Adverse drug reaction reporting in a multicenter surveillance study

To evaluate the performance of a multicenter, prospective surveillance program in identifying adverse events, and to seek explanations for misclassification bias.

The design was a prospective observational study of patients with documented or suspected bacterial pneumonia.

Data were collected in 74 acute care hospitals across the US.

This evaluation was based on a consecutive sample of 1822 adult patients (> 18 years of age) with documented or suspected bacterial pneumonia who were being treated with a cephalosporin, a penicillin, or an aminoglycoside over a 3-month period. Patients were followed for the duration of antibiotic therapy and were excluded if antibiotic therapy was < 3 days or if the pneumonia was judged to be nonbacterial.

Clinical pharmacists documented patient demographics, concurrent illnesses and medications, antibiotic administration, relevant laboratory data, and the occurrence of nephrotoxicity and neutropenia.

Validity of investigators' identification of neutropenia and nephrotoxicity as compared with objective laboratory data was assessed by using sensitivity, specificity, and positive and negative predictive value measures.

Among the 1502 patients with sufficient data to evaluate neutropenia, there was agreement in 1270 patients (84.6%); likewise, among 1291 patients with sufficient data to evaluate nephrotoxicity there was agreement in 1186 patients (91.9%). Sensitivity of the researchers' assessments was 50.9% and 71.0% for neutropenia and nephrotoxicity, respectively. The negative predictive value was > 95% for both events.

Overall, this evaluation demonstrated that the Drug Surveillance Network can successfully identify targeted adverse events. Moreover, this study highlights the importance of validation for all types of outcomes-oriented research studies.