Hospital Medication Use Reporting from Electronic Medication Administration Records

Background: Electronic medication administration records (eMARs), on which nurses record medication administration events (MAEs) generated by electron medication orders, are probably more accurate than other medication use data sources but validated reports of eMAR data from commercial pharmacy information systems have been difficult to derive.

Objective: Develop a program-generated report based on hospital eMAR data and measure its accuracy.

Methods: The Cerner Command Language (CCL) program was developed in Discern Visual Developer to extract medication utilization data, allowing data filtering  based on parameters inputted (dates, hospital wards, specialty service, etc.) before program execution.  Data elements in the resulting reports also include patient identifiers, nursing unit, order specifics (i.e., medication name, dose, form, route, frequency), scheduled administration date and time, order result (given or not, reason for withholding if not given), and date and time of MAE charting in the eMAR.  The output was reported in spreadsheet form.  All report-generated MAEs for one patient from each of our hospital’s 29 patient-care units were selected from a report of MAEs for all inpatients over 14 days in June 2010 and validated via comparison with identical data elements abstracted from the electronic health records of those patients.

Results: Programming the report took 40 hours.  A total of 1,861 MAEs from 29 patients were validated.  Discrepancies between report-generated MAEs and those abstracted from patient records were identified for only 8 (0.4%) MAEs: MAEs recorded in patient records were missed by the program-generated report 5 times; and 3 MAEs not present in patient records were generated by the program-generated report.  The report correctly identified all 223 (12%) MAEs in which withholding of the scheduled medication dose was reported in patient eMARs.  eMARs in patient records reported the wrong medication dispensing location for 227 (12%) MAEs, all occurring after patient transfer from the location where the order was written; the correct location was included in the program-generated report, which extracted dispensing location from separate admission-transfer-discharge tables, in all instances.

Conclusions: Because our program-generated report of medication utilization from Cerner eMAR data was 99.6% accurate in all patient-care units, antimicrobial and other medication use rates based on these data should be similarly accurate.  Adaptation of our approach to the informatics infrastructures in other institution may facilitate reporting and benchmarking of valid eMAR-based antimicrobial use measures.