2022 World Conference on Lung Cancer (ePosters)-EP01.06-004. Eliciting Quantitative Smoking History by the Use of Natural Language Processing

Pdf Summary

This document discusses the use of Natural Language Processing (NLP) to extract quantitative smoking history from electronic health records (EHRs) and improve the identification of patients eligible for low dose CT scanning (LDCT) for lung cancer screening. Currently, structured data from EHRs can only identify patients with a current or past smoking history, but not the specific details needed to determine LDCT eligibility. By applying NLP to analyze unstructured data from physician notes in the EHR, the authors were able to identify a larger number of current and former smokers, as well as gather quantitative pack year data. They found that NLP was superior to structured data analysis in determining smoking behaviors and eligibility for LDCT. The authors suggest that using NLP to analyze EHRs can help healthcare organizations identify eligible patients more efficiently, reducing the reliance on social marketing campaigns and individual patient interviews for smoking history. The study used the MIMIC-III database, a collection of critical care patient data, and employed the Ember Platform for NLP analysis. The results showed the potential of NLP to significantly increase the number of eligible patients for LDCT screening and improve patient management. The document highlights the importance of accurate smoking history for lung screening studies and reimbursement, as well as the challenges in obtaining this information from EHRs. The use of NLP and unstructured data analysis can enhance the identification of eligible patients and streamline the screening process. Overall, the study demonstrates the potential benefits of NLP in healthcare analytics and the future implications for improving patient care.

Asset Subtitle

John C. Ruckdeschel

Meta Tag

Speaker John C. Ruckdeschel

Topic Early Detection and Screening - Risk Stratification

Keywords

Natural Language Processing

quantitative smoking history

electronic health records

low dose CT scanning

lung cancer screening

physician notes

pack year data

structured data analysis

MIMIC-III database

patient management