Sachin Kheterpal et al., –

Despite advances in critical care, sepsis is a major source of patient suffering and mortality worldwide, with nearly 50 million cases and 11 million deaths each year. Early recognition is crucial to successful treatment of sepsis with anti-infectives, intravenous fluids, and circulatory support, and researchers have long sought to use artificial intelligence (AI) to identify patients showing early signs of the condition. But the development of AI-based sepsis prediction has been complex and hindered by shortcomings—eg, one algorithm, trained between 2008 and 2010, showed substantial deterioration in performance over time. Adoption has also outpaced external validations with proprietary models, such as the Epic Sepsis Model, widely adopted in the USA.

Read the full article in The Lancet.