Decision analysis model of rapid versus deferred antibiotic initiation in patients with suspected sepsis in the emergency department.

Abstract

Sepsis remains a major health concern with high associated mortality. Adequate treatment involves the use of antibiotic therapy although the timing of antibiotics is controversial. A decision analysis model of antibiotic initiation was created to determine optimal management of patients with suspected sepsis. Two decision trees were created using data from the published literature. A limited model used mortality as the primary outcome using the impact of antibiotic timing on rates of progression to shock and in-hospital mortality. The primary model included mortality and stewardship-related factors such as antibiotic avoidance and antibiotic-associated adverse events. Rapid initiation of antibiotics was defined as universal antibiotic administration within 3 h of presentation whereas deferred initiation included administration up to 6 h. Sensitivity analyses were performed to evaluate the effectiveness of each option. When considering only mortality, rapid initiation was the optimal strategy. When considering stewardship-related factors, rapid initiation of antibiotics maximized utility in only 40.6% of model iterations. One-way sensitivity analysis demonstrated rapid initiation of antibiotics was optimal when initiation times were above 1.33 h and the prevalence of infection was above 89.5%. Two-way sensitivity analysis demonstrated that as time to antibiotics increased, rate of true infection above which rapid antibiotics is optimal drops from just under 91% to approximately 88.5%. We constructed decision analysis models to characterize optimal conditions for antibiotic initiation in suspected sepsis. Our model suggests that the prevalence of infection needs to be approximately 90% for rapid initiation of antibiotics to be the optimal strategy.