Chart review versus an automated bioinformatic approach to assess real-world crizotinib effectiveness in ALK-positive NSCLC.

Abstract

Purpose The efficacy of targeted therapies such as crizotinib in anaplastic lymphoma kinase–positive non–small-cell lung cancer has been well established by multiple clinical trials. However, clinical trials involve highly selected participants, and manual data curation is resource intensive. With the increasing use of electronic medical records, there is potential for the development of electronic algorithms that could quickly generate outcomes data, but the validation of such algorithms requires comparison with historical methods, such as retrospective chart review. Materials and Methods Using a cohort of patients with anaplastic lymphoma kinase–positive non–small-cell lung cancer, we performed manual chart review to retrospectively evaluate time on treatment (TOT) for crizotinib and cytotoxic chemotherapies. Thirty-three patients were identified, with a total of 70 regimens administered. We developed a computational algorithm to mine electronic charts for crizotinib therapy data and correlated that with manually curated data. Results Among the 24 patients who received crizotinib, the median TOT was 8.5 months. The computational algorithm was able to extract TOT data for 15 out of 24 patients treated with crizotinib (62.5%). Most of the patients for whom data could not be automatically extracted were treated within a clinical trial. Pearson’s correlation coefficient between the two methods was 0.39 ( P = .15); however, there were five outliers as a result of incorrect provider notation and nonstandard treatment patterns that severely skewed the correlation. Conclusion An automated method of extracting patient TOT is feasible but requires additional optimizations to handle outliers. We are currently working on improving the algorithm to better correct for outliers. Automated generation of treatment data potentially represents a viable approach to perform retrospective outcomes analysis.