Early detection of adverse drug events within population‐based health networks: application of sequential testing methods

Abstract

Active surveillance of population-based health networks may improve the timeliness of detection of adverse drug events (ADEs). Active monitoring requires sequential analysis methods. Our objectives were to (1) evaluate the utility of automated healthcare claims data for near real-time drug adverse event surveillance and (2) identify key methodological issues related to the use of healthcare claims data for real-time drug safety surveillance. We assessed the ability to detect ADEs using historical data from nine health plans involved in the HMO Research Network's Center for Education and Research on Therapeutics (CERT). Analyses were performed using a maximized sequential probability ratio test (maxSPRT). Five drug-event pairs representing known associations with an ADE and two pairs representing 'negative controls' were analyzed. Statistically significant (p < 0.05) signals of excess risk were found in four of the five drug-event pairs representing known associations; no signals were found for the negative controls. Signals were detected between 13 and 39 months after the start of surveillance. There was substantial variation in the number of exposed and expected events at signal detection. Prospective, periodic evaluation of routinely collected data can provide population-based estimates of medication-related adverse event rates to support routine, timely post-marketing surveillance for selected ADEs.