AI Models Show Promise for Early Sepsis Detection
A recent demonstration highlighted the use of H2O Driverless AI for clinical decision support, specifically for the early detection of sepsis in the ICU. The application shows the growing maturity of AI tools designed to analyze real-time patient data and provide actionable insights. For successful adoption, such tools must be tightly integrated into EHR workflows to support clinical judgment without causing alert fatigue.
- AI models have demonstrated high accuracy in predicting sepsis, with some achieving an area under the receiver operating characteristic curve (AUROC) between 0.68 and 0.99, outperforming traditional scoring systems. However, a 2023 study found Epic's sepsis model missed two-thirds of cases, prompting the company to release an updated version to reduce false positives and improve timeliness. - To transition from the ICU into informatics, the key credential is the Informatics Nursing Certification (NI-BC) from the American Nurses Credentialing Center (ANCC). Eligibility requires an active RN license, a BSN, two years of full-time practice, 30 hours of informatics continuing education, and 1,000 to 2,000 hours of informatics practice in the last three years. - You can leverage critical care expertise for health IT roles by positioning yourself as a clinical subject matter expert on technology projects, where workflow knowledge is highly valued. Gaining practical experience can be done internally by participating in EHR implementation teams or quality improvement initiatives that focus on system optimization. - The interoperability standard enabling AI tools to integrate with EHRs is HL7 FHIR (Fast Healthcare Interoperability Resources). It uses modern web-based APIs to define data formats and elements for exchanging electronic health records, allowing different systems to communicate seamlessly. - Common complaints from frontline nurses about EHRs like Epic include cumbersome workflows that increase documentation time and reduce direct patient care, as well as frequent, inaccurate alerts from predictive models that lead to alert fatigue. Nurses have reported that such systems can feel counter-intuitive and were not designed in concert with clinicians. - The 21st Century Cures Act and its associated ONC Final Rule are key federal policies that mandate against "information blocking." This requires hospitals to provide patients with access to their electronic health information without charge and adopt standardized APIs, directly influencing health IT priorities and EHR configurations.
