To devise a local strategy for eradication of a hospital-wide outbreak caused by carbapenem-resistant Klebsiella pneumoniae (CRKP).

Quasi-experimental, before-and-after, interrupted time-series study.

A 1,000-bed tertiary-care university teaching hospital.

Retrospectively, all relevant data were collected from the medical records of patients with CRKP infections from May 2006 through April 2007, the preintervention period. From May 1, 2007, through May 1, 2010, the postintervention period, the intervention was applied and prospectively followed. The 5 key elements of this strategy were an emergency department flagging system, the building of a cohort ward, the eradication of clusters, environmental and personnel hand cultures, and a carbapenem-restriction policy. The demographic and clinical parameters of patients colonized by and/or infected with CRKP were collected from medical records.

A total of 10,680 rectal cultures were performed for 8,376 patients; 433 (5.16%) and 370 (4.4%) were CRKP-colonized and CRKP-infected patients, respectively, and 789 (98%) of 803 patients were admitted to the CRKP cohort ward. The CRKP infection density was reduced from 5.26 to 0.18 per 10,000 patient-days (P<.001), and no nosocomial CRKP infections were diagnosed. Twenty-three percent of environmental cultures were found to be positive. Meropenem use was reduced from 283 ± 70.92 to 118 ± 74.32 defined daily doses per 1,000 patient-days (P<.001).

This intervention produced an enormous impact on patient location, surveillance cultures, and antibiotic policies and a massive investment in infection control resources.

To determine the attributable (direct) mortality and morbidity caused by carbapenem-resistant Klebsiella pneumoniae bacteremia.

A matched retrospective, historical cohort design, using a stepwise procedure to stringendy match the best control subjects to the best case subjects.

A 1,000-bed tertiary-care university teaching hospital.

Case subjects were defined as adult patients with carbapenem-resistant K. pneumoniae bacteremia during the period from October 2005 through October 2008. Control subjects were defined as patients who were very similar to case subjects except that they did not have bacteremia.

Matching potential control subjects to case subjects was performed at a 1:1 ratio using a computerized record system. The criteria used included same hospitalization period, similar Charlson comorbidity index, same underlying disease, same age within 10 years, and same sex. Demographic and clinical characteristics were collected from medical records.

During the study period, 319 patients developed an infection due to carbapenem-resistant K. pneumoniae. Of these 319 patients, 39 (12.2%) developed a bloodstream infection, for an overall rate of 0.59 episodes of carbapenem-resistant K. pneumoniae bacteremia per 10,000 patient-days. We excluded 7 patients from our study, leaving a total of 32 case subjects in our cohort. Case subjects were significandy more likely than control subjects (n = 32) to require care in an intensive care unit (12 case subjects [37.5%] vs 3 control subjects [9.4%]), ventilator support (17 case subjects [53.1%] vs 8 control subjects [25%]), and use of a central venous catheter (19 case subjects [59.4%] vs 9 control subjects [28.1%]). For case subjects, the crude mortality rate was 71.9% (ie, 23 of the 32 case subjects died); for control subjects, the crude mortality rate was 21.9% (ie, 7 of the 32 control subjects died) (P < .001. For case subjects, the attributable mortality was 50% (95% confidence interval [CI], 15.3%-98.6%). A mortality risk ratio of 3.3 (95% CI, 2.9-28.5) was found for case subjects with carbapenem-resistant K. pneumoniae bacteremia.

Patients with carbapenem-resistant K. pneumoniae require more intensive and invasive care. We have shown that the crude and attributable mortality rates associated with carbapenem-resistant K. pneumoniae bacteremia were striking.