Using an anesthesia information management system to prove a deficit in voluntary reporting of adverse events in a quality assurance program

Objective. A deficit is suspected in the manual documentation of adverse events in quality assurance programs in anesthesiology. In order to verify and quantify this, we retrospectively compared the incidence of manually recorded perioperative adverse events with automatically detected events. Methods. In 1998, data of all anesthetic procedures, including the data set for quality assurance of the German Society of Anaesthesiology and Intensive Care Medicine (DGAI), was recorded online with the Anesthesia Information Management System (AIMS) NarkoData4 (R) (Imeso GmbH). SQL (Structured Query Language) queries based on medical data were defined for the automatic detection of common adverse events. The definition of the SQL statements had to be in accordance with the definition of the DGAI for perioperative adverse events: A potentially harmful change of parameters led to therapeutic interventions by an anesthesiologist. Results. During 16,019 surgical procedures, anesthesiologists recorded 911 (5.7%) adverse events manually, whereas 2966 (18.7%) events from the same database were detected automatically. With the exception of hypoxemia, the incidence of automatically detected events was considerably higher than that of manually recorded events. Fourteen and a half percent (435) of all automatically detected events were recorded manually. Conclusion. Using automatic detection, we were able to prove a considerable deficit in the documentation of adverse events according to the guidelines of the German quality assurance program in anesthesiology. Based on the data from manual recording, the results of the quality assurance of our department match those of other comparable German departments. Thus, we are of the opinion that manual incident reporting seriously underestimates the true occurrence rate of incidents. This brings into question the validity of quality assurance comparisons based on manually recorded data.