SOPalign: A Tool for Automatic Estimation of Compliance with Medical Guidelines

Luke van Leijenhorst, Arjen P. de Vries, Thera Habben Jansen, Heiman Wertheim

Abstract: SOPalign is a tool designed for hospitals and other healthcare providers in the Netherlands to automatically estimate the compliance of internal standard operating procedures (SOPs) for employees with the national guidelines. In this tool, users can upload the SOPs of their hospital and the recommendations from the most recent guidelines. SOPalign will then link the individual recommendations from the guidelines to the relevant passages of text in the SOPs and determine whether these passages are compliant with the recommendations. To link the SOP passages to the recommendations from the guideline, we make use of a Semantic Textual Similarity (STS) model based on the siamese BERT-network architecture. For efficiency reasons, we only apply the STS model to sentences that exceed a threshold in n-gram cosine similarity. To estimate compliance of SOPs with guideline recommendations, we have fine-tuned pre-trained language models using two different Dutch Natural Language Inference (NLI) datasets.