An Automated Line-of-Therapy Algorithm for Adults With Metastatic Non-Small Cell Lung Cancer: Validation Study Using...

Area of Science:

• Oncology
• Health Informatics
• Data Science

Background:

• Accurate extraction of line-of-therapy (LOT) information from electronic health records (EHR) and claims data is crucial for analyzing longitudinal changes in systemic anticancer therapy in real-world settings.
• Existing methods for LOT extraction require validation and refinement to improve accuracy and applicability.

Purpose of the Study:

• To validate and refine an open-source line-of-therapy (LOT) algorithm using a retrospective cohort analysis.
• To compare the algorithm's output against results from a blinded manual chart review for accuracy assessment.

Main Methods:

• A retrospective cohort of 500 adult patients with metastatic non-small cell lung cancer treated with systemic anticancer therapy (2011-2018) was analyzed.
• Patients were divided into training (n=350) and test (n=150) cohorts, stratified by simple (one LOT) and complex (multiple LOTs/maintenance therapy) cases.
• The LOT algorithm was refined using training data and subsequently evaluated on the test cohort.

Main Results:

• Algorithm refinement significantly reduced discordance in simple cases from 16 to 8 instances, with zero discordance in the test cohort.
• For complex cases, discordance decreased from 68 to 62 instances, with 37 instances in the test cohort.
• Percentage agreement for single LOT cases improved from 89% to 93% post-refinement and in the test cohort; however, precise matching decreased with more unique regimens.

Conclusions:

• The refined LOT algorithm demonstrates improved accuracy in extracting line-of-therapy information from EHR data.
• Identified sources of discordance highlight areas for further algorithm refinement and underscore the need for clearer definitions in medical literature regarding LOTs and maintenance therapy.