OBJECTIVES: Evaluating the patient journey necessitates multiple data sources including patient reported outcomes (PRO) and administrative claims data. These data are best understood when collected, deterministically linked, and analyzed for a patient cohort. Challenges in that process such as time, cost, and patient availability sometimes require researchers to rely on the underlying assumption that cohort data are generalizable to the populations they represent. The purpose of this study was to compare characteristics of a claims-based patient cohort from the Optum Research Database (ORD) identified via protocol requirements from a previously conducted claims-linked PRO study to the characteristics of patients in that PRO study, thus validating the generalizability of the broader claims-based cohort.

METHODS: Inclusion/exclusion criteria were used from previously conducted claims-linked PRO study of commercial patients with asthma to identify the larger sample. A total of 29,094 patients were identified (versus 428 in the PRO study). Descriptive analyses, independent t-tests, and Chi-square tests were used for comparison of results.

RESULTS: Results indicate that participants in the PRO study were older (Mean 49.8/47.1, p<.01) and more females comprised the PRO sample (67%/56%, p<.01). Charleson Comorbidity Index was consistent across both groups (Mean=1.2). Importantly, for the main outcome variable of asthma exacerbations, there was no difference between the two groups in the baseline (p=.5) or follow-up period (p=.8).

CONCLUSIONS: Although small demographic differences in study participants between a claims-linked PRO sample and larger claims-based sample were found, there were no differences in the study outcomes. These differences reflect those most likely to complete surveys and can be adjusted for in future analyses. Although, ideally, claims data should be deterministically linked for optimal results, this study demonstrates that a larger claims-based sample, from a broad administrative claims database such as ORD, can be a generalizable source of outcomes data when resources are limited.