Presentation (CTI)

OBJECTIVES: Retrospective chart abstraction in real-world studies often requires sites to review a large number of records across multiple timepoints, creating substantial burden when the sampling frame greatly exceeds the number of patients needed. In these settings, selecting effective patient identification strategies becomes a methodological/operational challenge. Tailored approaches can reduce workload, minimize screening inefficiencies, and mitigate bias. The study objective was to describe a structured, reproducible framework for selecting sampling strategies in large-scale chart review studies, informed by recent real-world study experience.
METHODS: Three global gastroenterology/hepatology chart review studies were synthesized to develop a generalized sampling-strategy framework. Methodological and operational criteria for patient identification and sampling were extracted, compared, and summarized across studies.
RESULTS: Key framework elements and summarized steps for three studies (13 countries; ~2,900 patients) include: 1) Non-random versus random sampling: Non-random approaches were deprioritized due to wide variability in site populations (10-1000+ patients) and the need to minimize selection bias. 2) Sampling level: Two studies used site-level random sampling to meet per-site caps and balance country-level contributions; the third also applied disease-level sampling due to patients having multiple diseases. 3) Operational feasibility: Site feasibility assessments evaluated electronic medical record (EMR) functionality, informing the EMR-filtering workflows and screening log requirements. 4) Efficient identification of sampling frame: EMR-filtering workflows were tailored to each study utilizing electronically searchable eligibility criteria (e.g., diagnosis, treatments, age), to efficiently identify sampling frames. 5) EMR capability variance: Due to differences in EMR search functionality, screening logs were implemented to maintain traceability and support study-specific random-selection algorithms.
CONCLUSIONS: Sampling strategy selection in retrospective chart review studies should follow a structured framework that balances scientific validity with operational feasibility. By accounting for EMR variability, screening workload, and risk of selection bias, this approach supports method selection that reduces site burden, enhances workflow efficiency and supports timely, reliable evidence generation.