Further characterization of the influence of crowding on medication errors

Our prior analysis suggested that error frequency increases disproportionately with Emergency department (ED) crowding. To further characterize, we measured this association while controlling for the number of charts reviewed and the presence of ambulance diversion status. We hypothesized that errors would occur significantly more frequently as crowding increased, even after controlling for higher patient volumes.

We performed a prospective, observational study in a large, community hospital ED from May to October of 2009. Our ED has full-time pharmacists who review orders of patients to help identify errors prior to their causing harm. Research volunteers shadowed our ED pharmacists over discrete 4- hour time periods during their reviews of orders on patients in the ED. The total numbers of charts reviewed and errors identified were documented along with details for each error type, severity, and category. We then measured the correlation between error rate (number of errors divided by total number of charts reviewed) and ED occupancy rate while controlling for diversion status during the observational period. We estimated a sample size requirement of at least 45 errors identified to allow detection of an effect size of 0.6 based on our historical data.

During 324 hours of surveillance, 1171 charts were reviewed and 87 errors were identified. Median error rate per 4-hour block was 5.8% of charts reviewed (IQR 0-13). No significant change was seen with ED occupancy rate (Spearman's rho = -.08, P = .49). Median error rate during times on ambulance diversion was almost twice as large (11%, IQR 0-17), but this rate did not reach statistical significance in univariate or multivariate analysis.

Error frequency appears to remain relatively constant across the range of crowding in our ED when controlling for patient volume via the quantity of orders reviewed. Error quantity therefore increases with crowding, but not at a rate greater than the expected baseline error rate that occurs in uncrowded conditions. These findings suggest that crowding will increase error quantity in a linear fashion.