Integrating Precision Sepsis Risk Stratification into Systems-based Antimicrobial Stewardship

Abstract Early broad-spectrum antimicrobial administration in sepsis is strongly associated with improved survival. However, widespread implementation of time-based antibiotic mandates has substantially increased antimicrobial utilization, contributing to escalating global antimicrobial resistance. This creates a fundamental tension in sepsis care: aggressive empiric therapy optimizes short-term outcomes for individual patients but threatens long-term antimicrobial effectiveness at a population level. This perspective frames the “sepsis–antimicrobial stewardship tension” as a systems-level problem driven by diagnostic uncertainty, organizational incentives, and guideline implementation. We review emerging evidence demonstrating heterogeneity in the benefit of early antibiotics and highlight how rigid quality metrics, such as the US Centers for Medicare & Medicaid Services SEP-1 measure, may exacerbate unnecessary antimicrobial use. Finally, we discuss how advances in artificial intelligence, machine learning, and host-response diagnostics can enable precision risk–guided antimicrobial timing. Integrating these tools into clinical workflows via health systems engineering offers a path towards addressing antimicrobial resistance.