Hierarchical Models for Multiple, Rare Outcomes Using Massive Observational Healthcare Databases

Abstract

Clinical trials often lack power to identify rare adverse drug events (ADEs) and therefore cannot address the threat rare ADEs pose, motivating the need for new ADE detection techniques. Emerging national patient claims and electronic health record databases have inspired post-approval early detection methods like the Bayesian self-controlled case series (BSCCS) regression model. Existing BSCCS models do not account for multiple outcomes, where pathology may be shared across different ADEs. We integrate a pathology hierarchy into the BSCCS model by developing a novel informative hierarchical prior linking outcome-specific effects. Considering shared pathology drastically increases the dimensionality of the already massive models in this field. We develop an efficient method for coping with the dimensionality expansion by reducing the hierarchical model to a form amenable to existing tools. Through a synthetic study we demonstrate decreased bias in risk estimates for drugs when using conditions with different true risk and unequal prevalence. We also examine observational data from the MarketScan Lab Results dataset, exposing the bias that results from aggregating outcomes, as previously employed to estimate risk trends of warfarin and dabigatran for intracranial hemorrhage and gastrointestinal bleeding. We further investigate the limits of our approach by using extremely rare conditions. This research demonstrates that analyzing multiple outcomes simultaneously is feasible at scale and beneficial.

Figure 1

Mode estimates and 95% bootstrap confidence intervals (gray) of the log relative risk for each drug and their simulated relative risk (black) across two conditions with different prevalence. The first 10 covariates represent the estimates from one condition with a prevalence of 20 patients; the second 10 represent estimates from the condition with high prevalence, affecting 1000 patients; and the last 10 covariates represent a second condition with low prevalence, affecting 10 patients. Using the multiple outcomes in an aggregated approach (a) produces less appropriate estimates than the hierarchical outcomes approach (b).

Figure 2

Mode estimates and 95% bootstrap confidence intervals for the effect of dabigatran (light gray) and warfarin (dark gray) on gastrointestinal hemorrhage (GIH) and intracranial hemorrhage (ICH), compared to an aggregated outcome where GIH and ICH are exchangeable.

Figure 3

Mode estimates of the log relative risk for each drug for a common, rare, or aggregated outcome. The common outcome is vomiting blood (VB), dark gray triangles. The rare outcome is chronic gastrojejunal ulcer with hemorrhage and obstruction (CGJUHO), light gray circles. The aggregated outcome is CGJUHO or VB, black squares. The estimates for CGJUHO and VB rely on the hierarchical structure.