OP4-4 - Pragmatic SIMEX-type method to correct for measurement error in time-varying prescription-based exposures

Prescription claims have been increasingly used in drug safety and effectiveness studies to determine individual drug exposures. However, due to non-adherence to the prescribed treatment, drug exposures based on prescriptions do not always accurately represent the actual drug intake. Such discrepancies induce measurement error in the assessment of the true drug exposure, which is known to result in biased inference in naive analyses.

We developped a method specifically tailored to correct for measurement error in the analyses of time-varying prescription-based exposures. The proposed approach relies on a pragmatic adaptation of the simulation-extrapolation (SIMEX) method, a well-known bias correcting method for exposure measurement error. In contrast with SIMEX, our proposed method does not require any assumptions on the measurement error process, which facilitates its implementation in a wide variety of applications.

We conducted a simulation study to evaluate to what extent the proposed method reduces bias. Observed prescription histories were resampled from an empirical cohort and true drug intake simulated under various plausible assumptions about non-adherence patterns. Preliminary results indicate that the proposed method performs well in simulation, by reducing the bias due to measurement error, and this for various exposure metrics that might be of interest in different pharmacoepidemiologic applications.

In conclusion, accounting for measurement error in time-varying prescription-based exposures is challenging, but crucial to avoid biased analyses. Pragmatic methods, such as the one proposed, are a promising avenue to more appropriately account for complex types of measurement error.