Mortality and antibiotic timing in deep learning-derived surviving sepsis campaign risk groups: a multicenter study

Area of Science:

• Medical research
• Artificial intelligence in healthcare
• Clinical decision support

Background:

• Current Surviving Sepsis Campaign (SSC) guidelines recommend antibiotic timing based on sepsis probability and shock presence.
• Objective patient stratification and outcome analysis based on these recommendations are lacking.

Purpose of the Study:

• To objectively stratify sepsis patients using deep learning models.
• To analyze patient outcomes based on antibiotic timing within risk-stratified groups.

Main Methods:

• Observational cohort study using data from 2016-2024 across two health systems.
• Two deep learning models stratified patients into four risk groups based on sepsis probability and shock likelihood.
• Primary outcome was short-term mortality (in-hospital mortality or hospice transition).

Main Results:

• 34,087 adult patients with potential sepsis were analyzed.
• Mortality rates varied significantly across risk groups.
• Patients with possible sepsis and low shock risk showed similar mortality with antibiotics administered within 1 or 3 hours versus later.
• Patients with probable sepsis had lower mortality when antibiotics were given within 1 hour of triage.

Conclusions:

• A more lenient antibiotic administration time may be acceptable for low-risk sepsis patients (possible sepsis, unlikely shock).
• Prompt antibiotic administration (within 1 hour) is associated with lower mortality in probable sepsis patients, irrespective of shock risk.
• Further prospective studies are needed to confirm these findings and refine antibiotic timing guidelines.