Risk factors and mortality outcomes of extended-spectrum beta-lactamase producing Escherichia coli bacteremia: A retrospective cohort study from two Indonesian referral hospitals

Introduction

Bacteremia or the presence of bacteria in blood can develop into bloodstream infection which, if left untreated, can lead to death from sepsis or septic shock.^1–3 It is very difficult to find bacteria in blood unless in severe cases and it may not correlate with clinical features at the time of infection.^2,4 In several studies, bacteremia was most often caused by gram-negative bacteria with Escherichia coli (E. coli) being the most common etiology with an incidence rate of 50 to 60 cases per 100,000 population.^5–7

Based on the World Health Organization (WHO) global priority pathogens list of antibiotic-resistant bacteria (https://www.who.int/news/), extended-spectrum beta-lactamase (ESBL)-producing Enterobacteriaceae is included in the first priority (critical) group and ESBL-producing E. coli is the most reported.⁸ ESBL is an enzyme produced by certain bacteria capable of hydrolyzing broad-spectrum beta-lactam antibiotics such as penicillin, first-, second-, and third-generation cephalosporins (ceftriaxone, cefotaxime, and ceftazidime), and monobactams (aztreonam), causing multidrug resistance (MDR).^8,9

In Indonesia, the prevalence of ESBL from various isolates is fairly high at 60–85%.^10,11 The “One Health” surveillance research in public health, agriculture, and animal and livestock sectors conducted by the Indonesian Ministry of Health Research and Development found that the proportion of ESBL-producing E. coli bacteremia reached 57.7%.¹² Meanwhile, the mortality rate of E. coli bacteremia in several studies ranged from 18.2–30.6%.^13–15 Most bacteremic patients develop sepsis and the excessive use of antibiotics can cause antibiotic resistance, therefore prolonging the length of hospital stay which may in turn lead to increased mortality.^13,14,16

Several studies have identified several risk factors for ESBL-producing E. coli and associated mortality.^16–19 Demographic factors, antibiotic resistance patterns, sources of infection, and disease severity in each hospital are deemed necessary to be evaluated as risk factors for ESBL-producing E. coli and associated mortality that clinicians concern themselves with these factors in the management of patients with infections caused by antibiotic-resistant bacteria.

Methods

An observational retrospective cohort study was conducted using the medical records of all inpatients from January to December 2019 at two referral hospitals in Jakarta, Indonesia, namely Cipto Mangunkusumo National Hospital and Persahabatan Central General Hospital. This study was approved by the Ethics Commission of The National Institute of Health Research and Development (LB.02.01/2/KE.184/2018) on 8 May 2018, the Health Research Ethics Committee of the Faculty of Medicine, Universitas Indonesia-Cipto Mangunkusumo National Hospital (KET-169/UN2.F1/ETIK/PPM.00.02/2022) on 21 February 2022, and the Health Research Ethics Committee of Persahabatan Central General Hospital (42/KEPK-RSUPP/07/2018) on 31 July 2018.

The target population was patients with positive E. coli blood cultures which were taken once during hospitalization in all age groups. Patients who were treated for less than 48 hours, had negative E. coli blood cultures or with incomplete medical record data were excluded. Positive E. coli blood cultures were phenotypically differentiated into ESBL- and non-ESBL-producing colonies using the VITEK®2 compact system algorithm (bioMérieux, France). Patient characteristics, length of stay, source of infection, comorbidity, and several risk factor variables for ESBL-producing E. coli infection including age, history of antibiotic therapy, utilization of medical devices, Charlson Comorbidity Index (CCI), and history of hospitalization were extracted from medical records. Further analysis was conducted for patients with ESBL-producing E. coli bacteremia. An effort was made to avoid missing data bias by excluding patients with incomplete records and differential measurement bias was addressed by uniformizing confirmation tests between all patients.

Elaboration on the process of subject inclusion is available in the Underlying data (Supplementary Figure 1).²⁰ The sample size was determined by calculating the formula for comparing the difference between two independent proportions. A total of 118 subjects were determined to be needed for adequate hypothesis testing. The details are available in the Underlying data (Supplementary Figure 2).²⁰

The basic and clinical characteristics of the research subjects are presented in Table 1 and as underlying data.²⁰ Categorical data are presented as proportions. The data are presented as means and standard deviations if they were normally distributed and median and minimum-and-maximum values if not. Risk factors for acquiring ESBL-producing E. coli bacteremia and risk factors for mortality in patients with such conditions were analyzed. Bivariate analysis using Chi-square or Fisher’s exact test was performed to examine the significance of the odds ratio (OR). A multivariate logistic regression test was subsequently performed to identify significantly contributing factors for variables with a bivariate analysis p value of less than 0.25. Statistical analyses were conducted with IBM SPSS Statistics version 20 (RRID:SCR_002865).

Results

A total of 116 subjects were analyzed in this study. The basic characteristics of patients with E. coli bacteremia are shown in Table 1. The most common source of infection was the gastrointestinal and intra-abdominal tract. More than half of the patients had a history of hospitalization and a history of antibiotic therapy, mostly with cephalosporins. Most patients were treated with empirical antibiotic therapy with the same drug group. Malignancy was the most frequent comorbidity. Most were found to have CCI scores of greater than two. Medical devices were commonly utilized among patients with urinary catheters and nasogastric tubes being the most common. Most had sepsis and nearly half was lethal.

Bivariate analysis showed significant statistical differences (p <0.05) in crude OR for ESBL-producing E. coli bacteremia between age, utilization of medical devices, CCI scores, and history of antibiotic therapy. Variables with p-values of <0.25 were included for multivariate logistic regression. The analysis yielded statistically significant risk from utilization of medical devices and history of antibiotic therapy. Specifically, it showed that history of antibiotic therapy and the use of medical devices increases the risk of developing ESBL-producing E. coli bacteremia by 2.78 and 3.21 times respectively (Table 2).

The proportion of mortality between patients with ESBL- and non-ESBL-producing E. coli bacteremia was 55.7% and 38.2% respectively (p = 0.059). Patients infected by the ESBL-producing strain were further analyzed for mortality risk factors (Table 3). Out of all the variables, only sepsis produced a statistically significant difference in mortality (p = 0.011) in the bivariate analysis. Furthermore, multivariate logistic regression analysis of the eligible variables (age, length of stay, utilization of mechanical ventilator, empirical antibiotic therapy, and sepsis) showed that only age (OR 15.0; 95% CI 1.54 to 146.02; p = 0.020) and sepsis (OR 6.5; 95% CI 1.91 to 22.16; p = 0.003) produced statistically significant differences in mortality.

Discussion

In this study, the proportion of patients with ESBL-producing E. coli bacteremia was 52.58%. This result was higher compared to some other studies.^15,16,18 Both ESBL and non-ESBL-producing E. coli bacteremia can occur in various age groups.^21,22 Epidemiological studies reported the proportion of E. coli bacteremia ranging from 18–55% in all age groups.²³ A 2017 study by Nivesvivat et al. in a Thai tertiary care hospital found that the prevalence of ESBL-producing Enterobacteriaceae bacteremia in children has tended to increase since 2010.²⁴ Bacteremia by ESBL-producing Enterobacteriaceae significantly associated with higher mortality than its non-ESBL-producing counterpart. The study found that this condition was significantly more prevalent in children with comorbidities such as chronic illnesses.²⁴ The median age of subjects in our study was 48 years (range 0–89 years) with males being the predominant gender (52.6%). Around 18.9% of patients were less than 18 years old, which ranged from neonates to adolescents. Our findings were similar to previous studies that described the age distribution of bacteremic patients which ranged from pediatric to elderly patients.^7,22,25

According to the multivariate analysis, patients aged 18 years or older were more than twice as likely to experience ESBL-producing E. coli bacteremia though this was not statistically significant. Neonates and the elderly are groups susceptible to infection. Children tend to experience more severe forms of bacteremia due to their immature immune systems. Neonates can be exposed to bacteria by vertical transmission during pregnancy or from the mucosa of the birth canal, especially in the presence of bacterial colonization. Furthermore, E. coli strains found in children aged less than three months were reported to be genotypically more virulent than those found in adults.²⁶ Meanwhile, diminished immune responses along with the presence of other risk factors in geriatric patients increase the risk of MDR infection.

The most common sources of infection were the gastrointestinal and intra-abdominal tracts (37.1%). No significant difference was observed in the proportion of infection sources between the ESBL and non-ESBL groups. Many other studies also found gastrointestinal tract as a source of infection.^19,27,28 E. coli is a gram-negative bacteria found in the normal intestinal flora of healthy individuals which may become pathogens if present in an excessive amount.^29,30 Transmission of E. coli can occur by direct contact or fecal-oral through contaminated food and/or water. Hence, some types of E. coli can cause diarrhea, gastrointestinal infections, urinary tract infections, respiratory tract infections, and other diseases.^22,29 Therefore, the source of the infection site for E. coli bacteremia can come from various organ systems. The most common sources of infection related to mortality from ESBL-producing E. coli bacteremia is from the gastrointestinal and respiratory tracts. Different results were reported in several other studies, the urinary tract was the most common source of infection associated with mortality in E. coli ESBL bacteremia.^13,18,31

In this study, history of antibiotic therapy was found in as many as 63.8% of patients with E. coli bacteremia and 77.1% ESBL-producing E. coli bacteremia. Previous studies reported a history of antibiotic therapy in 40 to 80% of ESBL-producing E. coli bacteremia.^17–19 Additionally, the multivariate analysis showed that a history of antibiotic therapy was a significantly associated risk factor for ESBL-producing E. coli bacteremia. Other studies have also demonstrated the same results.^16–18,28 History of antibiotic therapy was reported to be a significant risk factor for ESBL-producing bacteremia in adults but not in children.^18,19,21,32 Furthermore, the most widely used type of antibiotic is the beta-lactam group, especially cephalosporins, which are widely used as an empirical antibiotic due to their broad spectrum of activity against gram-positive and negative bacteria and even some resistant strains.³³ Inadequate empirical antibiotic therapy during treatment in the intensive care unit (ICU) or in the regular ward may lead to colonization of ESBL-producing bacteria and potentially lead to antibiotic resistance.³⁴

Many studies have reported the use of medical devices as a risk factor for MDR bacteremia.^16,18,31,35 In this study, utilization of medical devices was also significant as a risk factor for ESBL-producing E. coli bacteremia. Medical devices are thought to be a mediator for the entry of bacteria into the blood which can then progress to bacteremia. Trauma to the skin and mucosal surfaces causes epithelial antigen-presenting cells (such as dendritic cells or Langerhans cells) to capture bacterial antigens, they then migrate to lymph nodes, and are presented to T-lymphocytes.^1,2

The Charlson Comorbidity Index is commonly used for the assessment of risk factors and predictors of prognosis.³⁶ Although not a risk factor, we found a statistically significant difference was observed in the OR between ESBL and non-ESBL-producing E. coli bacteremia patients with CCI scores of greater than two. Nguyen et al. and Kaya et al. also reported that comorbidity was not proven as a risk factor for ESBL-producing E. coli bacteremia.^16,27 In this study, malignancy was the most common comorbidity. Patients with malignancies are susceptible to infection by ESBL-producing E. coli, a condition which may increase mortality in such a population.³⁷ Patients are generally immunosuppressed due to the underlying disease and the treatment of the disease itself.³⁸ The use of medical devices and invasive procedures further puts patients with malignancy in higher risk of MDR infection.^3,22,23 The innate and adaptive immune systems play a major role in microbial clearance along with the spleen and liver which filter bacteria entering blood circulation. The introduction of bacteria to the bloodstream may resolve spontaneously or develop into a bloodstream infection that leads to sepsis.^1–3 Coexistence of bacteremia with malignancy may result in a vicious cycle as it renders patients with malignancy in a constant immunosuppressive and hyperinflammatory state, which consequently affects the innate and acquired immune systems even in the recovery period.³⁸

We also found that history of hospitalization was not a statistically significant risk factor for ESBL-producing E. coli bacteremia. However, different results were reported by Xiao et al. who found that a history of hospitalization in the past three months doubled the risk of developing ESBL-producing E. coli bacteremia.²⁸ A study conducted by Su et al. reported that elderly patients with a history of hospitalization were more than three times as likely to be infected by ESBL-producing E. coli or K. pneumoniae bacteremia even though they had been discharged more than 360 days ago.³⁹ The use of urinary catheters is cited as one of the most common sources of infection and medium for bacterial colonization. Utilization of such medical devices is often associated with a history of hospitalization.^29,30,37 Moreover, patients with a history of hospitalization are at a higher risk of bacterial colonization and subsequent MDR bacteremia due to exposure from MDR bacteria at the ICU or regular ward and previous administration of antibiotics. The aforementioned factors along with the fact that our study was conducted at two referral hospitals predominated by patients with malignancies who have had a history of multiple hospitalizations (and consequently have already been exposed to MDR bacteria several times) may provide a plausible explanation for our findings.

There was no significant difference in the proportion of mortality between subjects with ESBL-producing and non-ESBL-producing bacteremia. Sianipar et al. previously reported no significant difference in the proportion of mortality between ESBL- (30.6%) and non-ESBL-producing (22.2%) E. coli bacteremia.¹⁴ Kang et al. reported a mortality proportion of 25.6% in ESBL-producing E. coli bacteremia patients with no meaningful difference between the groups for gender and age.¹⁵ We also found that administration of empirical antibiotic therapy did not have a significant difference in the incidence of mortality. Research by To et al. demonstrated that administration of carbapenem and non-carbapenem empirical antibiotic therapy had no significant effect on 30-day mortality.³¹ Appropriate and compatible definitive antibiotic therapy is required in bloodstream infections if proven to be caused by MDR bacteria. Delay in antibiotic administration is associated with increased morbidity and mortality.⁴⁰ Therefore, prompt identification of the offending bacteria in bloodstream infections is crucial.

The multivariate analysis indicated that age and the presence of sepsis are significant risks for mortality in patients with ESBL-producing E. coli bacteremia. Rodriguez et al. also reported that patients with sepsis due to ESBL-producing E. coli are over three times and over four times more likely to experience 30-day and 14-day mortality respectively.¹⁸ Other factors such as length of stay, use of mechanical ventilators, and empirical antibiotic therapy were not found to be statistically significant, but these factors are often associated with other factors including history of antibiotic therapy and history of hospitalization. The effect of the presence of ESBL-producing strains of E. coli on bacteremia mortality outcomes is still controversial. Several studies have found an association between infection by ESBL-producing E. coli and mortality while another study reported contradictory results showing no difference in mortality between the two patient groups.⁷

Across the numerous studies that have been conducted to assess the risk factors for ESBL-producing E. coli bacteremia, there exist differences in demography, bacterial patterns, and causes of bacteremia in each hospital. This study, which is part of the “One Health” study conducted by the Centre for Biomedical Research and Development and Basic Health Technology of the Indonesian Ministry of Health, provides a well-detailed investigation of risk factors for ESBL-producing E. coli bacteremia in two referral hospitals in Indonesia.¹²

Several limitations are present in this study. First, the retrospective nature of this study renders it to rely on secondary data obtained from the paper medical records of the two hospitals previously mentioned. Therefore, the occurrence of missing data was unavoidable and perfect representation of the data would be difficult to achieve. Second, several factors such as an onset of bacteremia before positive culture results, risk factors for mortality such as SOFA score, duration of empirical antibiotic therapy, and the type and duration of definitive antibiotic therapy were not accounted for in this study. Third, the ESBL-producing strains were only identified phenotypically and not genotypically. Nevertheless, the results of this study can be applied clinically and are also useful in stratifying risk factors for bacteremia caused by ESBL-producing E. coli. Clinicians should stay vigilant for these risk factors to limit mortality and morbidity from ESBL-producing E. coli.

Conclusions

To summarize, history of antibiotic therapy and utilization of medical devices were significant risk factors for ESBL-producing E. coli bacteremia. The proportion of mortality in ESBL-producing E. coli bacteremia patients was 55.7% with age and sepsis being its independent risk factors. Some of the other risk factors were found to not be significantly associated with the incidence of ESBL-producing E. coli infection. Nevertheless, it is still necessary for clinicians to identify these risk factors to stratify patients at risk of infection by MDR bacteria. Hence, the identification of risk factors for MDR bacterial infection remains an important issue as their current treatment options are mostly limited to carbapenems.⁴⁰

Consent

Written informed consent for publication of the patients’ details was obtained from the patients and guardians of the patients.