Corinne Hohl

Preliminary evidence suggests that 19-54% of patients diagnosed with adverse drug events (ADEs), unintended harm from medications, will be re-exposed to the culprit medication upon hospital discharge. The studies from which these estimates arise are small, descriptive, or rely solely on administrative data. Previous research reports an estimated 6-28% of adverse drug events can be identified using administrative data; these data are not well poised to examine adverse events. By triangulating multiple data sources (prospective, chart review, administrative claims data), we sought to: (1) Examine the proportion of re-exposure to culprit medication upon hospital or emergency department discharge, and identify any risk factors for culprit medication re-exposure using Cox regression models; (2) Investigate methods to identify medication non-adherence using claims data, and examine repeat medication non-adherence descriptively; and (3) Examine how administrative health data perform in identifying adverse drug events, including medication non-adherence, by calculating sensitivity and specificity, and conducting logistic regression. The results of our analyses indicated that re-exposure to culprit medication occurred for 45.2% of adverse drug events, though this was largely driven by necessary, irreplaceable medications. Re-exposure varied by type of ADE and duration of medication use. Analyses indicated that no method to identify medication non-adherence in administrative datasets performed well, but the proportion of days covered was the best measure. Finally, as expected, administrative claims data performed poorly in identifying adverse drug events. The administrative data source from which events were more readily detected was hospital data. Overall, the results of this dissertation suggest that while a large proportion of adverse drug events result in subsequent medication re-exposure, these may not be inappropriate re-exposures.ivFurther research is needed to draw this distinction. These results also demonstrate that adverse drug events, including medication non-adherence, are poorly identified in administrative claims data, and caution should be used when interpreting research that relies solely on their use. Previous work in the field has likely underestimated the burden of adverse drug events. Where possible, research on adverse drug events should include data that are prospectively collected at the point of care.

View record