Objective: To investigate the sensitivity and efficiency of medical record review for detecting adverse occurrences.
Design: A prospective observational study. Adverse occurrences were defined broadly as events that caused or had potential to cause adverse patient outcomes. Between August 1, 1990 and March 1, 1991 a research nurse reviewed written medical records and attended nursing morning report.
Setting: A university hospital.
Patients: Inpatients on two general medical wards and the medical and cardiovascular intensive care units.
Results: There were 317 occurrences identified (0.2 per patient day): 56% were medication related, 20% were new medical conditions, 11% were procedure complications, 5% were patient dissatisfaction related, 4% were equipment related, and 3% were accidents. Review of physician progress notes was the most sensitive means of detecting new conditions (sensitivity 73%), procedure complications (sensitivity 100%), equipment-related occurrences (sensitivity 71%), and adverse drug reactions (sensitivity 89%). Review of nursing notes was the most sensitive means of detecting patient dissatisfaction (sensitivity 69%) and accident-related occurrences (sensitivity 100%). Passive reporting through incident reports identified only 9% of all adverse occurrences. Surveillance was time-consuming: concurrent review of all written sources in a 250-bed hospital would require 5.5 technicians.
Conclusions: Surveillance for adverse occurrences is expensive. If selected adverse occurrences were identified as being important for quality improvement, surveillance could be focused efficiently on limited sources. The number of adverse occurrences discovered by surveillance will be heavily dependent on the choice of data source (physician progress notes, nursing notes, etc.). Comparisons among hospitals can be made only when systems for detection have similar validity.