Beyond Human Error: Building Intelligent Resilience for Medication Safety in the ICU

Comprehensive Summary

The intensive care unit (ICU) is a care setting vulnerable to medical errors in relation to medication safety. There exists a surveillance gap where many errors go unnoticed in the demanding workflow of the ICU. This review proposes the implementation of a five-layer Intelligent Safety Stack to address the surveillance gap by integrating artificial intelligence (AI) tools with human oversight. The first layer of the safety stack oversees the addition of standardized error classifications into existing electronic health record systems to allow for machine learning (ML) training. The second layer involves using ML to proactively identify patients at high risk for medical errors, which transforms the AI tools from being a reactive system to a system that has the capability of identifying errors before they produce harm. Layer 3 addresses alarm fatigue and proposes using ML to analyze and filter out inappropriate warnings so that clinicians can focus on high priority warnings. The fourth layer of this intelligent safety stack involves overseeing transition of care processes by using generative AI to compile patient records into organized reconciliation lists that can be reviewed by the human user in a more efficient manner. Finally, the last layer of the stack involves using a bi-directional smart pump for infusion safety monitoring to prompt digital alerts in the case of error during medication administration.

Outcomes and Implications

The implementation of AI in ICU settings to promote medication safety will produce a pronounced shift in the way we surveil for errors. There will be a shift from relying on human based detection to intelligent, proactive tools integrated within the ICU workflow. Although human oversight is necessary for decision making, the integration of artificial intelligence as a means to bridge the current surveillance gap in medication errors seems to be a promising step towards supporting clinicians in preventing harmful medication errors.