians identify patients at high risk, prevent harmful events, and optimize the allocation of scarce resources.

In this presentation, I will outline our general workflow and showcase one of our projects. We will also discuss key challenges in designing, developing, and implementing AI tools for healthcare. How can we better leverage the abundance of free-text data in psychiatry? How do we ensure effective collaboration between these technologies and human clinicians? How should we evaluate these tools, and do we have evidence that their benefits outweigh their costs?

Speaker: Jakob Grøhn Damgaard

Zoom link: https://aarhusuniversity.zoom.us/j/65788597296