The ASP, ID team, and intensive care unit (ICU) physicians approved a plan to reduce use of empiric carbapenems in the ICU and use colistin, tigecycline, or both for patients confirmed with or at high risk for A. baumanii infections. ID physicians evaluated the clinical severity and hemodynamic stability of each patient and had final discretion to prescribe either colistin or tigecycline.

We included all ICU admissions in the study, even recurrent admissions. This ICU has a multidrug-resistant organism surveillance program that collects a sputum sample for culture every third day for intubated patients with abundant secretions. We used these cultures for our evaluation. We considered any culture sample outside this practice a duplicate and excluded it from our analysis. During the study period, we did not modify infection control practices. The study was approved by the institutional review board of SGHUMC.

We retrieved data from the hospital’s computerized ordering system and examined medical records of all ICU admissions during February 1, 2016–January 31, 2017. Clinical data included patient demographics, admission diagnosis, and presence of mechanical ventilation. During February 1–June 30, 2016 (period 1), patients received colistin/carbapenem combination therapy for A. baumanii infections. During July 1, 2016–January 31, 2017 (period 2), we applied our intervention. We recorded the total number of bacterial cultures collected from the ICU and noted the site and date of sampling.

We considered the isolation density the number of clinical isolates/1,000 patient-days and the rate of ventilator-associated pneumonia (VAP) the number of VAP events/1,000 patient-days. We defined variables according to guidelines for XDR A. baumanii from the US Centers for Disease Control and Prevention and World Health Organization (10). We calculated case-fatality and VAP rates following guidelines from the American Thoracic Society and Infectious Diseases Society of America (11).

We grouped antimicrobial drugs into 5 categories: group 1, antimicrobial drugs that do not require ID preapproval, such as third-generation cephalosporins, amoxicillin/clavulanic acid, and quinolones; group 2, oral vancomycin and metronidazole used for Clostridioides difficile therapy; group 3, imipenem and meropenem; group 4, broad-spectrum carbapenem-sparing regimens, including piperacillin/tazobactam, cefepime, ceftazidime, amikacin; and group 5, the XDR A. baumanii–active antimicrobial drugs colistin and tigecycline. We measured antimicrobial drug consumption by DDD per 1,000 patient-days (Table 1).

We sent 48 laboratory-confirmed A. baumanii isolates, 31 collected during period 1 and 17 during period 2, to IHU-Méditerranée Infection, Aix-Marseille, France, for testing. Samples underwent 4 types of testing: matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (Microflex; Bruker Daltonics, https://www.bruker.com); antimicrobial susceptibility testing by disk diffusion method and interpreted according to the European Committee of Antimicrobial Susceptibility Testing 2017; real-time PCR to screen for carbapenemase-encoding genes; and multilocus sequence typing to determine genetic relationships among the isolates.

The ICU admitted 536 patients during the study period; 3 were readmissions. Patient characteristics between the 2 periods were statistically similar (Table 1). Throughout the study, the incidence of A. baumanii VAP decreased from 154.9 to 38/1,000 patient-days (p = 0.007) and the A. baumanii VAP case-fatality ratio dropped from 79 to 12/1,000 patient-days. Non–A. baumanii VAP incidence decreased from 62 to 51/1,000 patient-days. ICU mortality rates from all causes remained unchanged between period 1 and period 2 (Table 1).

Figure 1

Figure 1. Isolation density of Acinetobacter baumanii in sputum cultures versus carbapenem consumption in the intensive care unit (ICU) of Saint Georges Hospital University Medical Center, Beirut, Lebanon, during February 1, 2016–January 31,...

Consumption of group 1 and group 4 antimicrobial drugs was statistically similar during the 2 periods (Table 1). Carbapenem consumption decreased by 59%, a total of 318 DDD/1,000 patient-days, and overall restricted antimicrobial drug consumption dropped 637 DDD/1,000 patient-days (p<0.005). Because isolation of A. baumanii decreased substantially, colistin consumption also decreased by 55%, from 20 DDD/1,000 patient-days in period 1 to 9 DDD/1,000 patient-days in period 2 (p = 0.019) (Figure 1). Tigecycline consumption remained statistically unchanged (84 DDD/1,000 patient-days in period 1, 62 DDD/1,000 patient-days in period 2). Of note, group 2 C. difficile therapy consumption dropped by 231 DDD/1,000 patient-days (p = 0.042), a 51% decrease that likely mirrors reduction in C. difficile infections.

Figure 2

Figure 2. Isolation density of Acinetobacter baumanii and non–A. baumanii in the intensive care unit (ICU) of Saint Georges Hospital University Medical Center, Beirut, Lebanon, during February 1, 2016–January 31, 2017. Rates are...

The A. baumanii isolate density in sputum cultures decreased by 70.7%, from 82 to 24/1,000 patient-days, positively correlating with the fall in carbapenem consumption (p = 0.004) (Figure 1). The number of non–A. baumanii multidrug-resistant (MDR) isolates did not increase (Figure 2).

All 48 A. baumanii isolates carried extended-spectrum β-lactamase bla_TEM-1 genes. The 31 isolates from period 1 were XDR; 30 carried the class D carbapenemase bla_oxa-23 gene, and 1 carried the bla_oxa-24 gene. Multilocus sequence typing revealed 3 sequence types (STs) in period 1: ST2, 29/31 (93.5%); ST699, 1/31 (3%); and ST627, 1/31 (3%) (Table 2). In period 2, A. baumanii ST2 disappeared; 58.8% (10/17) of isolates belonged to ST25 and 5.9% (1/17) belonged to ST99. The remainder belonged to 6 new STs, assigned ST1200, 1201, 1202, 1203, 1204, and 1205 (35.2%). Of the 17 isolates from period 2, 6 carried the bla_oxa-23 gene, 5 the bla_oxa-24 gene, and 3 both genes.

Overall, XDR A. baumanii isolation decreased by 64.7% from period 1 to period 2. In addition, isolates from period 2 were more antimicrobial-susceptible than in period 1: 64.8% (11/17) sensitive to ceftazidime and cefepime, 17.6% (3/17) to piperacillin/tazobactam, and 17.6% (3/17) to carbapenems (Table 2).

Our prudent use of antimicrobial drugs did not increase mortality rates and had a dramatic effect on antimicrobial consumption and MDR A. baumanii isolate density. A longer study period and larger sample likely would reveal additional effects on XDR infections and outcomes. Many factors could have affected the study results, including patient referrals and seasonality. However, the microbiological findings strongly point to high rates of carbapenem consumption as a sustaining factor in survival of XDR A. baumanii ST2 in our facility. By reducing carbapenem consumption, we broke a vicious cycle.

In the era where clinicians must manage severely ill, MDR-colonized patients, relying on existing guidelines is not enough. A creative, multidisciplinary approach with knowledge of local epidemiology is key to controlling MDR and XDR infections. Investing time in accurate diagnosis and implementing targeted carbapenem-sparing strategies for initial treatment is only possible through trusted collaboration between ID and ICU physicians. The dedication of the ASP and microbiology departments at this facility is an example of a successful active surveillance program for antimicrobial drug consumption and resistance profiles, especially when developing standards of care tailored to meet an institution’s needs.