Case-Case-Control Study of Patients with Carbapenem-Resistant and Third-Generation-Cephalosporin-Resistant Klebsiella pneumoniae Bloodstream Infections

Abstract

Strains of third-generation-cephalosporin-resistant Klebsiella pneumoniae (3GCRKP) and carbapenem-resistant K. pneumoniae (CRKP) are rapidly spreading. Evidence is needed to establish whether differences exist between patients at risk for 3GCRKP and those at risk for CRKP bloodstream infections (BSIs); thus, this retrospective case-case-control study was conducted to determine if the risk factors for these two infections differ. The inclusion criteria for cases were positive blood cultures for K. pneumoniae, first episode of BSI, age of ≥18 years, and susceptibility results indicating resistance to either third-generation cephalosporins (3GCRKP group) or carbapenems and cephalosporins (CRKP group). Controls were patients admitted for ≥72 h and were matched to cases by month/year and medical unit. Variables of interest were analyzed by univariate analysis, and those of significance were analyzed by logistic regression. In total, 111 patients with 3GCRKP BSIs and 43 patients with CRKP BSIs were matched to 154 controls. Multivariate analyses of 3GCRKP case and control groups demonstrated that a length of stay (LOS) of >40 days (odds ratio [OR], 17.7; 95% confidence interval [CI], 3.7 to 84.3), the use of antibiotics in the past 90 days (OR, 4.3; 95% CI, 1.5 to 11.9), and the presence of a central venous catheter (OR, 4.1; 95% CI, 1.3 to 13.4) were independent risk factors. Multivariate analyses of the CRKP case and control groups demonstrated that a LOS of >40 days (OR, 13.5; 95% CI, 2.9 to 62.8) and the use of antibiotics in the past 90 days (OR, 5.9; 95% CI, 1.3 to 26.5) were independent risk factors. Similar factors put patients at risk for these two types of K. pneumoniae BSIs.

INTRODUCTION

Resistance to beta-lactam agents by Gram-negative bacilli (GNB) is reaching crisis levels (1–3). While multiple mechanisms of GNB resistance to beta-lactams exist, the most common form is the production of beta-lactamase enzymes (4). Extended-spectrum beta-lactamases (ESBLs) confer resistance to penicillins, aztreonam, and most cephalosporins. Carbapenems are generally stable against ESBLs. Genes encoding for ESBLs have spread worldwide among many different species of GNB, and these organisms are a major threat to public health (4, 5).

In recent years, carbapenemase production has begun to spread among Enterobacteriaceae (6, 7). These carbapenem-resistant Enterobacteriaceae (CRE) are resistant to all beta-lactam agents and frequently harbor resistance genes for multiple classes of drugs. Some isolates that are resistant to all tested agents have been described (8). Most CRE are Klebsiella pneumoniae, a common pathogen in human infection that causes infections of the bloodstream, urinary tract, lungs, skin, and other sites (7, 9). Bloodstream infections (BSIs) caused by these organisms have a high attributable mortality rate, estimated to be as high as 50% (10–12).

Several studies have explored risk factors associated with the development of infection with ESBL-producing organisms and with CRE (12–17). However, studies comparing risk factors between the two types of infections have been lacking. Since available treatment options for both types of resistant organisms are limited, it would be useful to determine if there are differential risk factors that can predict either type of resistance pattern in an infected patient. Since patients with CRE infections frequently require therapy with toxic antimicrobials, such as polymyxins and aminoglycosides, the ability to predict these infections by risk factors would be useful.

This study was performed to determine the risk factors for BSIs caused by third-generation-cephalosporin-resistant K. pneumoniae (3GCRKP) and carbapenem-resistant K. pneumoniae (CRKP). After risk factors were determined, we compared them to see if any differential risk factors existed between the groups. A secondary objective was to determine if exposure to specific antibiotic classes predisposes patients to BSIs with either resistance type.

MATERIALS AND METHODS

Patients.

This was a study of patients at Temple University Hospital, a tertiary-care academic medical center in Philadelphia, Pennsylvania, from June 2005 through October 2010. To minimize the biases associated with direct comparisons between infected populations, we utilized a case-case-control design. Case patients were matched with control patients on a 1:1 basis by location (hospital unit) and time (within 30 days). Control patients who were discharged after <72 h were not included since they did not have an adequate opportunity to develop a BSI. To be included, case patients had to be ≥18 years of age with a positive blood culture for K. pneumoniae with resistance to either third-generation-cephalosporins (3GCs) with carbapenem susceptibility (3GCRKP group) or carbapenems and 3GCs (CRKP group). Susceptibilities were determined by the BD Phoenix system using breakpoints from the 2009 Clinical and Laboratory Standards Institute (CLSI) M100-S19 performance standards, published before the breakpoints for Enterobacteriaceae resistance to carbapenems and cephalosporins were revised (18). Patients without clinical evidence of infection and episodes of K. pneumoniae BSI after the initial index infection were excluded. This study was approved by the Temple University Institutional Review Board.

Data collection.

The data collected on subjects included information on comorbidities, clinical status, intensive care unit (ICU) admission, previous hospital admissions, antibiotic use, presence of urinary and/or central catheters, length of stay, medical procedures, and severity of illness. The severity of illness was determined by calculation of the Pitt bacteremia score (13) and the acute physiology and chronic health evaluation II (APACHE II) score (19). These were calculated for case patients at the time that the index blood culture was collected. They were not calculated for control patients because these patients were not necessarily infected. The degree of comorbidity was determined by the Charlson comorbidity index (20). Hospitalizations within the 90 days before admission were considered recent.

Statistical analysis.

Within the two case groups, univariate comparisons of each covariate of interest between the case group and the matched control group were made using the chi-square test, stratifying on the matched pairs. Significant covariates were then selected for multivariate analysis by stepwise forward logistic regression with a P value of 0.1. Conditional logistic regression was used to account for pair matching. All calculations were made using Stata 13 software (Stata Corp, College Station, TX). Statistical significance was set at a P value of ≤0.05.

RESULTS

During the study period, we enrolled 111 patients with 3GCRKP BSIs and 43 patients with CRKP BSIs. The demographics for these patients and their matched controls are listed in Table 1. Table 2 lists the characteristics of the patients during and immediately before their hospital stay. There was a high degree of acute illness in both case groups, with mean (± standard deviation) APACHE II scores of 17 ± 7 and 19 ± 7 in the 3GCRKP and CRKP case groups, respectively. Notable characteristics included high percentages of patients on ventilators and with both urinary and central venous catheters. The mortality rates in the two case groups and the control group were high and were significantly higher in the two case groups. In the 3GCRKP group, 32% of the patients died, compared with 13% of the controls (odds ratio [OR], 3.6; 95% confidence interval [CI], 1.2 to 10.5). In the CRKP group, 45% of the patients died, compared with 18% of the controls (OR, 3.8; 95% CI, 1.4 to 9.9).

TABLE 1.

Patient demographics

Demographic	Data for indicated group (no. of patients)
	3GCRKP (111)	3GCRKP control (111)	CRKP (43)	CRKP control (43)
Age (yr)a	57	52	56	58
Male sex (%)	55	43	61	61
Charlson comorbidity indexa	3	2	3	3
Pulmonary disease (%)	35	34	50	41
Cardiovascular disease (%)	28	34	55	57
Diabetes (%)	39	23	32	34
Immunocompromised (%)	17	19	18	16
Recent hospital admission (%)	62	38	43	25
Recent ICU admission (%)	32	13	32	9
Antibiotics in past 90 days (%)	39	18	41	15

^aMeans are shown for continuous variables.

TABLE 2.

Patient characteristics

Characteristic	Data for indicated group (no. of patients)
	3GCRKP (111)	3GCRKP control (111)	CRKP (43)	CRKP control (43)
Admitted from (%):
Home	68	77	59	70
Nursing home	14	6	18	5
Outside hospital	19	17	23	25
Mechanical ventilation (%)	62	50	84	34
Dialysis (%)	27	6	27	18
Urinary catheterization (%)	79	58	80	52
Central catheter (%)	79	43	75	41
Length of stay (days)a	63	20	54	19
Length of ICU stay (days)a	21	7	22	7
No. of antibiotic classes	4	3	5	3
Total antibiotic daysa	25	11	27	10
Pitt bacteremia scorea	3	NAb	4	NA
APACHE II scorea	17	NA	19	NA
Mortality (%)	32	13	45	18

^aMean values are shown.

^bNA, not applicable.

Table 3 contains the results of the univariate analyses. As expected, many factors were associated with infection in the 3GCRKP and CRKP groups. In the 3GCRKP group, the factors associated with infection included recent hospital admission, the use of antibiotics within 90 days, a length of stay of >40 days, liver disease, dialysis, the presence of a central venous catheter, and the use of penicillins, cefepime, carbapenems, vancomycin, metronidazole, or fluoroquinolones. Fewer patients were in the CRKP group than in the 3GCRKP group, and fewer factors were significantly associated with infection. Significant factors in this group included recent ICU admission, the use of antibiotics within 90 days, mechanical ventilation, and antifungal therapy. Mutual factors associated with infection in both groups were recent ICU admission, the use of antibiotics within 90 days, a length of stay of >40 days, and the presence of a central venous catheter.

TABLE 3.

Univariate analyses

Risk factor	3GCRKP group		CRKP group
	Odds ratio	Confidence interval	Odds ratio	Confidence interval
Previous admissions
Recent hospital admission	2.9	1.3–6.3a	—b	—
Recent ICU admission	2.7	0.9–7.7	38.5	2.4–624a
Antibiotics in past 90 days	4.4	1.7–11.7a	5.6	1.3–26.5a
Current admission
Length of stay > 40 days	24.0	5.7–101.9a	7.2	1.3–38.9a
Renal disease	2.4	0.8–7.2	—	—
Liver disease	4.7	1.1–19.6a	—	—
Dialysis	5.3	—	—	—
Mechanical ventilation	—	—	10.6	1.3–88.6a
Central catheter	4.3	1.7–11.1a	3.1	0.6–15.7a
Antifungal therapy	—	—	5.4	1.1–27.1a
Penicillins	6.7	1.4–31.9a	1.2	0.4–3.4
Cephalosporins	2.8	0.8–10.5	—	—
Fourth-generation cephalosporins	5.8	1.0–34.0a	1.9	0.4–8.9
Carbapenems	17.9	1.5–216.6a	—	—
Vancomycin	6.7	1.8–25.5a	1.4	0.4–5.2
Metronidazole	6.5	1.1–37.1a	1.7	0.5–5.6
Fluoroquinolones	3.4	1.3–12.4a	—	—
Aminoglycosides	3.0	0.5–18.9	—	—
Colistin	8.9	0.4–203.6	—	—
Tigecycline	3.3	0.4–30.2	—	—

^aP < 0.05.

^b—, not entered into multivariate model (P > 0.1).

The results of the multivariate analyses are presented in Tables 4 and 5. Logistic regression identified three factors that remained associated with 3GCRKP BSI: the use of antibiotics within 90 days (OR, 4.3; 95% CI, 1.5 to 11.9), a length of hospital stay of >40 days (OR, 17.7; 95% CI, 3.7 to 84.3), and the presence of a central venous catheter (OR, 4.1; 95% CI, 1.3 to 13.4). Two factors remained associated with CRKP BSI in a fully adjusted model: the use of antibiotics within 90 days (OR, 5.9; 95% CI, 1.3 to 26.5) and a length of hospital stay of >40 days (OR, 13.5; 95% CI, 2.9 to 62.8).

TABLE 4.

Significant^a risk factors by multivariate analysis of 3GCRKP case and control groups

Risk factor	Adjusted odds ratio	Confidence interval
Antibiotic in past 90 days	4.3	1.5–11.9
Length of hospital stay >40 days	17.7	3.7–84.3
Central venous catheter	4.1	1.3–13.4

^aSignificance was defined as a P value of <0.05.

TABLE 5.

Significant^a risk factors by multivariate analysis of CRKP case and control groups

Risk factor	Adjusted odds ratio	Confidence interval
Antibiotic in past 90 days	5.9	1.3–26.5
Length of hospital stay >40 days	13.5	2.9–62.8

^aSignificance was defined as a P value of <0.05.

DISCUSSION

This study evaluated the risk factors for Klebsiella pneumoniae BSIs for two different resistant phenotypes. Third-generation cephalosporin resistance in Klebsiella species is primarily mediated by ESBLs and has spread worldwide throughout Klebsiella and other Enterobacteriaceae (4). While it is the more common of the two mechanisms, the difficulty in treating infection caused by ESBL-producing GNB was recently superseded by the rising prevalence of carbapenemases such as K. pneumoniae carbapenemase (KPC) (6). Due to concomitant multiple-resistance mechanisms, the CRE that produce carbapenemases are generally resistant to most or even all classes of antibiotics and may not be effectively treated by any of them. As CRE continue to emerge, it would be useful to determine which patients with K. pneumoniae BSIs are likely to have either type of resistance. This study was conducted to determine if differential risk factors can be used to improve empirical therapy in patients at risk for CRE BSIs.

Our study suggests that differentiation between risk factors for BSIs caused by 3GCRKP and CRKP may not be possible. Of the three significant risk factors for 3GCRKP BSIs, CRKP patients shared two: increased length of stay and previous antibiotic use. Only central venous catheter use differentially predicted 3GCRKP BSIs, and this may have not been significant in the CRKP group only because of lesser power due to the low number of patients. Therefore, CRKP may simply be a new resistant phenotype of K. pneumoniae that causes BSIs in the same type of patients who contract BSIs with 3GCRKP. Instead of relying on risk stratification, surveillance cultures for patients at high risk of resistant GNB BSI may be a useful option (21, 22).

On multivariate analysis, no antibiotic classes emerged as risk factors for either type of resistant K. pneumoniae BSI. However, overall antibiotic use within 90 days significantly predicted both phenotypes of K. pneumoniae BSI. Studies of risk factors for CRKP infection have differed in their findings on antibiotic classes as risk factors (10–12, 14–17). We chose the case-case-control or double case-control study design to more accurately ascertain risks for resistant infection (23–25). A study design that directly compared CRKP BSI cases to cases with 3GCRKP BSIs may have exaggerated the risks of antibiotic use. Our use of matching by location and time gave case and control patients similar opportunities for infection with resistant K. pneumoniae without amplifying the influence of antibiotic use, as has been seen in some studies comparing vancomycin-resistant and vancomycin-susceptible enterococci (23). We also chose to match control patients for each resistant group separately instead of analyzing the same controls with each resistant group. This approach was taken because with the relatively rare finding of resistant K. pneumoniae (particularly CRKP), we thought it was important to match patients with controls who had an opportunity to develop infection with the same organism.

This study is limited by several factors. The analysis was retrospective in nature, and therefore it is possible that some patients were not truly infected. We chose to study BSIs to decrease the influence of this factor, since it is rare that patients with positive blood cultures with K. pneumoniae have the lab result disregarded and do not receive antibiotic therapy. This would be more difficult with patients who have cultures from nonsterile sites, such as in pulmonary sources in patients with possible pneumonia. Also, since we studied patients at a single center, it is possible that the results are not generalizable. Since our microbiology laboratory did not report ESBL production in Enterobacteriaceae through the entire period of the study (this practice began in 2008), we chose to group patients by resistance to beta-lactam subclasses instead of proven enzyme production, which may not be as reliable. Also, the definition for CRKP we used was derived from CLSI guidelines that are no longer standard, since the breakpoints for carbapenem susceptibility in Enterobacteriaceae have been lowered. We were unable to assess the isolates for clonal relatedness, although the cases were spread out throughout the years of the study, making a high degree of relatedness unlikely. Finally, we were able to evaluate only hospital records for risk factors and did not have access to other records of preadmission, nonhospital antibiotic exposure, or other factors that are not documented in hospital charts.

In conclusion, the use of antibiotics within 90 days and a length of stay of >40 days were found to predispose patients for BSIs caused by 3GCRKP or CRKP. Central venous catheters predisposed patients for 3GCRKP BSIs. Infections with either phenotype occur in similar patients.