OBJECTIVES: A concern in database studies is the index date definition, which impacts how survival and exposure outcomes are measured/reported. Most studies use first medical code as standard. This analysis compared the validity of this approach against a multi-coding approach in two real-world databases of patients with ovarian cancer (OC).

METHODS: OCEANIA, a retrospective cohort study, assessed data from patients with OC from a Brazilian health claims (2015–2020) database and Argentinian hospital (2010–2019) database. In the standard approach, the index date (proxy for diagnosis) was defined as either the first date of an OC-related medical code (Argentina) or an ICD-10 code assignment (Brazil). In the multi-coding approach, the index date was determined by reviewing patient records to identify the first event/procedure that could retrospectively be considered OC-related ≤180 days prior to the first OC code. Mean, median and standard deviation (SD) of patient follow-up durations were calculated.

RESULTS: In the Brazilian database, 2529 patients with OC were identified. Mean follow-up duration from the standard index date to last database input was 542.9 (SD=472.3; median 394.0) days. Using the multi-coding approach, mean follow-up changed to 623.5 (SD=476.5; median 479.0) days. Among 2126 patients that had their index updated, mean follow-up increased by 95.8 (SD=53.9; median 97.0) days. In the Argentinian database, 741 patients with OC were identified. Mean follow-up duration was 938.7 (SD=925.7; median 664) days with the standard approach. The multi-coding approach increased the mean follow-up to 965.2 (SD=932.3; median 709.0) days. Among 347 patients that had their index updated, mean follow-up period increased by 76.0 (SD=176.8; median 31) days.

CONCLUSIONS: The choice of index date for database studies requires careful consideration to ensure accurate interpretation of time-dependent outcomes, which may help minimise bias and improve data validity.