University of Twente Student Theses

Login

Generating natural language triage advice at Dutch out-of-hours primary care centers

Koehorst, G.S. (2023) Generating natural language triage advice at Dutch out-of-hours primary care centers.

[img] PDF
2MB
Abstract:When a patient calls an out-of-hours primary care center, a triagist goes through different steps to assess the urgency of the situation. These steps are entered into a software system. This software is provided by Topicus under the name VIPLive Spoed EPD, for a large part of the Dutch out-of-hours primary care centers. Triagists are faced with high workloads at these centers. As such, there is value in reducing their workload. Triagists provide an advice which is recorded as natural language. In this study, we aimed to generate this advice based on historical data. The generated advice could be presented as a suggestion, which could reduce the typing workload of triagists. Given their great success in various natural language generation (NLG) tasks, this study aimed to apply pre-trained language models to tackle this problem. Different models were finetuned and their performance was compared. To assess the usefulness of the proposed Rob2Rob model in practice, a user study was conducted which involved 4 triagists and 3 general practitioners. This study showed that the Rob2Rob is capable of generating appropriate advice in the majority of cases. However, domain experts expressed their doubts about the usefulness of the model outputs in practice.
Item Type:Essay (Master)
Clients:
Topicus, Deventer, Netherlands
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:54 computer science
Programme:Computer Science MSc (60300)
Link to this item:https://purl.utwente.nl/essays/97782
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page