Primary data, claims data, and linked data in observational research: the case of COPD in Germany

BACKGROUND: Real-world evidence (RWE) can inform patient management decisions, but RWE studies are associated with limitations. Linkage of different RWE data types could address such limitations by enriching data and improving scientific quality. Using the example of chronic obstructive pulmonary disease (COPD) in Germany, this study assessed the value of data linkage between primary and secondary data sources for RWE. METHODS: Post hoc analysis of data from an observational RWE study, which used prospectively collected data and data from an insurance claims database to assess treatment adherence and persistence in patients with COPD in Germany. Patient-level primary data were collected from the prospective observational study (primary dataset, N = 636), and claims data from the sickness fund AOK Nordost (claims dataset, N = 74,916). Primary and claims data were linked at a patient level using insurance numbers (linked dataset). Patients in the linked dataset were indexed at date of study inclusion for primary data and matched calendar date for claims data. Agreement between primary and claims data was examined for patients in the linked dataset based on comparisons between recorded sociodemographic data at index, comorbidities (primary: any recorded; claims: pre-index), prescriptions for COPD therapies (type and date) and exacerbations in the 12-month post-index period. RESULTS: The linked dataset included primary and claims data for 536 patients. Fewer comorbid patients were reported in primary data compared with claims data (p < 0.001), with overall agreement between 63.6% (hypertension) and 90.5% (osteoporosis). Number of prescriptions for COPD therapies per patient was lower in primary versus claims data (3.7 vs 10.3 prescriptions, respectively), with only 24.5% of prescriptions recorded in both datasets. Only 11.5% of exacerbations (moderate or severe) were recorded in both datasets, with 15.5% recorded only in primary data and 73.0% recorded only in claims data. CONCLUSION: Our study highlighted discrepancies between primary and claims data capture for this population of German patients with COPD, with lower reporting of comorbidities, COPD therapy prescriptions and exacerbations in primary versus claims data. Study findings suggest that data linkage of primary and claims data could provide enrichment and be useful in fully describing COPD endpoints. ELECTRONIC SUPPLEMENTARY MATERIAL: The online version of this article (10.1186/s12931-018-0865-1) contains supplementary material, which is available to authorized users.