Sowing the Seeds to Reap the Promise of AI in Laboratory Medicine

Diagnostic precision in clinical microbiology and infectious disease has rapidly advanced in the past twp decades, correlating with the rise of clinical genomics and proteomics, and now with the emergence of artificial intelligence (AI) for advanced data analytics. Arguably, nowhere is this more promising than for the complex and high-burden disease of sepsis, a syndrome which is a heterogenous group of clinical entities captured under one moniker. Despite major advances in our understanding of sepsis pathophysiology and its clinical manifestations, current diagnostic criteria remain limited to an operational definition anchored to downstream organ dysfunction, nonspecific biomarkers and clinical characteristics that are invariant to the underlying mechanism of disease. All these points account for the dearth of high-accuracy clinical tools that provide actionable results. The SEPSIS-SHIELD study, published by Liesenfeld et al. in September 2025 in Nature Medicine (1), signals a welcome change on the horizon. This non-interventional, prospective study, conducted in a cohort of 1222 patients presenting to the emergency room (ER) with concern for infection, silently deployed the proprietary TriVerity assay (Inflammatix, Inc.), to inform 2 critical decision points in the sepsis care trajectory, (a) the need for antibiotics (by predicting a bacterial vs viral etiology vs noninfected) and (b) the need for intensive care unit (ICU)-level care. The TriVerity assay is an FDA-cleared test that performs isothermal amplification on 29 host mRNA targets correlated with the presence and type of infection, and severity of illness. The assay reports semiquantitative results along a scale from least to most likely.