Background: Sepsis, severe sepsis, and septic shock remain important contributors to hospital utilization, morbidity and mortality in the United States. Timely interventions including blood cultures, intravenous crystalloid infusions, and antibiotics make up the 3-hour sepsis bundle which represents current standard of practice, and the target of hospital quality reporting. Many aspects of sepsis care are still under discussion including best diagnostic criteria, optimal timing of treatment, and management of heterogeneity. Isolating the impact of treatment timing from other patient factors has proved challenging.

Methods: We performed a retrospective observational analysis of patients with severe bacterial illness admitted to a 6 hospital network from 2011 to 2017. The primary outcome in this study was a composite of inpatient mortality or length of stay more than 10 days, stratifying by different definitions of sepsis: 2+ SIRS criteria, 2+ SOFA score, both, or neither. We then investigated the ability to predict antibiotic latency, defined as the time from presentation to administration using patient factors prior to treatment including demographics, vitals, labs, and orders collected prior to 1, 3, and 6 hours of presentation. Finally to disentangle the impact of treatment timing from patient factors we compared a cohort that received antibiotics before time “t” against a propensity score matched control that did not across all time points between 15 minutes and 24 hours.

Results: We identified 21,425 patients admitted with severe bacterial infection, of which 29.4% had no sign of sepsis, 35% were SIRS positive, 8% were SOFA positive, and 27.3% were both SIRS and SOFA positive. In all groups antibiotic latency followed a J curve relationship with higher mortality and length of stay within the first hour and after the fourth hour. Patients were observed to have higher acuity when receiving antibiotics within the first hour, and greater complexity when receiving antibiotics late. Patient factors were able to predict with good accuracy whether antibiotics were administered within 1 hour (AUC 0.72), within 3 hours (AUC 0.74), and within 6 hours (AUC 0.76). Earlier antibiotics administration was associated with improved outcomes against matched controls for all time points later than 2.5 hours for all groups including patients who lacked any diagnostic criteria of sepsis.

Conclusions: Timing of antibiotics for patients admitted with infection is a predictable function of severity of illness and complexity. Earlier antibiotic timing was associated with improved outcomes independent of risk or diagnosis of sepsis, and easier to observe in lower risk patients. Based on these data, future efforts to improve outcomes include education of at-risk individuals to present earlier to the hospital. Additionally, decision support tools should focus on patients who benefit specifically from timely treatment and not merely identifying sepsis or absolute risk. Finally, risk adjusted latency of care is likely a better measure of quality than a 1 hour or 3 hour sepsis bundle adherence.