The goal of this clinical trial is to learn if an educational intervention with audit and feedback on physicians and health care professionals who participate in antimicrobial treatment decisions can reduce the use of antimicrobials in adult patients admitted to a sample of Brazilian intensive care units (ICUs). The educational intervention is based on a literature review of current recommendations for a more rational use of antimicrobials and microbiological tests in daily ICU practice.

The main questions it aims to answer are:

• Does the educational intervention reduce the antimicrobial consumption in the intensive care units?
• Does this educational intervention aiming to reduce antimicrobial utilization in accordance with the latest guidelines have any safety signals regarding ICU mortality rates or ICU length-of-stay?

Researchers will compare (1) ICUs sequentially randomized to this quality improvement educational intervention aimed at improving antimicrobial utilization to (2) the same ICUs at months where the educational intervention has not been delivered yet.

Each participant ICU will transition to the quality improvement intervention approximately each month, starting at July, 2025. This quality improvement intervention is based on current recommendations for antimicrobial stewardship from regulatory agencies and medical societies, including cognitive aids for physicians to improve decision-making regarding the commencement of antimicrobials, their duration and antimicrobial time-outs. The investigators hypothesize that intensivists (ICU doctors) need to embrace antimicrobial stewardship as a core competence of their daily activities.

The SAFE-REDUCE trial is a hybrid effectiveness-implementation type 1 trial, where the investigators intend to implement a quality improvement (QI) intervention aiming to improve antimicrobial utilization in a sample of intensive care units. Their hypothesis is that this QI intervention will reduce antimicrobial consumption in the studied intensive care units. For this implementation, due to logistical constraints, the investigators can only train one intensive care unit at a time and therefore they designed a stepped-wedge cluster randomized trial to evaluate the impact of this QI intervention.

The trial will be conducted in 10 intensive care units currently participating in the IMPACTO-MR platform, a registry focused on evaluating the impact of intensive care unit acquired infections and multidrug resistant infections. IMPACTO-MR routinely collects data from the participating ICUs, including standard ICU quality improvement data. Additionally, the platform collects data on antimicrobial utilization, reported ICU-acquired infections and microbiological culture results. In the SAFE-REDUCE trial, the investigators will take advantage of this ongoing registry to test if the educational QI intervention can modulate antimicrobial utilization while maintaining the trends of ICU mortality and length-of-stay.

Each participant ICU will go through a control phase, in which they will follow their current practice, and they will be transitioned to an intervention phase, in which the quality improvement intervention will be delivered to the intensive care unit.

The QI intervention will provide cognitive aids to ICU physicians aiming at improving their decision-making for antimicrobial initiation, antimicrobial time-outs (i.e, routine revisions of the actual need for antimicrobials) and, when an infection is highly likely, deciding appropriate, shorter durations of antimicrobials. The investigators will provide intensive care units with educational sessions, a written guide for antimicrobial initiation and duration decision-making, while also accounting for key diagnostic stewardship considerations. Following an initial site visit, the investigators will collect monthly point-prevalence surveys in the ICUs regarding antimicrobial utilization in one day of the month, for the duration of the trial. During these point-prevalence surveys, feedback will be provided to the ICU provider to be delivered to the ICU team. Additionally, during the first site visit, ICUs will be encouraged to evaluate barriers and facilitators to implement the educational intervention in their units, and they will be allowed to adapt further local training based on this assessment.

All participating ICUs of the SAFE-REDUCE stepped-wedge cluster randomized trial will be transitioned to the quality improvement intervention. The investigators included a baseline period (one month before implementation, expected on June, 2025, where only routinely collected data from the IMPACTO-MR platform will be linked to SAFE-REDUCE) in all ICUs and a phase-out phase (for three months, expected to occur on May-Jul, 2026) to evaluate whether any intervention benefits sustain after the investigators stop the point-prevalence surveys and feedback.

The investigators calculated sample size based on Hemming et al. suggested method. Considering that the study ICUs screened for participating in the SAFE-REDUCE trial would have between 10 and 30 ICU beds, most of them with 20 ICU beds, assuming a conservative occupation rate of 85%, the investigators estimated that 600 patient-days per period would be included, with each participating ICU transitioning to the intervention period in 10 steps. The investigators estimate the intervention will be able to reduce up to 20% the antimicrobial consumption in the ICU and they anticipate a variable baseline rate of antimicrobial utilization, considering an intracluster correlation coefficient (ICC) of 0.1. The investigators tested different ICCs, baseline daily defined doses and reductions in their level and the sample size yielded > 90% power in all simulations.

Given the analysis at the cluster level, informed consent was waived from the institutional review board at the coordinating center and at each participating hospital institutional review boards. They agreed with a minimal risk study (i.e., participation in this study will not incur higher risk than usual random variability observed in clinical practice) and the impossibility of obtaining informed consent to withdraw data utilization, as the primary outcome data is collected at the level of the unit, not at the level of the participants.

Intensive care units inclusion Criteria:

• ICU leadership acceptance to participant in the quality improvement intervention;
• Hospital infection control leadership acceptance to participant in the quality improvement intervention;
• ICU participation in the IMPACTO-MR platform with high quality data;
• ICU potential for quality improvement based on a subjective assessment of the ICU and hospital infection leaderships

Intensive care units exclusion Criteria:

- Absence of local IRB approval

Participants inclusion Criteria:

• All patients admitted to the intensive care unit

Participants exclusion Criteria:

• Patients younger than 18 years-old will be excluded from individual-level analysis

Individual patient data (IPD) might be shared upon request and approval by the study steering committee after the study completion.