Development and validation of an algorithm to identify drug-induced anaphylaxis in the Beijing Pharmacovigilance Database.

Background Pharmacovigilance databases are utilized to identify serious adverse drug events (ADEs). In China, very few studies have evaluated the validity of using pharmacovigilance databases to identify drug-induced anaphylaxis (DIA). Objective We aimed to develop and validate an algorithm to identify DIA using the Beijing Pharmacovigilance Database (BPD). Setting ADEs from the BPD mainly spontaneously reported from 94 hospitals in Beijing, China. Method Using the diagnoses, we developed an algorithm to identify potential DIAs from the BPD between January 2004 and December 2014. A sample of 500 patients was randomly selected for chart abstraction. Two physician adjudicators assessed whether DIA occurred using the published clinical criteria as the gold standard. Main outcome measure Positive predictive values (PPVs) and 95% confidence intervals of the algorithm and algorithm criteria components were calculated. Results 500 patients (53.2% female; the mean age 48.2 years) with potential DIA were selected using the algorithm. 444 were adjudicated as having anaphylaxis by physicians. The PPV of the overall algorithm was 88.8% (95% CI 86.0-91.6%). PPV for the algorithm only using specific diagnoses of "anaphylactic shock", "anaphylactic reaction", and "anaphylactoid reaction [severe]" was 89.6% (95% CI 86.6-92.4%); this partial algorithm identified 387 (87.2%) DIAs. The diagnosis that identified the most DIAs (83.8%) was "anaphylactic shock", with a PPV of 91.6% (95% CI 88.9-94.3%). Conclusion The overall algorithm identified a greater number of DIAs than the algorithm that only used specific diagnoses; however, its PPV was slightly lower. We were able to identify DIAs with the algorithm we developed.