Carbapenem-Resistant Klebsiella pneumoniae Outbreak in a Neonatal Intensive Care Unit: Risk Factors for Mortality

Abstract

Aim

The aim of our study was to determine the factors associated with mortality in neonates with carbapenem-resistant Klebsiella pneumoniae (CRKP).

Material and methods

This retrospective, single-center study was conducted in the Neonatal Intensive Care Unit of Harran University Faculty of Medicine between January 2017 and July 2018 who had CRKP growth in their blood, urine or cerebrospinal fluid cultures. The discharged group was designated as the control group (Group 1), whereas the group that faced mortality was classified as the case group (Group 2). The demographic data, clinical findings and laboratory and microbiological results of the two groups were compared to identify risk factors.

Results

A total of 58 patients (36 in Group 1 and 22 in Group 2) exhibited CRKP growth during the study period. Low birth weight (p = 0.039), previous antifungal (p = 0.002) or amikacin use (p = 0.040), congenital anomalies (p = 0.002), total parenteral nutrition (TPN) administration (p = 0.002), surgery (p = 0.035), thrombocytopenia (p = 0.007), low platelet mass index (p = 0.011), elevated C-reactive protein (p = 0.004), high carbapenem minimum inhibitory concentration (MIC) (p = 0.029) and high amikacin MIC (p = 0.019) were associated with mortality. In a multivariate regression analysis, previous antifungal use (p = 0.028), congenital anomalies (p = 0.032) and TPN use (p = 0.013) were independent factors in predicting mortality.

Conclusion

Previous antifungal use, congenital anomalies and TPN use were found to be independent risk factors for mortality in neonates with CRKP infection.

INTRODUCTION

Klebsiella pneumoniae is a Gram-negative, lactose-producing, immobile, aerobic and rod-shaped bacterium from the Enterobacteriaceae family that causes nosocomial urinary tract infections, pneumonia, septicemia and soft tissue infections. Low birthweight, prematurity, previous surgical procedure, presence of intravascular catheter, parenteral nutrition, intubation, tracheostomy, rupture of the amniotic sac more than 24 h before birth, prolonged hospitalization, underlying metabolic disease, neutropenia and previous antibiotic use, including carbapenems, glycopeptides, aminoglycosides and ß-lactam antibiotics have all been reported as important risk factors of K. pneumoniae infection in newborns [1–4]. Studies have demonstrated that rate of mortality due to Gram-negative pathogens in Neonatal Intensive Care Units (NICUs) ranges between 10% and 67.7% and increases in the presence of resistance to carbapenem, bacteremia, inappropriate/delayed empirical antibiotic use and underlying diseases [4–8].

For diagnosis, it is essential to reproduce the causative agent in culture, and early diagnosis and appropriate antibiotic treatment can be lifesaving. Although carbapenems are an effective group of antibiotics against bacilli that produce Gram-negative beta-lactamase, these bacteria have developed resistance to it due to the widespread use of broad-spectrum antibiotics, leading to worldwide outbreaks of carbapenem-resistant infections [9]. One of the important resistance mechanisms developed by these bacteria is plasmid-mediated carbapenemase production, whereas the other one is the absence or decreased expression of outer membrane proteins [10]. While K. pneumoniae carbapenemases and New Delhi metallo-beta-lactamase (NDM) are effective on Class A and Class B carbapenemases, respectively, NDM is more common in the pediatric group [11]. Carbapenem-resistant Klebsiella pneumoniae (CRKP) infections have also been reported to exhibit 98% fluoroquinolone and 50% aminoglycoside (amikacin and gentamicin) resistance [12, 13]. Tigecycline and colistin are antibiotic therapies which can be successfully administered to patients that include preterm infants with very low birthweight [12, 14–16].

In our study, CRKP patients admitted to the NICU within an 18-month period were evaluated retrospectively to determine the factors associated with mortality.

MATERIALS AND METHODS

Study design and data collection

This retrospective, single-center study was performed in the NICU of Harran University Faculty of Medicine (a tertiary intensive care unit with a capacity of 65 incubators) in Sanliurfa, Turkey. Patients hospitalized between January 2017 and July 2018 who had CRKP growth in their blood, urine or cerebrospinal fluid (CSF) cultures were included in the study. Patients with congenital anomalies incompatible with survival were excluded from the study. The study was approved by the local ethics committee.

The following information of the patients were retrieved from their files and evaluated: demographic data (gestational age, sex, delivery method, birthweight, antenatal diagnosis, maternal age, number of pregnancies and consanguineous marriage); interventions in the delivery room; Apgar scores; underlying diseases; surgical procedures; mechanical ventilation support; presence and duration of a central catheter; total parenteral nutritional support; blood, urine or CSF culture results and antibiotic susceptibility; antibiotics used prior to growth; inotropic support needed during the treatment process; time to initiation of sensitive antibiotics; duration of hospitalization; biochemical tests (renal and liver functions and serum electrolytes); hematological results; and medication use. Two groups were established from the patients diagnosed with CRKP: those who were discharged (Group 1; control group) and those who faced mortality (Group 2; case group). Patients were compared in terms of these variables to determine the risk factors for mortality.

Definitions

Standard definitions for nosocomial infections and clinical sepsis were used according to the Center for Disease Control and Prevention [17]. Multidrug resistance was defined as resistance to at least three different antibacterial groups against Gram-negative pathogens. The hematological and biochemical values at the onset of infection were taken into consideration. Thrombocytopenia was defined as the platelet count <150 × 10³/µl. Serum levels of C-reactive protein (CRP) >1 mg/dl were defined as elevated. Platelet mass index (PMI) was calculated from the blood samples obtained in cases of suspected sepsis before beginning antibiotic therapy and was defined as the product of platelet count and mean platelet volume (MPV) (PMI = platelet count × MPV/10³) fl nl⁻¹.

Microbiologic methods and antibiotic policy of our NICU

BD BACTEC ™ FX40 (Becton-Dickinson, USA) automated blood culture system was used to isolate bacteria from blood. Samples of tracheal aspirates, urine and CSF were inoculated in routine culture media (sheep blood agar, chocolate agar and eosin-methylene blue agar plate). VITEK 2 automated systems were used for the identification of Gram-negative bacilli and the determination of antimicrobial susceptibility (bioMerieux, Marcy-l'Étoile, France). Anti-microbial susceptibility to colistin was investigated using the approved disc-diffusion method, based on the guidelines of the Clinical Laboratory Standards Institute. The unit empiric antibiotic treatment policy for suspected early-onset sepsis was penicillin and aminoglycoside. We used vancomycin or teicoplanin and cefatoxime for suspected late-onset sepsis as empirical antibiotic therapy for both groups. Carbapenems were preferred if the clinical condition of the infant was deteriorating.

Statistical analysis

Number Cruncher Statistical System 2007 Statistical Software (UT, USA) was used for statistical analysis. The authors worked in NICU of Harran University Faculty of Medicine only during the study period, and the number of patients followed-up in NICU has increased after the authors started working at the said hospital; therefore, the study was performed without sample size calculation. Descriptive statistical methods (mean, standard deviation, median, frequency, and ratio) were used to evaluate study data. In addition, the Kolmogorov–Smirnov test and box plots were used to evaluate the normal distribution of data. The Student’s t-test was used for comparison of variables with normal distribution, whereas the Mann–Whitney U-test was used for comparison of variables with non-normal distribution. Chi-square test and Fisher’s exact test were used to compare the qualitative data. Logistic regression analysis was used for multivariate evaluations of risk factors that could have an effect on mortality. The results were evaluated with 95% confidence interval (CI) and statistical significance was defined as p < 0.05.

RESULTS

The total number of patients admitted to the NICU between January 2017 and July 2018 was 1704. Of the 70 patients (4.1% of all inpatients) who exhibited CRKP growth during the study period, data were available for only 58. There were 36 (62.0%) patients in Group 1, and 22 (37.9%) patients in Group 2. While there was no significant difference in terms of gestational age between the groups, mean birth weight was significantly lower in Group 2 (1859 ± 843 g) compared with Group 1 (2368 ± 917 g) (p = 0.039). There were no significant differences in the number of small gestational age (birth weight < 10th percentile) infants, extremely low birth weight (ELBW) (birth weight < 1000 g) infants, and premature infants (gestational age < 37 weeks) between the groups. The median duration of hospital stay in Group 1 was 49.5 (16–268) days, and the median age at the time of death in Group 2 was 50 (3–78) days. Surgical operation rate and total parenteral nutrition (TPN) were used significantly higher in the mortality group (p = 0.035 and p = 0.002, respectively), whereas there was no significant difference in ventilation and in the presence and duration of a central catheter. In Group 2, there was a high probability of congenital anomaly (p = 0.002); whereas the probability of transient tachypnea of the newborn was significantly higher in the discharge group (p = 0.037) (Table 1). There was also no significant difference in terms of the time to initiation of susceptible antibiotics and the postnatal age at which growth was detected. CRKP was isolated from blood cultures of 39 (67.2%) patients and urine cultures of 17 (29.3%) patients. None of the neonates had growth in CSF or in multiple samples. CRKP ratio in blood culture was 86.4% in Group 2 and 55.6% in Group 1, and the difference was significant (p = 0.015). Aminoglycoside and antifungal use prior to growth was significantly higher in Group 2. Platelet count (p = 0.030) and PMI (p = 0.011) were significantly lower in the mortality group. The rate of thrombocytopenia and the rate of elevated CRP were significantly higher in Group 2 (p = 0.007 and p = 0.004, respectively) (Table 2). While mean antibiotic minimum inhibitory concentration (MIC) values for carbapenem and amikacin were significantly higher in the mortality group (p = 0.029 and p = 0.019, respectively), there were no differences in mean MICs with regards to the other antibiotics (Table 3).

Due to the small sample size of the study, only seven important risk factors affecting mortality were included in the logistic regression analysis (Table 4). The evaluation of the effects of aminoglycosides use and antifungal use, surgery, elevated CRP, congenital anomalies, thrombocytopenia and use of TPN on mortality using logistic regression analysis revealed that the model was significant (F = 35.527; p < 0.01) and the model’s descriptive coefficient (86.2%) was excellent. After using a logistic regression model, previous antifungal drug use (p = 0.028), congenital anomalies (p = 0.032) and use of TPN (p = 0.013) were determined as independent risk factors for mortality in newborns with CRKP infection. The odds ratio for the effect of antifungal use was 13.25 (95% CI 1.32–132.31), for the effect of congenital anomalies was 11.47 (95% CI 1.23–106.94), and for the effect of use of TPN was 36.71 (95% CI 2.14–628.39). While the effects of aminoglycoside use, surgery, elevated CRP and thrombocytopenia variables were significant in the univariate evaluation, they were not significant within the model in the multivariate evaluation (p > 0.05) (Table 4).

DISCUSSION

Our study demonstrated that previous antifungal use, congenital anomalies and use of TPN were independent risk factors for mortality due to CRKP infection in newborn infants. Previous studies evaluated the outcomes of multidrug-resistant Gram-negative bacilli bacteremia in NICU patients [18] and risk factors and outcomes associated with Gram-negative carbapenem-resistant bacteremia in neonates [5], risk factors for mortality in neonates with K. pneumoniae infection [6] and risk factors for CRKP in neonates and children [3]. To the best of our knowledge, this is the first study that have evaluated the risk factors for mortality only in neonates with CRKP infection to date. Unlike other studies, we only included neonates who produced CRKP in their cultures in the study.

Nosocomial infections contribute an important amount in neonatal deaths, with K. pneumoniae, other Gram-negative rods (Escherichia coli, Pseudomonas spp, Acinetobacter species) and Staphylococcus aureus often being identified as the causative agents [19]. CRKP was detected in the cultures of 70 of 1704 patients admitted to the NICU during the study period. The CRKP infection rate in our unit constituted 4.1% of the NICU hospitalizations, which is similar to the ratio reported by Saleem et al. [6] (3.7%). Two studies discovered a mortality rate that ranged from 10% to 23% due to carbapenem-sensitive Gram-negative sepsis and from 33% to 40% due to carbapenem-resistant Gram-negative sepsis in NICU [5, 7]. Saleem et al. [6] reported that mortality rate was 16.3% for late-onset K. pneumoniae sepsis in newborn patients. We only included neonates with progressive CRKP in our study. Consistent with two studies performed in neonates, we found that the mortality rate was 37.9% [5, 7]. On the basis of these studies [5–7] and our study, we considered that the risk of mortality in newborn patients with CRKP sepsis was higher compared to those with carbapenem-sensitive K. pneumoniae sepsis.

Some studies performed in newborn infants reported several risk factors for mortality due to Gram-negative sepsis. Saleem et al. [6] reported that male gender, ELBW, thrombocytopenia and failure to achieve microbiological clearance were significantly associated with mortality due to late-onset K. pneumoniae sepsis in neonates. Similar to our study, the rate of mechanical ventilation was higher in the mortality group in their study; however, it was not significant in the final model. Nonetheless, unlike our work, a study in children with CRKP infection demonstrated that mechanical ventilation was an independent risk factor of mortality [20]. Although the rate of mechanical ventilation was higher in mortality group in our study, it was not significant. This may be attributed to small sample size of the study. We illustrated that thrombocytopenia was a dependent risk factor for mortality which is consistent with the study of Saleem et al. [6]. Tsai et al. [18] revealed that the presence of infectious complications after bacteremia and underlying secondary pulmonary hypertension were independent risk factors for overall mortality in neonates with multidrug-resistant Gram-negative bacilli bacteremia. Although they found that congenital anomalies were significantly risk factor for mortality, on a multivariate analysis, it did not have significance. However, we illustrated that congenital anomalies were independent risk factors for mortality. Nour et al. [5] stated that male gender and existence of infectious complications were independent risk factors associated with mortality due to carbapenem-resistant Gram-negative late-onset sepsis in neonates. They established that previous antifungal use was a dependent risk factor for mortality, whereas it was an independent risk factor in our study. In contrast to our study, a study showed that use of TPN was not associated with mortality due to Klebsiella pneumonia sepsis in neonates [6]. Tsai et al. [18], on the other hand, stated that the use of TPN caused a significant increase in the risk of mortality due to multidrug-resistant Gram-negative bacteremia in newborn infants, however, they revealed this significant increase disappeared in the regression analysis. Consistent with Tsai et al.’s study [18], we showed that use of TPN was independent risk factor for mortality due to CRKP infection in neonates.

In neonatal sepsis, thrombocytopenia occurs as an inflammatory response to increased breakdown of platelets and their insufficient production, which results in the release of immature platelets into the circulation. MPV reflects platelet function, which may increase in neonatal sepsis [21]. Low PMI has been identified as a marker for transient tachypnea of the newborn, bronchopulmonary dysplasia, intraventricular hemorrhage, necrotizing enterocolitis and retinopathy of prematurity [22, 23]. A study revealed that preterm infants with sepsis had lower PMI compared to those without sepsis [22]. In the present study, PMI was found as a dependent risk factor for mortality in neonates with CRKP infection. We considered that a decreased PMI was due to thrombocytopenia, as the MPV value did not change significantly between the groups.

Imipenem or meropenem, polymyxin and tigecycline are used as monotherapy or combination therapy in CRKP infections [24]. In a meta-analysis of studies conducted on adults, it has been shown that combination therapy with at least two agents reduces mortality from 38.7% to 27.4% compared with therapies with a single agent [25]. A study performed in 2016 reported that a combined therapy with meropenem-containing regimens seems to be the best option in severely ill CRKP children [26]. In a current study on neonates, where nearly all participants received combination therapy, it was determined that combination therapy with fosfomycin was as effective as combination therapy with meropenem/panipenem and that fosfomycin played an important role in the treatment of CRKP infections [1]. Most patients of CRKP present extremely high (>32 μg/ml) carbapenem MICs. Therefore, the benefit of incorporating carbapenem in a combination regimen depends on the MIC value of the infectious agent toward carbapenem. Thus, high-dose carbapenem-combined regimens could be effective in cases with relatively low or moderately elevated carbapenem MIC, whereas it could not be beneficial for extremely high carbapenem MICs [27]. Consistent with our study, Zhang et al. [20] revealed that MIC of meropenem >8 μg/ml was associated with higher mortality compared with MIC ≤1 μg/ml in children with CRKP bloodstream infection. We administered a combination of colistin and meropenem to our patients with CRKP. Based on our study and previous studies [20, 27], since patients with high carbapenem MIC value (>8–16 μg/ml) had increased risk of mortality and extremely high (>32 μg/ml) carbapenem MICs could not benefit from carbapenem, a different antibiotic choice may be suggested for combination regimen for these patients instead of carbapenem.

K. pneumoniae is present in the normal gastrointestinal and vaginal flora. The first measures that were reviewed with the Infection Control Committee of our hospital were the washing of hands, disinfection and sterilization of equipment with suitable agents, regular maintenance of invasive catheters, cleaning of drug and formula preparation areas and isolation of the patients with CRKP infections. Although the specific stages in which there were shortcomings could not be identified precisely, a significant decrease was observed in the rate of infections caused by all agents as a result of these basic measures.

Our study had several limitations. First, it was a retrospective study. For this reason, molecular analysis for carbapenemase typing in CRKP strains could not be performed as it was not part of the clinical routine. In addition, due to the retrospective nature of this study, some findings that may be closely associated with CRKP infection may have been overlooked. Second, we could not determine the stage at which infection control measures were inadequate. Finally, due to the single-center nature of the study and a small sample size, the results should be generalized with caution.

In conclusion, we found that Gram-negative bacteria are an important cause of neonatal sepsis in Turkey. Due to the ever-increasing antibiotic resistance, it is essential to know the risk factors, associated with mortality, to take the necessary preventive measures against these factors, and to initiate early and correct antibiotic therapy so as to reduce morbidity and mortality in the future. It is also important to pay attention to infection control measures, particularly hand hygiene, and perform close clinical monitoring for indicators of infection in neonates in high-risk groups, especially those with low birth weight, surgery, congenital anomalies, thrombocytopenia, TPN use, previous aminoglycosides or antifungal use, to avoid unnecessary or prolonged antibiotic therapy and shorten the duration of hospitalization.

ACKNOWLEDGMENTS

We thank Fadile Yildiz Zeyrek, PhD and Mehmet Bayraktar PhD, Department of Microbiology, Harran University Medical Faculty, Sanliurfa, Turkey for their support in this report.

REFERENCES