AI Models Target Sepsis and ICU Alarm Fatigue
Researchers are prioritizing the development of interpretable, machine learning-based models for real-time sepsis diagnosis to improve ICU outcomes. Separately, startups are using AI to reduce the number of non-actionable alarms in ICUs, addressing a key source of cognitive overload for clinical staff.
- One machine learning model, the Artificial Intelligence Sepsis Expert (AISE), can predict the onset of sepsis in ICU patients 4 to 12 hours before it would be clinically recognized by analyzing 65 different features from EMR and high-resolution vital sign data. - AI-based alarm management systems have been shown to reduce the number of false alarms in ICUs by as much as 92.25% and decrease the total number of notifications caregivers receive by up to 99.3%. - To qualify for the American Nurses Credentialing Center (ANCC) Informatics Nursing Certification (RN-BC), registered nurses generally need a bachelor's degree, two years of full-time RN experience, and either 2,000 hours of informatics practice or 1,000 hours plus 30 hours of continuing education in informatics within the last three years. - A significant source of frustration for nurses using EHRs like Epic is the physician-centric design, which can lead to redundant data entry and excessive clicking. A survey of over 9,000 nurses revealed that more than two-thirds believe poor EHR usability and the burden of digital documentation contribute to their job dissatisfaction. - The Office of the National Coordinator for Health Information Technology (ONC) and Centers for Medicare & Medicaid Services (CMS) have finalized rules mandating the use of Fast Healthcare Interoperability Resources (FHIR) APIs. This is intended to give patients greater access to their health information and facilitate data exchange between providers and payers. - The HL7 FHIR standard is foundational for the next generation of healthcare AI applications, providing a standardized way to represent and exchange data for use in clinical decision support and other AI-driven tools. This interoperability is crucial for integrating the various IT systems—such as EHRs, lab systems, and billing systems—that often lack seamless communication. - Nurses frequently report that the time required to navigate and document in EHR systems reduces the time they can spend at the bedside with patients, leading to feelings of caring for the computer rather than the patient. Common requests for EHR improvement from nurses include more mobile-friendly interfaces, voice-enabled documentation, and integrated tools to support handoffs. - Recent EHR implementations are associated with increased stress and time pressure for nurses, particularly within the first six months. Cognitive failures were found to be highest among nurses who had experienced an EHR implementation within the preceding 6 to 12 months.
