Facilitating post-surgical complication detection through sublanguage analysis

Identification of postsurgical complications is the first step towards improving patient safety and health care quality as well as reducing heath care cost. Existing NLP-based approaches for retrieving postsurgical complications are based on search strategies. Here, we conduct a sublanguage analysis study using free text reports available for a cohort of patients with postsurgical complications identified manually to compare the keywords identified by subject matter experts with words/phrases automatically identified by sublanguage analysis. The results suggest that search-based approaches may miss some cases and the sublanguage analysis results can be used as a base to develop an information extraction system or support search-based NLP approaches by augmenting search queries.