Using computable knowledge mined from the literature to elucidate confounders for EHR-based pharmacovigilance
Description
INTRODUCTION: Confounding bias threatens the reliability of observational data and identifying confounders poses a significant scientific challenge. We hypothesize that adjustment sets of literature-derived confounders could also improve causal inference. This paper shows how to exploit literature-derived knowledge to identify confounders for causal inference from observational data. We show how semantic constraint search over literature-derived computable knowledge helps reduce confounding bias in statistical models of EHR-derived observational data. METHODS: We introduce two methods (semantic vectors and string-based confounder search) that query the literature for potential confounders and use this information to build models from EHR-derived data to more accurately estimate causal effects. These methods search SemMedDB for indications TREATED BY the drug that is also known to CAUSE the adverse event. For evaluation, we attempt to rediscover associations in a publicly available reference dataset containing expected pairwise relationships between drugs and adverse events from empirical data derived from a corpus of 2.2M EHR-derived clinical notes. For our knowledge-base, we use SemMedDB, a database of computable knowledge mined from the biomedical literature. Using standard adjustment and causal inference procedures on dichotomous drug exposures, confounders, and adverse event outcomes, varying numbers of literature-derived confounders are combined with EHR data to predict and estimate causal effects in light of the literature-derived confounders. We then compare results from the adjustment and inference procedures with naive ($\chi^2$, reporting odds ratio) measures of association. RESULTS AND CONCLUSIONS: Logistic regression with ten vector space-derived confounders achieved the most improvement with AUROC of 0.628 (95\% CI: [0.556,0.720]), compared with baseline $\chi^2$ 0.507 (95\% CI: [0.431,0.583]). Bias reduction was improved more often in modeling methods using more rather than less information, and using semantic vector rather than string-based search. We found computable knowledge useful for improving automated causal inference, and identified opportunities for further improvement, including a role for adjudicating literature-derived confounders by subject matter experts.