P0021 - Integrating and Evaluating the Usability and Utility of Smart Pump Information in Detecting Medication Administration Errors

Background: As the concept of the learning health system (LHS) has come to fruition, clinical informatics has been emphasized as a necessary component. At present, electronic health records (EHRs) are the central focus of many of these efforts given their role as the primary source of data, but many other sources of data exist, including other digital devices commonly used in healthcare such as smart infusion pumps. When combined with prescribing data from the EHR, smart pump record (SPR) data is capable of shedding light on the actions taking place in the medication use lifecycle from prescription to administration.

Objective: To merge and chronologically align EHR data with SPR data to identify discrepancies in medication administration and to develop novel algorithms to characterize the magnitude of the identified discrepancies and the concordance between the ability of the medication administration record (MAR) and SPR data to detect these discrepancies.

Methods: We evaluated medication administrations for nine high-risk continuous intravenous infusion medications administered to neonatal patients admitted to the NICU. An algorithm was developed to merge the EHR and SPR data and link the SPRs to their potential medication orders by patient ID and medication information. The aligned data was reviewed by a clinical research coordinator and two pediatric physicians manually to identify medication administration discrepancies. Measures of the magnitude of discrepancy and concordance of the detection of discrepancies between MAR and SPR data were also calculated.

Results: Seventy-one percent (71%) of smart pump records (SPRs) in 2014 and 66% in 2016 contained both valid patient-IDs and medication names making them usable for data federation. The investigative team reviewed 2307 medication orders with 10575 MARs and 233975 SPRs. A total of 321 MAR and 682 SPR discrepancies were identified with Epinephrine, Dopamine, and Vasopressin having the highest discrepancy rates. Approximately 58%/67% of identified MAR/SPR discrepancies were overdose and 13%/18% were substantial overdose – a dose 100% greater than the prescribed dose. Lastly, SPR data was found to have more discrepancies than MAR data with only 21.7% discrepancy detection concordance.

Conclusions: The majority of discrepancies were identified through SPRs indicating that SPRs may be a more reliable data source for discrepancy or error detection when compared to MARs. By detecting discrepancies at an increased rate than MARs, the use of SPRs becomes relevant to increasing clinical safety. Ultimately, it is crucial that MAR and SPR data be integrated to effectively detect, identify, and mitigate medication errors occurring in the clinical setting. Further studies should investigate meaningful use of MAR and SPR data to inform effective error prevention strategies.