Improving Lung Cancer Screening Rates Through an Evidence-Based Electronic Health Record Smoking History

Abstract

Background: Lung cancer is prevalent worldwide, with 2.1 million new cases and 1.8 million deaths in 2020. In the United States, an estimated 131 880 lung cancer deaths are expected to occur in 2021, with most detected in later stages. Smokers are 15 to 30 times more likely to develop or die from lung cancer. Local Problem: Our community residents were more likely to be diagnosed with lung cancer in later stages (62%) compared with 56% nationally, resulting in an increased community mortality rate. Intervention: Evidence-based changes in an electronic health record system supported identification and referral of high-risk patients for low-dose computer tomography to improve early lung cancer detection rates. Results: Early-stage lung cancer detection increased 24%. Conclusions: Interprofessional teams used technology to adopt evidence-based practice and improve health outcomes in their communities.

Enhancing an enterprise data warehouse for research with data extracted using natural language processing

“…We focus on the integration of the NLP to extract smoking behavior into the extract, transform, load (ETL) process in the local EDW4R at the University of Florida Health (UF Health). These data are not reliably collected in structured form, however, their availability as structured data allows computation of clinical and research-relevant information such as identifying patients who fit guidelines for lung cancer screening [28][29][30][31][32][33] and recruitment for cancer-related clinical trials [34][35][36][37]. We explain how this architecture can be generalized for future NLP models that extract different concepts.…”