Empirical validation of the reverse parametric waiting time distribution and standard methods to estimate prescription durations for warfarin

PurposeIn many prescription databases, the duration of treatment for the single prescription is not recorded. This study aimed to validate 2 different types of approaches for estimating prescription durations, using the oral anticoagulant warfarin as a case. MethodsThe approaches undergoing empirical validation covered assumptions of a fixed daily intake of either 0.5 or 1.0 defined daily dose (DDD), as well as estimates based on the reverse parametric waiting time distribution (rWTD), with different sets of covariates. We converted estimates of prescription duration to daily dose and compared them to prescribed daily dose as recorded in a clinical registry (using Bland-Altman plots). Methods were compared based on their average prediction error (logarithmic scale) and their limit of agreement ratio (ratio of mean error1.96 SD after transformation to original scale). ResultsEstimates of daily doses were underestimated by 19% or overestimated by 62% when assumptions of 0.5 or 1.0 DDD were applied. The limit of agreement ratio was 6.721 for both assumptions. The rWTD-based approaches performed better when using the estimated mean value of the inter-arrival density, yielding on average negligible bias (relative difference of 0 to 2%) and with limit of agreement ratios decreasing upon additional covariate adjustment (from 6.857 with no adjustment to 4.036 with the fully adjusted model). ConclusionsComparing the different methods, the rWTD algorithm performed best and led to unbiased estimates of prescribed doses and thus prescription durations and reduced misclassification on the individual level upon inclusion of covariates.