Detect-HAI - Detection of Hospital Acquired Infections through language technology
(In Swedish: Detektion av vårdrelaterade infektioner genom språkteknologisk analys av elektroniska patientjournaler) 

Project manager: Hercules Dalianis
Participants from DSV:  Martin Hassel, Hideyuki Tanushi, Mia Kvist and Gunnar Nilsson
Participants from Karolinska University Hospital: Elda Sparrelid and Gunnar Ekeving
Participants from Uppsala University: Claudia Ehrentraut
Participants from KTH: Markus Näsman, Olof Jacobsson
Project duration: January  - December 2012, January 2013 - June 2014
Funding: Karolinska University Hospital and DSV,  January  - December 2012.
Funding: Vinnova Innovation mot Infektion, October 2012 - June 2014.

The goals of the project

  • To construct a system that automatically can detect if a hospital aquired infection (HAI) has occurred by analyzing clinical free text written in Swedish in a patient record, specifically in an epicrisis.
  • To detect HAI with a precision with over 90 percent and a recall (sensitivity) on around 50 percent.

  • Long term goal is to monitor hospital aquired infections and consquently diminish adverse events in hospital care.

Electronic patient records and language technology

Digital or electronic patient records is a vast source for reuse since a large part of the health care process is documented in free text. As language technology has been developed and matured in recent years one can use automated tools to structure and extract important information. This can be used to improve the health care process in general, diminish adverse events and specifically hospital acquired infections.

Published papers

Ehrentraut, C., M. Ekholm, H. Tanushi, J. Tiedemann and H. Dalianis. 2016. Detecting hospital-acquired infections: A document classification approach using support vector machines and gradient tree boosting, Health Informatics Journal, DOI: 10.1177/146045821665647, pp. 1–19, pdf

Jacobson, O. and H. Dalianis. 2016. Applying deep learning on electronic health records in Swedish to predict healthcare-associated infections. Proceedings of the 15th Workshop on Biomedical Natural Language Processing, BioNLP 2016, pages 191–195, Berlin, Germany, August 12, 2016, pdf.

Ehrentraut, C., Kvist, M., Sparrelid, M. and Dalianis, H. 2014. Detecting Healthcare-Associated Infections in Electronic Health Records - Evaluation of Machine Learning and Preprocessing Techniques, in the Proceedings of the 6th International Symposium on Semantic Mining in Biomedicine (SMBM 2014). Bodenreider, O., Oliveira, J.L., Rinaldi, F. (Eds.), Aveiro, Portugal, pdf.

Tanushi, H., Kvist , M. and Sparrelid, E. 2014. Detection of Healthcare-Associated Urinary Tract Infection in Swedish Electronic Health Records. Studies in Health Technology and Informatics, vol 207, 330-339, pdf.

Ehrentraut, C, H. Tanushi, H. Dalianis and J. Tiedemann. 2012. Detection of Hospital Acquired Infections in sparse and noisy Swedish patient records. A machine learning approach using Naïve Bayes, Support Vector Machines and C4.5. In the proceedings of the Sixth Workshop on Analytics for Noisy Unstructured Text Data, AND, December 9, 2012 held in conjunction with Coling 2012, Bombay, pdf.