[Knowledge discovery in clinical psychiatry: learning from electronic health records]

Today, almost every psychiatric care institution registers information concerning the care they provide in an electronic health record (EHR). By analyzing these health care data with innovative and advanced techniques, they can be an important source of new knowledge in the near future, and thereby contribute to improving psychiatric care.

To investigate how data from EHRs can provide relevant knowledge and insights for psychiatric care.

We designed and discussed solutions for some technical, organizational and ethical barriers surrounding unlocking health care data, in order to make analysis possible. We then analyzed the obtained health care data using techniques from knowledge discovery, the process in which new and useful information is extracted from data. We used techniques from data visualization, machine learning and natural language processing, among others, to demonstrate which types of results can be achieved.

Our approach showed that it is possible to find new and interesting insights that are hidden in EHRs on an aggregated level, in collaboration with healthcare professionals and patients. In particular we showed how the risk of violent behavior can effectively and accurately be assessed based on clinical text in the EHR.

After addressing some of the important challenges surrounding analyzing EHR data, learning from data from EHRs is a new and interesting approach with clear potential for improving psychiatric care.