Effect of computer order entry on prevention of serious medication errors in hospitalized children.

OBJECTIVE: Although initial research suggests that computerized physician order entry reduces pediatric medication errors, no comprehensive error surveillance studies have evaluated the effect of computerized physician order entry on children. Our objective was to evaluate comprehensively the effect of computerized physician order entry on the rate of inpatient pediatric medication errors.
METHODS: Using interrupted time-series regression analysis, we reviewed all charts, orders, and incident reports for 40 admissions per month to the NICU, PICU, and inpatient pediatric wards for 7 months before and 9 months after implementation of commercial computerized physician order entry in a general hospital. Nurse data extractors, who were unaware of study objectives, used an established error surveillance method to detect possible errors. Two physicians who were unaware of when the possible error occurred rated each possible error.
RESULTS: In 627 pediatric admissions, with 12,672 medication orders written over 3234 patient-days, 156 medication errors were detected, including 70 nonintercepted serious medication errors (22/1000 patient-days). Twenty-three errors resulted in patient injury (7/1000 patient-days). In time-series analysis, there was a 7% decrease in level of the rates of nonintercepted serious medication errors. There was no change in the rate of injuries as a result of error after computerized physician order entry implementation.
CONCLUSIONS: The rate of nonintercepted serious medication errors in this pediatric population was reduced by 7% after the introduction of a commercial computerized physician order entry system, much less than previously reported for adults, and there was no change in the rate of injuries as a result of error. Several human-machine interface problems, particularly surrounding selection and dosing of pediatric medications, were identified. Additional refinements could lead to greater effects on error rates.