OBJECTIVES: Large nationwide databases are commonly used to evaluate healthcare outcomes based on retrospective analyses. However to our knowledge, no guidelines currently exist on how to proceed to the patient selection in these types of electronic health records (EHR). The objective of this study was to formalise the patient selection process for EHR databases.

METHODS: A generalizable patient selection process was drafted based on existing methods in the literature and experts opinion. Two different independent reviewers approved the process by testing it on different EHR databases using examples of population of interest.

RESULTS: A descriptive flowchart has been developed to guide users through the patient selection process. The first step involves the review and understanding of which types of codes are available in the database (for example diagnosis codes, prescription codes, or a combination of the two). When such codes are not available, specific algorithm can be developed based on available clinical guidelines, the published literature and experts opinion to identify the target population of interest. The algorithm may include minimum and maximum follow-up times, age at baseline, and comorbid conditions. The authors report on the critical validation step, based on the published literature and review by clinical experts. The use of this process results in a more robust approach to selecting patients.

CONCLUSIONS: This is the first standardised process to be developed to select patients in EHR databases. It should be considered for future research when selecting patients in real-world evidence studies.