Early adverse drug event signal detection within population-based health networks using sequential methods: key methodologic considerations

Abstract

Purpose: Active surveillance of population-based health networks may improve the timeliness of detection of adverse events (AEs). Our objective was to expand our previous signal detection work by investigating the effect on signal detection of alternative study specifications.

Methods: We compared the signal detection performance under various study specifications using historical data from nine health plans involved in the HMO Research Network's Center for Education and Research on Therapeutics (CERT). Five drug-event pairs representing generally accepted associations with an AE and two pairs representing "negative controls" were analyzed. Alternative study specifications related to the definition of incident users and incident AEs were assessed and compared to our previous findings.

Results: Relaxing the incident AE exclusion criteria by (1) including members with prior outpatient diagnoses of interest and (2) halving (to 90 days) the time window specified to define incident exposure and diagnoses increased the number of members under surveillance and as a consequence increased the number of exposed days and diagnoses by about 10-20%. The alternative specifications tend to result in earlier signal detection by 10-16 months, a likely consequence of more exposures and events entering the analysis.

Conclusions: This paper provides additional preliminary information related to conducting prospective safety monitoring using health plan data and sequential analytic methods. Our findings support continued investigation of using health plan data and sequential analytic methods as a potentially important contribution to active drug safety surveillance.