EMDAI: A machine learning based decision support tool for Emergency Medical Dispatch

Important results from the project

The project has met the goals of developing and validating machine learning models for use at Sjukvårdens Larmcentral, to develop an open source software package to implement the developed models, and to integrate them into a user interface in the dispatch system. In order to correct data quality problems that emerged during the project´s preparation phase, an update was implemented in May 2019, which had a good effect on data quality. A randomized trial of the methods is now undergoing ethics board review.

Expected long term effects

In the project we have developed risk assessment models, published a validation of these in a scientific journal, evaluated the assessment process in another article, and made the tools that we use to implement them available under an open-source license. Efforts to implement the tools in the clinic were delayed due to data quality problems, but are now proceeding.

Approach and implementation

The execution project changed as Uppsala Clinical Recearch Center´s role in the project was instead carried out by the Uppsala Ambulance Service. The tool´s implementation in clinical use was delayed as an update in decision support system to address data quality issues was introduced, which entailed that additional data had to be collected. The qualitative studies on the work process also led to the intervention being reworked to better meet the needs of the business. The remaining activities in the project were carried out according to plan.

External links

openTriage: A platform for deploying machine learning models in clinical decision support systems Using trigger tools to identify triage errors by ambulance dispatch nurses in Sweden: an observational study Projektets hemsida vid Uppsala Center for Prehospital Research A validation of machine learning-based risk scores in the prehospital setting