Background Smart infusion pumps affect workflows as they add alerts and alarms in an information-rich clinical environment where alarm fatigue is already a major concern. An analytic approach is needed to quantify the impact of these alerts and alarms on nursing workflows and patient safety. Objectives To analyze a detailed infusion dataset from a smart infusion pump system and identify contributing factors for infusion programming alerts, operational alarms, and alarm resolution times. Methods We analyzed detailed infusion pump data across four hospitals in a health system for up to 1 year. The prevalence of alerts and alarms was grouped by infusion type and a selected list of 32 high-alert medications (HAMs). Logistic regression was used to explore the relationship between a set of risk factors and the occurrence of alerts and alarms. We used nonparametric tests to explore the relationship between alarm resolution times and a subset of predictor variables. Results The study dataset included 745,641 unique infusions with a total of 3,231,300 infusion events. Overall, 28.7% of all unique infusions had at least one operational alarm, and 2.1% of all unique infusions had at least one programming alert. Alarms averaged two per infusion, whereas at least one alert happened in every 48 unique infusions. Eight percent of alarms took over 4 minutes to resolve. Intravenous fluid infusions had the highest rate of error-state occurrence. HAMs had 1.64 more odds for alerts than the rest of the infusions. On average, HAMs had a higher alert rate than maintenance fluids. Conclusion Infusion pump alerts and alarms impact clinical care, as alerts and alarms by design interrupt clinical workflow. Our study showcases how hospital system leadership teams can leverage infusion pump informatics to prioritize quality improvement and patient safety initiatives pertaining to infusion practices.

No abstract

No abstract

Objective To detect unusual infusion alerting patterns using machine learning (ML) algorithms as a first step to advance safer inpatient intravenous administration of high-alert medications. Materials and Methods We used one year of detailed propofol infusion data from a hospital. Interpretable and clinically relevant variables were feature engineered, and data points were aggregated per calendar day. A univariate (maximum times-limit) moving range (mr) control chart was used to simulate clinicians' common approach to identifying unusual infusion alerting patterns. Three different unsupervised multivariate ML-based anomaly detection algorithms (Local Outlier Factor, Isolation Forest, and k-Nearest Neighbors) were used for the same purpose. Results from the control chart and ML algorithms were compared. Results The propofol data had 3,300 infusion alerts, 92% of which were generated during the day shift and seven of which had a times-limit greater than 10. The mr-chart identified 15 alert pattern anomalies. Different thresholds were set to include the top 15 anomalies from each ML algorithm. A total of 31 unique ML anomalies were grouped and ranked by agreeability. All algorithms agreed on 10% of the anomalies, and at least two algorithms agreed on 36%. Each algorithm detected one specific anomaly that the mr-chart did not detect. The anomaly represented a day with 71 propofol alerts (half of which were overridden) generated at an average rate of 1.06 per infusion, whereas the moving alert rate for the week was 0.35 per infusion. Discussion These findings show that ML-based algorithms are more robust than control charts in detecting unusual alerting patterns. However, we recommend using a combination of algorithms, as multiple algorithms serve a benchmarking function and allow researchers to focus on data points with the highest algorithm agreeability. Conclusion Unsupervised ML algorithms can assist clinicians in identifying unusual alert patterns as a first step toward achieving safer infusion practices.