Quantitative bias analysis for outcome phenotype error correction in comparative effect estimation: an empirical and synthetic evaluation


This interactive web application reports results from an empirical and synthetic evaluation of quantitative bias analysis for outcome phenotype error correction.


Outcome phenotype measurement error is rarely corrected in comparative effect estimation studies in observational pharmacoepidemiology. Quantitative bias analysis (QBA) is an outcome misclassification correction method that algebraically adjusts person counts in exposure-outcome contingency tables based on the magnitude of misclassification. The extent that QBA minimizes bias is unclear because few systematic evaluations have been reported. We empirically evaluated QBA impact on odds ratios (OR) in several comparative effect estimation scenarios. We estimated non-differential and differential phenotype errors with internal validation studies using a probabilistic reference standard. Further, we created a synthetic analytic space defined by outcome incidence, uncorrected ORs, and phenotype errors to identify which input combinations would produce invalid results indicative of input errors. We evaluated impact with relative bias [(OR-ORQBA)]/ORĂ—100%]. Results were considered invalid if any contingency table cell was corrected to an implausible negative number. Empirical bias correction was greatest in lower incidence scenarios where uncorrected estimates were larger. Synthetic bias correction increased as uncorrected estimates increased and incidence decreased. The invalid proportion of synthetic scenarios increased as uncorrected estimates increased at incidence levels. QBA produced invalid results in common, low incidence pharmacoepidemiology scenarios indicating problematic phenotype error input. This demonstrates the importance of estimating outcome phenotype errors to assess QBA feasibility before implementing comparative effect estimation studies.

Below are links for study-related artifacts that have been made available as part of this study:

Table 3. Fitted propensity model
Figure 2. Preference score distribution