Risk factors for bacteriuria with carbapenem-resistant Klebsiella pneumoniae and its impact on mortality: a case–control study
The objective of this study was to evaluate the mortality of and risk factors for bacteriuria due to carbapenem-resistant Klebsiella pneumoniae (CRKp) versus carbapenem-susceptible K. pneumoniae (CSKp) producing extended spectrum β lactamase (ESBL).
This was a retrospective case–control study in which 135 case-patients with bacteriuria due to CRKp were compared with 127 control patients with CSKp producing ESBL. In a first step, multivariate Cox regression and Kaplan–Meier survival analysis models were used to determine the difference in mortality between the two groups and risk factors for mortality. In a second step, a univariate analysis was used to identify risk factors for CRKp colonization.
There were no significant demographic or clinical differences between the groups. In-hospital mortality in the study and control groups was 29 and 25 %, respectively (non-significant difference). Multivariate analysis revealed that the most important risk factor for mortality in both groups was being bed ridden [hazard ratio 2.2, 95 % confidence interval (CI) 1.23–3.93; P = 0.008]. Patients with CRKp bacteriuria had a longer hospitalization time with a mean ± standard deviation of 28 ± 33 days compared to 22 ± 28 days in the control group (P < 0.05). Several univariate risk factors for acquiring CRKp bacteriuria were identified: antibiotic use [odds ratio (OR) 1.93, 95 % CI 1.18–3.17, p = 0.008], especially colistin (OR 2.04, 95 % CI 1.04–4.02; P = 0.036), presence of a urinary catheter (OR 2.09, 95 % CI 1.2–3.63; P = 0.008), surgery (OR 3.94, 95 % CI 1.85–8.37; P = 0.0002), invasive procedures (OR 3.06, 95 % CI 1.61–5.8; P = 0.0004), and intensive care unit admission (OR 2.49, 95 % CI 1.18–5.37; P = 0.015).
Bacteriuria caused by CRKp as compared that caused by CSKp was not found to be a risk factor for death.
KeywordsCarbapenem resistanceKlebsiella pneumoniaeMulti-drug resistanceBacteriuriaUrinary tract infectionNosocomial infection
Klebsiella pneumoniae is an important pathogen that may cause infections in many different parts of the body, including urinary tract infections (UTI), hospital acquired pneumonia, intra-abdominal infections, and primary bacteremia [1, 2]. K. pneumoniae accounts for only 1–2 % of UTI episodes among otherwise healthy adults, but for 5–17 % of episodes of complicated UTI, including infections associated with indwelling bladder catheters [3, 4].
During the last two decades, K. pneumoniae has acquired plasmids coding for extended spectrum β-lactamase (ESBL) enzymes and as such has become resistant to many β-lactam antibiotics. One of the few clinical options for treating infections due these resistant bacteria are carbapenem antimicrobials. Therefore, the recent emergence of carbapenem-resistant pathogens is of particular concern. First, treating patients with these resistant bacteria is a therapeutic challenge [3, 5]. Second, the mechanism of carbapenem resistance in K. pneumoniae is encoded in plasmids, which can be readily exchanged between various bacterial species. For example, a KPC-2 gene (the most commonly isolated gene in multi-drug resistant K. pneumoniae) was isolated in Escherichiacoli clones of four patients in Israel . Third, the standard microbiological methods currently used to detect bacterial resistance have a limited ability to detect carbapenem-resistant K. pneumoniae (CRKp) .
Beginning in the first months of 2006, an unusual increase in the number of patients infected or colonized with CRKp has been observed in many hospitals in Israel . In a study conducted at our medical center, the prevalence of asymptomatic carriers among hospitalized patients was 5.4 % .
The prognostic impact of bacteremia from CRKp as opposed to carbapenem-susceptible K. pneumoniae (CSKp) has been evaluated in several studies. In a retrospective study comparing 53 patients with CRKp infection and those with CSKp infection there was no difference between in-hospital mortality between the groups . Conversely, a retrospective study of 48 patients in Israel reported higher mortality rate among patients with CRKp infection as opposed to CSKp infection . These studies included a mix of patients with bacteremia as well as infections in other body sites. A retrospective cohort of 190 patients with bacteremia in Israel had an infection-related mortality of 48 % for carbapenem resistance, 22 % for ESBL producers, and 17 % for susceptible K. pneumoniae .
Risk factors for acquiring infection caused by CRKp have been evaluated in several studies. In one study, risk factors for acquiring an infection with CRKp included treatment with fluoroquinolones and anti-pseudomonal penicillins . In another study, the risk factors for CRKp isolation were poor functional status, intensive care unit (ICU) stay, and antibiotic treatment, particularly fluoroquinolones . These results conflict with the results of an earlier study of 30 patients which showed the use of fluoroquinolones to be protective against the emergence of CRKp. Nevertheless, treatment with cephalosporins and carbapenems was a risk factor in both studies .
The impact of CRKp bacteriuria on mortality has not yet been investigated. Therefore, we conducted a retrospective case–control study of bacteriuria caused by CRKp versus that cause by CSKp with positive ESBL. The primary endpoint of the study was mortality. The secondary endpoint was length of hospital stay from the date of urine sample collection until discharge or death.
The study was conducted in the Shaare Zedek Medical Center, a 700-bed university-affiliated general hospital in Jerusalem. From January 2006 to April 2009, 56 patients were hospitalized with CRKp bacteremia and 177 patients with CRKp bacteriuria. Since 2007, the number of new patients infected by CRKp has dramatically decreased due to the implementation of a strict isolation policy: patients infected or carriers with carbapenem-resistant Enterobacteriaceae are cohorted in isolation zones staffed with dedicated nurses. This zone is separated physically from other patients, and strict contact isolation measures are maintained. As there were only ±2.7 patients with CRKp bacteriuria each month during the study period, it was very difficult to conduct a prospective study. Therefore, a retrospective case–control study design was chosen, consisting of two independent groups. The study population consisted of hospitalized patients aged >14 years old. A retrospective search was performed of the microbiology laboratory records to identify all patients who had a documented episode of bacteriuria caused by CRKp during hospitalization from January 2006 to April 2009.
Data was retrieved from the microbiology laboratory records for 142 patients with bacteriuria caused by CRKp during the study period. One patient was excluded because of age (<14 years) and medical records were available for six patients, resulting in 135 patients being enrolled in the study. For each patient with CRKp bacteriuria, we randomly chose a patient with CSKp and positive ESBL bacteriuria [1, 2] who had been hospitalized during the same year as the CRKp patient, at a ratio of 1:1, for a total of 135 patients. The final control group was selected from a total of 488 patients and consisted of 127 patients.
The medical records were reviewed for data regarding demographics (age, gender, residence), physical disability and cognitive status, co-morbid conditions (diabetes mellitus, congestive heart failure, chronic obstructive pulmonary disease, renal disease, hepatic disease, immune compromise, malignancy), main indication for hospitalization, procedures prior to the positive index urine culture [placement of a central venous or a urinary catheter, admission to ICU, dialysis, invasive procedures (including cardiovascular and endovascular catheterization, endoscopic procedures, and tracheostomy), surgery, and mechanical ventilation], antibiotic therapy during hospitalization, before the positive urine culture, and the classes of antibiotics received. Parallel bacteremia was considered as a positive blood culture with K. pneumoniae with similar antibiotic susceptibility as the index urine culture, obtained within a range of 4 days: up to 2 days prior or subsequent to the positive urine culture. Cases with urinary cultures positive on admission were included in the analysis.
Bacteriuria versus UTI
The aim of our study was to distinguish between asymptomatic bacteriuria and UTI. UTI was defined as one or more of the following signs or symptoms: fever of >37.8 °C with dysuria, frequency, urgency, and/or supra-pubic pain, with growth of >105 colony forming units (CFUs) per milliliter from a properly collected midstream “clean-catch” urine sample . Nonetheless, we expected that the differentiation between clinical UTI and bacteriuria might be difficult within the context of a retrospective study and therefore attempted to retrieve data from a urine dipstick specimen for presence of leucocytes and nitrites from the same day as the index culture.
Identification and antimicrobial susceptibility testing were performed by the clinical microbiology laboratory using disk diffusion and E-test methods for determining the minimum inhibitory concentration (MIC). Carbapenem (imipenem and/or ertapenem) resistance was defined as ≥16 mg/l, a positive modified Hodge test  with imipenem, and a positive PCR for the KPC gene . The production of ESBL by CSKp was determined with the double disk method . In two studies, K. pneumoniae was the third most common pathogen isolated from our patients’ blood cultures, after E. coli and Staphylococcus aureus; 48 % of hospital-acquired Klebsiella isolates were ESBL-positive [17, 18].
Our hypothesis was that patients with CRKp bacteriuria would have a mortality rate of 44 % and patients with CSKp and positive ESBL bacteriuria would have a 12.5 % mortality rate. We based our hypothesis on data collected from our hospital [19, 20] and from a retrospective study in 2007 of patients with CRKp infection in different parts of the body: most of the patients in the latter study had bacteriuria . Given 135 patients in the CRKp group and 127 patients in the control group, the study had >99 % power to detect a 12.5 % difference in mortality at a significance level of 5 % (bidirectional). Continuous variables were compared between the study groups using the two-sample t test. The χ2 test was used to compare categorical variables, and the Kaplan–Meier survival analysis model was used to analyze the mortality rates in the two groups. We also used Cox regression to examine the effect of different individual continuous risk factors on death. Factors that were found to be statistically significant by a univariate analysis with a P value of <0.05 were included in a multivariate Cox regression model to estimate the attributable mortality of each factor. All tests were two-tailed, and a P value ≤0.05 was considered to be statistically significant. We used PASW Statistics software ver. 18 for the statistical analysis.
The study was approved by the Institutional Review Board of Shaare Zedek Medical Center.
Demographic and clinical characteristics of patients with CRKp or CSKp bacteriuria
CRKp (N = 135) (%)
CSKp (N = 127) (%)
Mean ± SD
77 ± 14
80 ± 13
Median ± range
UTI was reason for admission (%)
A univariate analysis of risk factors for acquiring CRKp bacteriuria
CRKp (N = 135) (%)
CSKp (N = 127) (%)
Previous antibiotic usea
Central venous catheter
Bacteremia with K. pneumoniaeb
Admission to ICU (%)
LOS in ICU, days (SD)
Of the patients in our study 80 % were treated with antibiotics from the day the index culture was collected until 5 days later: 36 % were treated with colistin, 5 % were treated with gentamicin, and 71 % were treated with a different antibiotic, not necessarily according to the susceptibility profile of the isolated organism. The reasons for the choice of antibiotic therapy included clinical assessment of asymptomatic bacteriuria, a good clinical response to a different antibiotic that was initiated before the result of the urine culture was known, or a parallel infection that required a different antibiotic treatment.
Carbapenem-resistant Klebsiella pneumoniae constitutes an emergent public health hazard. Recent studies have focused on risk factors for acquiring this organism and the impact of bacteremia on outcome. UTI account for as many as 40–45 % of nosocomial infections, with up to 3 % of bacteriuric patients developing bacteremia. Although the contribution of UTIs to a prolongation of the hospital stay and to extra costs is only 10–15 %, these infections are important reservoirs and sources for spread of antibiotic-resistant bacteria in hospitals . The impact of CRKp bacteriuria and the risk factors for acquiring a UTI caused by CRKp have not been investigated until now and were therefore the aims of this study.
In this retrospective case–control study we compared 135 patients with bacteriuria caused by CRKp with 127 matched controls (patients with CSKp that produce ESBL). We expressly selected the latter control group in order to have similar patient populations and to be able to determine the possible impact of the presence of CRKp.
We found a crude mortality rate of 27 %—29 % in the study group and 25 % in the control group. The Kaplan–Meier analysis did not reveal a significant difference in mortality between the two groups. Nonetheless, the mortality rate of these patients was considerably higher than that of patients admitted with drug-susceptible urosepsis, reported to be 12 % in earlier studies [19, 20]. In comparison, the mortality rate of patients with nosocomial bacteriuria without bacteremia has been reported to be <5 % [22, 23]. In a multivariant analysis we found that the most important factor for death was being bed ridden. This finding indicates that these patients are severely debilitated and die of other causes not related to bacteriuria. However, there was a trend towards a higher crude in-hospital mortality in the CRKp group, and patients with CRKp bacteriuria had a longer mean hospitalization [28 ± 33 vs. 22 ± 28 days (CSKp group); P = 0.037]. These findings suggest that CRKp bacteriuria is associated with an even worse outcome than patients with CSKp that produce ESBL (control group).
The univariate risk factors for CRKp bacteriuria were prior antibiotic use, presence of a urinary catheter, invasive procedures or operations, and admission to an ICU. These findings are similar to those found in other studies in which independent risk factors for CRKp bacteremia were poor functional status, admission to ICU, and antibiotic treatment [11, 24, 25]. In addition, these factors are similar to the risk factors for acquiring infection in less resistant bacteria, such as CSKp that produce ESBL . We believe that the use of cephalosporins was higher in the CRKp group due to its prophylactic use before surgery and invasive procedures, which is also a risk factor for CRKp bacteriuria. In terms of the use of colistin, we assume that it was more common in the CRKp group since some of the patients were known carriers of CRKp and therefore treated with colistin accordingly.
Since CRKp is a nosocomial pathogen, the risk factors mentioned above emphasize the importance of minimizing antibiotic use, permanent urinary catheters, invasive procedures during hospitalization and, especially, adherence to strict infection control measures in order to reduce the prevalence of resistant pathogens in the hospitalized patient [27–29].
There are several limitations to our study. First, we chose a retrospective design since there were (fortunately) insufficient numbers of patients with CRKp bacteriuria to conduct a prospective study. Secondly, we were unable to differentiate between asymptomatic bacteriuria and clinical UTI due to lack of information in the medical records. We defined UTI according to standard clinical and laboratory indicators (see “Methods”) . However, dementia had been diagnosed in more than half of the patients, which virtually precluded the collection of reliable historical information on complaints of UTI. Of the patients in our study, 41 % had fever of >37.8 °C, but of those, 65 % had alternative causes (other then UTI) for fever, mainly pneumonia. Results of dipstick urinalysis were available only in a small minority of patients. It is possible that UTI caused by CRKp carries a different prognosis than asymptomatic bacteriuria. The differentiation between asymptomatic and clinical UTI is an inherent problem with studies on nosocomial UTIs. We believe that the prevalence of symptomatic and asymptomatic UTI was probably quite similar between the two groups.
The third limitation relates to sample size. Since the impact of CRKp bacteriuria has not been investigated until now, we estimated the sample size needed for the study based on local data [19, 20] and on results of a retrospective study that included patients with infection caused by CRKp at several different locations in the body . The mortality rate that was observed in the latter study was 44 %, considerably higher than the rate observed in our study (29 %). In addition, in Schwaber’s study , the difference between mortality rates caused by CRKp and CSKp with positive ESBL was 31.5 %, which was also much higher than the 3.7 % difference observed in our study. Therefore, the sample size that was calculated based on these data was significantly smaller than the one that was actually required to obtain a result with statistical significance. It is possible that the study should be repeated with even larger numbers of patients, but we doubt that is feasible.
In summary, bacteriuria due to CRKp was not found to be a risk factor for death compared to bacteriuria due to CSKp testing positive for ESBL. However, the mortality rate of these patients was very high (29 %), much higher than that expected for nosocomial bacteriuria or even clinical UTI. These findings imply that bacteriuria caused by CRKp is an indicator of patients who are severely ill and who die of causes other than bacteriuria. This conclusion is also supported by the fact that the surviving patients have a relatively longer mean hospitalization stay.
Conflict of interest