This article was originally published here

Am J Prev Cardiol. December 3, 2021; 9: 100300. doi: 10.1016 / j.ajpc.2021.100300. eCollection 2022 March

ABSTRACT

OBJECTIVE: To determine whether natural language processing (NLP) of unstructured medical text can improve identification of patients with ASCVD not using high-intensity statin therapy (HIST) due to side effects associated with statins (SASE) and other reasons.

METHODS: The reviewers noted the reasons for not prescribing a HIST in notes from 1152 randomly selected patients across the VA health system treated for ASCVD but not receiving a HIST. The developers used reviewer annotations to train the Canary NLP tool to detect and extract notes that contained one or more of these reasons. Negative predictive value (NPV), sensitivity, specificity, and area under the curve (AUC) were used to assess the accuracy of detecting documents containing reasons when using structured, non-structured data. structured data extracted by PNL or from the two combined data sources.

RESULTS: At least one documented reason for not prescribing HIST occurred in 47% of notes. The most frequent reasons were SAEs (41%) and general intolerance (20%). When identifying notes containing a documented reason for not using HIST, the addition of unstructured data extracted by NLP in a meaningful way (p

CONCLUSIONS: NLP extraction of data from unstructured text may improve identification of reasons why patients are not on HIST compared to structured data alone. The additional information provided by NLP in unstructured free text form should help tailor and implement system-level interventions to improve the use of HISTs in patients with ASCVD.

PMID: 34950914 | PMC: PMC8671496 | DOI: 10.1016 / j.ajpc.2021.100300


Source link