Use of artificial intelligence to search the clinical narrative for information
Information overload has become a challenge for healthcare professionals, because not being able to retrieve critical information may compromise patient safety and the quality of health care provided.
Geir Thore Berge
PhD Candidate
Geir Thore Berge of the Faculty of Social Sciences at the University of Agder has submitted his thesis “Methods for automated structuring of health information for clinical decision support”, and will defend the thesis for the PhD-degree Thursday 20 August (trial lecture) and Friday 21 August (public defense) 2020.
He has followed the PhD Programme at the Faculty of Social Sciences with Specialization in Information Systems at the University of Agder.
Berge is a fellow from the Public Sector Ph.D. scheme, with funding from Sørlandet Hospital HF and the Research Council of Norway.
Summary of the thesis by Geir Thore Berge:
Use of artificial intelligence to search the clinical narrative for information
Today’s electronic health records (EHRs) often contain hundreds or even thousands of pages, and unstructured information (i.e., free text) may typically account for as much as 80-90% of the information.
Information overload
Information overload has become a challenge for healthcare professionals, because not being able to retrieve critical information may compromise patient safety and the quality of health care provided.
Increasing the proportion of the patient record that is structured, thereby making it computable (i.e., machine-readable), has the potential to help overcome this challenge and also cause other major positive impacts on healthcare. For example, computer systems may leverage such structured data to automatically recognize, compile, and analyze critical health information to provide healthcare professionals with advanced clinical decision support.
The research has focused on the application of machine learning and rules to automate the structuring of health information in EHRs for clinical decision support.
To be useful in a healthcare context, the results produced by such methods should be both accurate and interpretable. Moreover, the methods should not be demanding in terms of clinical and technical resources required. Furthermore, there are challenges associated with ethical issues, privacy, and legislation that must be adequately addressed.
A clinical decision support system
To explore the methods’ possibilities and limitations, the research project has developed a clinical decision support system that automatically searches for and classifies health information in EHRs. The system is the first of its kind in Norwegian healthcare to use artificial intelligence. The system has been tested by doctors and nurses at Sørlandet Hospital Trust to search for patient allergies, with promising results.
The lessons learned during the research project have been distilled and formulated into design principles which may serve as guidance for future similar implementations in healthcare. The results of the research have so far been published in three research articles, and several more are in the pipeline. Because the research has also created national interest, it has further been extensively disseminated outside of academia and through media reports.
Opponent ex auditorio:
The chair invites members of the public to pose questions ex auditorio in the introduction to the public defense, with deadlines. Questions can be submitted to the chair Anne Halvorsen on e-mail anne.halvorsen@uia.no
The thesis are available:
Please contact Cecilie Rygh Mawdsley at e-mail cecilie.mawdsley@uia.no, to obtain a PDF or a link to the thesis.
Disputation facts:
The Candidate: Geir Thore Berge (1975, Kristiansand) Bachelor degree (1988) and Masters degree (2003) from The University of Agder. Master Thesis Title:"Criteria for Enterprise Portal Solutions in Healthcare". Berge is an employee at the Sørlandet Hospital HF from 1997, and are now working as an ICT Advisor.
The trial lecture and the public defence will take place at internett, via the Zoom conferencing app (link below) Thursday 20 August (trial lecture) and Friday 21 August (public defense) 2020.
Dean of the Faculty of Social Sciences at the University of Agder, Anne Halvorsen, will chair the disputation.
The trial lecture at 15:00 hours Thursday 20 August
Public defence at 15:00 hours Friday 21 August
Given topic for trial lecture: “Current status and methods for interpretable AI in medicine”
Thesis Title: “Methods for automated structuring of health information for clinical decision support”
Opponents:
First opponent: Professor Gondy Leroy, PhD, Management Information Systems, University of Arizona, USA
Second opponent: Centre Director Stein Olav Skrøvseth, PhD, Norwegian Centre for E - Health Research
Professor Carl Erik Moe, Department of IInformation Systems, University of Agder, is appointed as the administrator for the assessment commitee.
Supervisors were Professor Bjørn Erik Munkvold, University of Agder (main supervisor), Professor Ole-Christoffer Granmo, University of Agder, Professor Emeritus Rune W. Fensli, University of Agder and Tor Oddbjørn Tveit, MD, Sørlandet Hospital HF / CAIR at the University of Agder (co-supervisors)