Bayesian Health Sepsis Flagging Device

The Bayesian Health Sepsis Flagging Device is an AI/machine learning-based software that continuously monitors patient data from electronic health records to aid healthcare providers in early detection and risk prediction of sepsis developing within 24 hours. It analyzes patient information including comorbidities, lab results, vital signs, medications, and procedures to output a "Sepsis Risk High" flag displayed in the EHR, intended for use in adult patients (≥18 years) in acute care settings as an adjunctive tool alongside clinical assessments.

The subject device uses an AI/ML algorithm with 115 patient parameters from EHR data and provides continuous monitoring with flags issued within an hour of new data availability, whereas the predicate uses 22 predetermined inputs and produces a single risk score without continuous monitoring. Both are software-only prescription-use devices designed to identify sepsis risk in adult patients, but the subject device operates independently of blood culture orders and applies to all adult hospitalized patients regardless of suspected sepsis.

IEC 62304:2006/A1:2015 (medical device software lifecycle processes), IEC 62366-1:2015/A1:2020 (human factors), ANSI AAMI HE75:2009/(R)2018 (human factors), and FDA guidance on cybersecurity, software verification/validation, and human factors engineering.

Both devices are Class II AI/ML-based software under 21 CFR 880.6316 intended to aid in sepsis prediction as adjunctive tools for licensed healthcare providers. Despite differences in continuous monitoring capability and independence from blood culture orders, they share the same fundamental intended use, technological approach, and operational principle. Clinical validation on 7,732 encounters demonstrated the device identifies sepsis risk within 24 hours of onset (encounter-level PPA 79.4%, NPA 89.5%), and flag-level performance analysis supports substantial equivalence to the predicate when accounting for the broader patient population served.