FDA adverse event reports are catalogued in the Manufacturer and User Device Experience (MAUDE) database. We developed and internally validated an instrument for analyzing MAUDE data, reviewing all reports in which a fatality occurred. Two trained reviewers independently analyzed each report, and a third resolved discrepancies or passed them to a committee for resolution.
One thousand two hundred eighty-four adverse events were reported between June 1993 and October 2008, of which 1,150 were failed defibrillation attempts. Thirty-seven automated external defibrillators never powered on, 252 failed to complete rhythm analysis, and 524 failed to deliver a recommended shock. In 149 cases, the operator disagreed with the device's rhythm analysis. In 54 cases, the defibrillator stated the batteries were low and in 110 other instances powered off unexpectedly. Interrater agreement between reviewers 1 and 2 ranged by question from 69.0 % to 98.6 % and for most likely cause was 55.9 % . Agreement was obtained for 93.7 % to 99.6 % of questions by the third reviewer. Remaining discrepancies were resolved by the arbitration committee.
MAUDE information is often incomplete and frequently no corroborating data are available. Some conditions not detected by automated external defibrillators during self-test cause units to power off unexpectedly, causing defibrillation delays. Backup units frequently provide shocks to patients.