Background: Carbapenem-resistant GNB infections are a serious public health problem worldwide, particularly due to the high mortality associated with them and the low number of therapeutic options. One approach to this challenge is the development of antimicrobial stewardship programs. Objective: We evaluated the impact of a carbapenem restriction program on reducing of bacterial resistance in an intensive care unit (ICU). Methods: A retrospective study conducted in 2 phases in the 80-bed ICU of an acute-care public hospital in Minas Gerais, Brazil. The preintervention phase lasted 16 months (January 2018–April 2019) and the second phase (carbapenem restriction), after the intervention, lasted 4 months (May–August 2019). The intervention was defined as carbapenem-sparing and the use of meropenem was authorized in 3 situations: (1) treatment of serious infections documented by extended-spectrum β-lactamase–producing Enterobacteriacea (ESBL); (2) therapeutic failure with the use of another antimicrobial; and (3) infectious disease recommendation. Data were obtained through consultation of electronic medical records and microbiological results, as standardized by the CLSI, for patients with a >48-hour stay in the ICU and who met the criteria for healthcare-associated infection (HAI) according to the CDC NHSN definition. Results: Before the intervention, on average, 50 cultures were obtained with positive results for multidrug-resistant GNB–MER-GNB (SD, 12.2) and in the intervention phase, this number was 31 cultures (SD, 12.8; P = .010). Average carbapenem consumption decreased significantly with corresponding increase in cefepime consumption in the same period (Fig. 1). The ATB (DDD per 1,000 patient days) before the intervention for carbapenems was 110.6 (SD, 97.1) and for cefepime was 8.2 (SD, 5.9). In the intervention phase, the ATB for carbapenems was 44.7 (SD, 38.5; P = .015) and for cefepime it was 32.0 (SD, 20.3; P < .001). In terms of multidrug resistance rate, before the intervention, 95 of 149 of Acinetobacter (64%) were resistant and during the intervention, 13 of 30 Acinetobacter (43%) were resistant (P = .043). Other GNB (Klebsiella, Proteus, Escherichia coli, and Pseudomonas) reduced the resistance rate, but without statistical significance. We observed a reduction in the HAI rate per MDR-GNB (Fig. 2): before the intervention, it was 22.7 (SD, 5.5) and during the intervention phase it was 16.5 (SD, 7.7; P = .07), although this change did not reach statistical significance. Nevertheless, the ICU Klebsiella infection rate did significantly decrease; it was 5.5 (SD, 1.9) before the intervention and 2.4 (SD, 1.8) after the intervention (P = .009). Conclusions: Short-term carbapenem restriction may be an effective strategy to reduce the incidence of carbapenem-resistant GNB infections in the ICU. The scarce arsenal available for the treatment of MDR-GNB and the high mortality rate justify the growing need for stewardship programs in Brazilian ICUs.