Abstract
Background: Diagnostic uncertainty in sepsis contributes to both undertreatment (delayed antibiotics) and overtreatment (unnecessary early antibiotics in nonsepsis patients). We hypothesized hospitals with strong sepsis undertreatment and overtreatment performance have better outcomes. Methods: We conducted a retrospective cohort study of hospitalized adults at 20 hospitals (Advocate Health, Michigan Medicine) between 2022 and 2024. Patients met ≥2 systemic inflammatory response syndrome criteria within 6 hours of presentation. Hospital performance was characterized using risk-standardized rates (RSRs) for undertreatment and overtreatment. “Diagnostic excellence” (DxEx) was prespecified as top-tertile performance in 1 measure without bottom-tertile performance in the other. The primary outcome was hospital mortality or hospice discharge. Secondary outcomes were hospital-free days alive at 28 days (HFD), length of stay (LOS), and total antibiotic days. Results: Among 152 779 patients, 49 109 (32.1%) received antibiotics within 3 hours and 16 361 (10.7%) ultimately met sepsis criteria. Undertreatment RSRs ranged from 29.3% to 46.8%. Overtreatment RSRs ranged from 21.5% to 37.2%. There was negative correlation between performance ranks (τ = −0.51; P = .01). Four hospitals met DxEx criteria. Receiving treatment at DxEx hospitals was associated with lower hospital mortality or hospice discharge (adjusted OR = 0.61; 95% CI = .41–.90; P = .01), greater HFD (adjusted difference = 0.59; 95% CI = .04–1.14; P = .04), reduced LOS (adjusted difference = −1.20; 95% CI = −2.23 to −.17; P = .02), and fewer antibiotic days (adjusted difference = −0.62; 95% CI = −1.28–.05; P = .07), versus non-DxEx hospitals. Conclusions: Hospital performance that jointly minimizes sepsis undertreatment and overtreatment was associated with improved patient outcomes, adjusting for patient and hospital factors. Findings highlight the importance of reducing diagnostic error in efforts to improve sepsis outcomes.
Funding Information
  • Agency for Healthcare Research and Quality

This publication has 39 references indexed in Scilit: