On Friday 16 May at 13:00, Erik Perfalk defends his PhD dissertation entitled "Using machine learning for clinical prediction in hospital psychiatry". 

The PhD project focused on machine learning (ML)-based clinical prediction modelling in psychiatry. The aims were to develop a prediction model for involuntary admission based on electronic health records (EHRs), and to investigate the trust in this type of technology among patients and healthcare staff.

ML has recently seen major advances, partly due to its ability to handle large amounts of complex data. EHRs contain a large amount of data, both structured (e.g., diagnoses, lab tests etc.), and unstructured in the form of clinical notes. If effectively represented, this data could potentially support ML-based prediction models for a range of clinical outcomes. However, if such models are to be implemented in clinical practice, it is essential that both patients and healthcare staff have trust in them.

This dissertation consists of three papers that explore different aspects of clinical prediction modelling in psychiatry. The first paper uses ML to predict involuntary admissions among psychiatric patients based on electronic health record data. The best performing model is based on predictors using both structured and unstructured (text) data. The second and third paper uses survey experiments to examine whether providing basic information about ML-based clinical decision support systems enhances patient and staff trust in these tools.

The summary is written by the PhD student.

The defence is public and takes place in auditorium J116-113, Entrance J, Aarhus University Hospital. Please see the press release for more information. 

Contact

PhD student Erik Perfalk
Mail: erperf@rm.dk  
Phone: +45 61669696

Read full press release