Do oral antineoplastic treatment data differ between Concerto Health AI Definitive Oncology EMR and Symphony Claims Data?

Abstract

e18180 Background: Electronic Medical Records (EMRs) are increasingly being used for health outcomes research because they contain a rich array of clinical data. Oral medications are captured via structured medication order fields or abstracted from the clinical notes; however, it’s unknown how these correspond with dispensations. The objective of this study was to better understand how oral antineoplastic orders captured from an EMR database compare to dispensations from pharmacy claims. Methods: The Concerto Health AI Definitive Oncology Dataset is a deeply curated cohort of patients (pts) derived from a wide range of oncology practices throughout the United States. Concerto Health AI has a co-exclusive partnership with ASCO CancerLinQ & collects data from multiple other sources. These data were deterministically linked to US Symphony Health’s (SH) Integrated Dataverse, an anonymized, HIPAA-compliant data set. Only advanced (stage IIIb/IV) non-small cell lung cancer (aNSCLC) pts with curated data via manual review by nurse practitioners & oral antineoplastics prescribed for lung cancer were included in this analysis. Among a subset of pts that had a dispensation within 90 days of their 1st EMR order, treatment duration was calculated as time from 1st to last order/claim (before a 30-day gap) plus 30 days supply, for EMR & claims, respectively. Results: Of 19,175 aNSCLC pts from Concerto, 8,864 pts were linked to SH. Among linked pts, there were 2,057 pt-drug instances; 750 (36.5%) were in both Concerto and SH, 1,160 (56.4%) in Concerto but not SH, and 147 (7.1%) in SH but not Concerto. Median time from 1st order in Concerto to 1st dispensation in SH was 16 days (interquartile range (IQR): 6, 78), with 76% of pt-drug instances having a dispensation within 90 days of 1st Concerto order. Median treatment duration was 44 (IQR: 30, 125) and 93 (IQR: 41, 241) days based on Concerto and SH, respectively. Conclusions: Concerto EMR orders identified more pts treated with oral antineoplastics, though treatment durations were longer using SH claims. Limitations of this analysis include use of orders from EMR instead of abstracted start and end dates and use of open claims with limited specialty pharmacy data. Linked datasets may provide more robust capture of orals data and warrant further research.