Ethics approval and consent to participate
The study protocol was reviewed and approved by the ethics committee of all participating institutions in the Japanese Association for Acute Medicine (JAAM) study group. Osaka University, the representative for the JAAM Multicenter Assessment for Sepsis Treatment and Outcome (MAESTRO) study, was responsible for the overall approval (IRB number 18323).
Design and setting
This multicenter, prospective, observational study, JAAM MAESTRO (UMIN000036349), was conducted in 17 intensive care units (ICUs) in Japan from July 2019 to August 2020.
Patients were eligible for this study if they met the following criteria: 1) were older than 16 years; 2) fulfilled the Sepsis-3 criteria , i.e., had a proven or suspected infection and an acute increase of 2 or more points in Sequential (Sepsis-Related) Organ Failure Assessment (SOFA) score; and 3) were diagnosed as having only new-onset infection.
Exclusion criteria included 1) patients with cardiopulmonary arrest on hospital arrival; 2) patients with the limitation of sustained life care or post-cardiopulmonary arrest resuscitation status at the time of sepsis diagnosis; 3) patients deemed ineligible as study participants by a research director; and 4) patients transferred from other hospitals.
Data were extracted from the MAESTRO database and compiled by the MAESTRO investigators. Collected variables included relevant patient information such as demographics, comorbidities, vital signs, laboratory data, and site of infection. We also obtained data on adherence to sepsis care bundles (specifically, the hour-1 bundle).
In-hospital mortality was identified as the primary outcome. Secondary outcomes were the number of ventilator-free days and ICU-free days, length of hospital stay, and condition at discharge.
Data collection was conducted as part of the clinical routine workup. The MAESTRO site investigators recorded all data throughout the patient’s hospital stay. In the case of missing data, the MAESTRO committee requested a reconfirmation of data extraction from the MAESTRO investigators.
Sepsis care bundles were defined according to SSC guidelines  as whether all bundle components were achieved within the appropriate time frame (i.e., 1 hour) and if they adhered to the indications. Thus, if a component of the bundle was not applicable, we treated achievement of the other components as completion of the bundle (i.e., in cases where administration of crystalloid and application of vasopressor were not indicated), and adherence was defined when the other three components were completed. For all patients, bundle initiation time was defined as the time of sepsis recognition in the emergency department, ward, or ICU. Sepsis recognition was based on clinical judgement, by which the physician-in-charge suspected sepsis at the initial evaluation. The timestamp was recorded in the database by the physician-in-charge.
We divided the patients into two groups, those receiving bundle-adherent care (bundle-adherent group) and those not receiving it (non-bundle adherent group). We performed univariate analyses of the characteristics of the patients in whom the hour-1 bundle was or was not completed within 1 hour. Continuous data are expressed as mean (SD) or median (interquartile range), depending on normality. Categorical variables are shown as proportions. We also evaluated the time to completion of each component of the hour-1 bundle.
The impact of non-adherence to the hour-1 bundle on risk-adjusted in-hospital mortality was estimated using an inverse probability of treatment weighting analysis with a propensity score. The propensity score for adherence to the hour-1 bundle was determined using a logistic regression with the following covariates as independent variables, which were specified a priori based on clinical experience and prior studies: patient age, sex, admission source (emergency department, ward, or in ICU), Charlson comorbidity index (CCI), mechanical ventilation use, and each organ score within the SOFA. In addition, after replacing time to completion of each component of the hour-1 bundle as a continuous variable, we performed a multivariable logistic regression analysis, adjusted for clinically plausible and relevant confounders equal to the covariates, to calculate the propensity score. No assumptions were made on these data because the number of missing data was low.
Because two components (measure initial lactate level and begin rapid administration of crystalloid) in the hour-1 bundle were completed in almost all of the patients, we performed the same analyses excluding these two components. Two-tailed p values < 0.05 were considered to indicate significance. All statistical analyses were performed using STATA software version 15.0 (Stata Corp, College Station, TX, USA).