(151) Enhancing Evidence Generation by Linking Randomized Clinical Trials to Real-World Data: A Validation Study from the INVESTED-Medicare Linkage

Background: The potential of integrating real-world evidence (RWE) with randomized controlled trial (RCT) findings to enhance regulatory, clinical, and other decision-making is increasingly recognized. However, the capacity of real-world data (RWD) to enhance the collection of patient characteristics and long-term outcomes in RCTs remains largely unexplored.

Objectives: We validated variables measured in claims against trial-reported variables as the gold standard by linking individual participant data from the Influenza Vaccine to Effectively Stop Cardio Thoracic Events and Decompensated Heart Failure (INVESTED) trial with Medicare claims data.

Methods: Following the conclusion of the INVESTED trial in 2019, we obtained consent for data linkage from US participants between 11/2021 and 11/2022. We then linked participants, who consented and provided their Medicare Beneficiary Identifiers (MBIs), to Medicare data. We compared baseline characteristics and outcomes measured in claims with the same variables captured through primary data collection for the trial.

Results: Out of 126 participants who consented and provided complete MBIs, 60 were successfully matched to Medicare data. Compared with the broader US trial participants, these matched participants were more likely to be older (standardized difference [std.diff]=0.70) and White (std.diff=0.55) and tended to have either no or mild symptoms according to the New York Heart Association functional classification. Agreement between claims-measured and trial-reported demographics was high, with sex matching 100% and only one discrepancy in age. Racial classification (White) had a sensitivity of 0.82 and specificity of 1. Outcomes generally showed low sensitivity but high specificity for cardiopulmonary hospitalizations (sensitivity=0.19, 95% CI=0.05–0.42; specificity=0.92, 0.79–0.98) and heart failure hospitalizations (sensitivity=0.50, 0.07–0.93; specificity=0.96, 0.88–1.00).

Conclusions: While claims data may not fully substitute for trial data in collecting baseline characteristics and outcomes, the observed accuracy in claims versus RCT measurements can inform quantitative bias analyses for long-term outcomes captured only in claims. The high specificity of the investigated outcomes, as captured in claims, suggests that relative risk estimates in RWE studies may not be significantly biased. If bias occurs, it is likely to be towards the null. However, the limited sample size of this study necessitates further research to substantiate these results.