ABSTRACT

The medication errors associated with intravenous (IV) administration may cause severe patient harm. Smart pumps with a built-in dose error reduction system (DERS) is one technique to help ensure the safety of IV administration in clinical settings. A drug limit alert, triggered by DERS, may be overridden, and could potentially cause patient harm, especially in high-risk medications. Some common analytical measures are frequency-based and only consider the overall drug performance rather than the severity impact from individual alerts. Unlike common measures, the IV medication harm index has been traditionally used to quantify risk of harm for individual alerts. Concerns about applying the IV harm index were addressed in our study. Therefore, the goal of this research is (1) to apply four analytical methods that quantify risk for the simulated individual alert-overridden infusions, (2) to evaluate the risk scores from different analytical methods, and (3) to propose better risk quantification methods with a higher correlation to risk benchmarks.

In this study, 20 pharmacists and 5 nurses were recruited to assess risk (adjusted for risk benchmarks) on the scenarios, created based on hospital alert data. In addition, the four analytical methods, applied to quantify risk for the scenarios, are the linear mixed models (LMMs) (method A), the IV harm index (method B), the matrix-based method (method C), and the analytical hierarchy process (AHP) method (method D). Method A used 7 alert factors (identified as key risk factors) to build models for risk prediction, and methods B and C used 2 out of 7 factors to obtain risk scores. Method D used pairwise comparison surveys to calculate the risk priorities. The quantified scores from the four methods were evaluated in comparison to the risk benchmarks. The results showed that, comparing the quantified scores from each method with the risk benchmarks separately, the risk scores from method A (ρ = 0.94) and method D (ρ > 0.70 in 26 out of 30 scenario types) were highly correlated to the risk benchmarks. When comparing all scenarios, the adjusted AHP scores, derived from method D combined with method A, have a positive correlation with the risk benchmarks (ρ > 0.87). However, the risk scores derived method B and method C did not have a positive correlation with the risk benchmarks.

This study demonstrates that the traditional IV harm index should include more risk factors, along with their interaction effects, for increased correlation with risk benchmarks. Furthermore, the LMMs with and without the AHP method, allow for better risk quantification methods where the quantified scores most correlated with the benchmarks. These methods can provide risk-based analytical support to evaluate alert-overridden infusions associated with the four high-risk medications in adult intensive care unit (AICU) and adult medical/surgical (AMS).