OBJECTIVES: Although randomized clinical trials (RCT) remain the gold standard for safety and efficacy evidence in healthcare research, real-world data (RWD) is increasingly used to capture larger and more representative patient populations, in a cost and time effective manner. However, the applicability of RWD to generate high-quality evidence should be carefully evaluated, especially in the context of product approval or reimbursement. Early identification of RWD sources is imperative to determine their suitability for evidence generation to support product strategy. We propose a methodological framework for conducting data source landscaping to identify fit-for-purpose real-world secondary data sources.

METHODS: Based on previous industry experience and detailed understanding of RWD source requirements for secondary database research, we used an inductive approach to develop a methodological framework to systematically identify, characterize, and select appropriate data sources.

RESULTS:

A five-step framework was developed to select data sources. The framework needs to be adapted for different research questions:

1. Data source identification: conduct a pragmatic literature search targeting specific populations, diseases, and retrospective study designs.

2. Data source extraction and categorization: extract data sources from included publications and categorize them into different data source types to assess suitability for potential study objectives.

3. Data source screening: de-prioritize data sources for technical or operational considerations.

4. Data source evaluation: assess each data source based on an evaluation matrix, including availability of key variables and accessibility of the data sources.

5. Data source selection: select data sources with high scores from the evaluation matrix for future feasibility assessment.

This framework was successfully applied to multiple landscaping exercises in a range of therapeutic areas to support real-world evidence generation.

CONCLUSIONS: This framework covers the essential steps in a data source landscaping. Following this exercise, a deep-dive feasibility assessment is recommended to determine completeness and quality of the data source.