Hospital-level sepsis bundle compliance and its association with hospital performance in sepsis and septic shock in a nationwide cohort study

The sepsis bundle was developed to standardize sepsis treatment and improve patient outcomes. Although public health authorities have promoted its use, the impact of hospital-level compliance on patient outcomes, particularly in settings without mandatory sepsis bundle compliance, remains unclear. This prospective cohort study included 11,926 patients aged 19 years or older diagnosed with sepsis or septic shock, admitted to 19 tertiary referral hospitals from September 1, 2019, to December 31, 2020. We analyzed the associations between hospital ranks based on in-hospital mortality and compliance with the sepsis bundle at 1, 3, and 6-h intervals, as well as individual elements of the bundle. Among 11,926 patients, 3396 patients died, and per-hospital mortality rates varied from 21 to 47%. Overall compliance with the complete bundle at one-, three-, and six-hour timepoints was 18%, 59%, and 81%. The ranks based on 1-h antibiotics element showed a marginally significant positive correlation with Spearman’s Rho of 0.44 ( P = 0.06) and Kendall’s Tau of 0.33 ( P < 0.05). The ranks based on 6-h antibiotics element and the complete bundle compliance showed negative correlations with Spearman’s Rho and Kendall’s Tau of − 0.64 ( P < 0.01) and − 0.42 ( P < 0.05); and − 0.52 ( P < 0.05) and − 0.45 ( P < 0.01), respectively. To further characterize these negative correlations, breaking down the six-hour cumulative compliance rates into one-to-three and three-to-six interval compliance rates showed that one-to-three interval compliance rates retained the negative correlation, while the latter did not. The other elements—the lactate, culture, fluid, and vasopressor—did not show correlations. At the hospital level, compliance with the sepsis bundle was associated with comparative hospital performance, but these associations varied in their statistical significance, direction, and strength depending on the time point and elements used to define the compliance. Given the observational, hospital-level design—limited to 19 centers and subject to unmeasured institutional confounders and potential documentation bias—these correlations should not be interpreted as causal. In particular, the negative correlations observed at later time points should not be read as evidence that bundle completion is harmful; they more plausibly reflect delayed recognition and other confounding processes. The findings suggest that the role of the sepsis bundle as a public health interventional measure warrants further evaluation at a hierarchical level appropriate to how it is actually implemented.