Epidemiology of Invasive Escherichia coli Infection and Antibiotic Resistance Status Among Patients Treated in US Hospitals: 2009–2016 Free

Abstract

Background

Published data is limited on the prevalence and risk of recurrence of extraintestinal invasive Escherichia coli infections (IEIs) in the United States.

Methods

The analysis included all inpatient and hospital-based outpatient visits occurring between 2009 and 2016 at hospitals with continuous microbiology data submission to the Premier Healthcare Database for 90 days before and 12 months after the admission or visit. IEI was defined as having positive E. coli culture from a normally sterile site (eg, blood, cerebrospinal fluid). The prevalence of IEI, 12-month risk of recurrent IEI, and antibiotic resistance were assessed.

Results

Overall, 144 944 725 hospital visits among 37 207 510 patients were analyzed, and 71 909 IEI events occurred in 67 583 patients, corresponding to an IEI prevalence of 0.50 events per 1000 visits and 1.82 events per 1000 patients. Recurrence was common: 26.9 per 1000 patients had a recurrent IEI in the 12 months after their infection. Most infections were community acquired (66.4%), and urosepsis was most common clinical syndrome (66.0%). The 30-day risk of IEI among patients undergoing transrectal ultrasound–guided prostate biopsy was high: 5.03 events per 1000 patients. Among all IEI cases with antibiotic susceptibility testing, 9.18% were resistant to extended-spectrum cephalosporins, 28.22% to fluoroquinolones, and 0.14% to carbapenems. Resistance to extended-spectrum cephalosporins increased from 5.46% to 12.97% during the 8-year study period.

Conclusions

This real-world study indicates a substantial burden of IEI and recurrent IEI exists in the United States, as well as increasing resistance to extended-spectrum cephalosporins. Future research should explore risk factors of recurrent IEI aiming to effectively prevent such infections.

As the most common gram-negative bacterial pathogen, extraintestinal pathogenic Escherichia coli causes a wide range of diseases affecting all age groups [1–5]. E. coli is also the leading cause of bloodstream infections (BSIs), nosocomial infections, and infectious complications after transrectal ultrasound–guided prostate biopsy (TRUS-PB), resulting in substantial morbidity and mortality rates worldwide [6–15]. Extraintestinal infections due to E. coli are associated with substantial economic burden and loss of productivity [3, 5, 16]. The total cost attributable to excess hospital stays due to antibiotic-resistant E. coli infections in Europe was estimated to be 18.1 million Euros in 2007 [8].

Emergence of multidrug-resistant E. coli subtypes, such as extended-spectrum β-lactamase–producing E. coli, poses major challenges for infection prevention and control in hospital settings [11, 13, 17–22]. According to the 2011−2014 National Healthcare Safety Network report [15], 15% of all reported US healthcare-associated infections were due to E. coli, and there was an increase in the percentage of antibiotic resistance among E. coli pathogens, especially related to fluoroquinolone resistance compared with previous years. Inappropriate empirical antibiotic use among patients with antibiotic-resistant E. coli infections have resulted in worse clinical outcomes [23, 24].

Although E. coli infections have been studied extensively globally, there is a limited data available on the overall disease burden of extraintestinal invasive E. coli infections (IEIs) in the United States [6]. The Centers for Disease Control and Prevention collects information on hospital-acquired IEIs (HA-IEIs), but no national data on community-acquired and healthcare-related community-acquired IEIs (HCA-IEIS) have been reported. The current study sought to address this gap, using real-world microbiology laboratory data from a large administrative hospital database. The main objectives of this observational study were to (1) estimate the prevalence of culture-confirmed extraintestinal IEIs, (2) describe the characteristics of patients with such infections, (3) assess antibiotic resistance status, and (4) explore the risk of recurrent IEI.

METHODS

Data Source

Data for this retrospective observational study were extracted from Premier Healthcare Database (PHD), which at the time of analysis contained deidentified data from over 760 million patient encounters, or 1 in every 4 hospital visits in the United States [25]. The PHD is a hospital-based all-payer database containing data from standard hospital discharge files for both inpatient and hospital-based outpatient visits from geographically diverse US hospitals. This repository includes patient demographics and disease states, information on all billed services (eg, medications, laboratory tests performed, diagnostics, and therapeutic services), and hospital characteristics. Patients can be tracked with a unique identifier within a single hospital system that shares the same billing structure. A subset of hospitals (about 25%) contribute microbiology laboratory data to the PHD. These hospitals submit all available data from clinical microbiology testing at their hospitals to PHD regardless of testing results. The data include specimen source, tests performed, and all observations for those specimens. Protocols for antimicrobial susceptibility testing are specific to each individual hospital.

Study Population

This analysis included all inpatient and hospital-based outpatient visits from 1 January 2009 to 31 December 2016 at PHD hospitals with continuous microbiology data submission during the study period. During the study period, there were 561 166 033 unique inpatient and outpatient visits/discharges at PHD hospitals among 147 288 241 patients, among which 144 944 725 (25.8%) visits/discharges and 37 207 510 patients (25.3%) had continuous microbiology data during the 90 days before the index admission date and 12 months after the index discharge date and were included in the analytic data set for this study (Supplementary Figure 1).

Study Variables

The main outcome variables included count and prevalence of IEI events, risk of IEI recurrence, and antibiotic resistance pattern of related IEI isolates. IEI was defined as the presence of positive (including presumptively positive) E. coli culture from blood (42.1%), cerebrospinal fluid (0.07%), or other normally sterile sites, including sterile body fluids (eg, pleural, peritoneal, or synovial fluid; 22.5%), bronchoalveolar lavage fluid (0. 63%), and internal body sites (eg, bone, organ, deep wound, and lymph node biopsies/cultures; 27.2%). Positive E. coli cultures obtained from the same patient were within 14 days were considered a single IEI event.

IEI was categorized into 10 clinical syndromes based on bacterial culture source and visit diagnosis codes, including urosepsis, intra-abdominal infection, meningitis, complicated pneumonia, wound infection, post–prostate biopsy–related (TRUS-PB vs other type) infection, neonatal or infant sepsis, neutropenic fever, other BSI, and other IEI. E. coli urosepsis required a positive blood culture with either a positive urine culture or specified International Classification of Diseases codes indicating an associated genitourinary infection. Definitions of specific types of IEIs and associated diagnosis codes from the International Classification of Diseases (Ninth Revision or Tenth Revision) are listed in Supplementary Table A. Recurrent IEI was defined as an IEI event that occurred 14 days to 12 months after discharge from an initial IEI-related hospital visit.

Antibiotic resistance was defined as having shown resistance to any of the drugs in each of the 3 antibiotic classes assessed in this study: (1) extended-spectrum cephalosporins (ESC4; cefepime, cefotaxime, ceftazidime, and ceftriaxone); (2) fluoroquinolones (ciprofloxacin, levofloxacin, and moxifloxacin; (3) carbapenems (imipenem, meropenem, doripenem, and ertapenem). E. coli infections were categorized as HA-IEI, HCA-IEI, or non–healthcare-related community-acquired IEI (NHCA-IEI).

Patient characteristics assessed included sex, age, year of discharge, race, admission type, and discharge status. Overall health status was determined by All Patient Refined Diagnosis Related Group (APR-DRG) severity of illness level [26], intensive care unit (ICU) admission and length of stay, and Charlson Comorbidity Index (CCI) and CCI comorbid conditions [27]. Hospital characteristics included bed capacity, teaching status, population served, and US census geographic region.

Statistical Analysis

Descriptive statistical analyses were performed on study variables. Continuous data were expressed as mean, standard deviation (SD), median, and interquartile range (IQR). Categorical variables were expressed as frequencies and percentages. The frequency distribution of specific patient characteristics and CCI comorbid conditions were estimated for the overall IEI sample and by infection type. Prevalence and length of ICU stay were also reported. Prevalence of IEI and antibiotic resistance were estimated overall and by hospital setting, year of discharge, age at discharge, and infection type. Antibiotic resistance was also assessed separately among patients with HA-IEI to facilitate comparisons to published report.

The 30-day risk of IEI among patients with prostate biopsy procedure was estimated by type of procedure (ie, TRUS-PB and other types). The 12-month risk of any recurrent IEI was estimated at case and patient levels for all patients with IEI during index visit and by year of index discharge, age, and infection type. The 12-month risk of recurrent urosepsis due to E. coli was estimated for both overall IEI sample and urosepsis sample. The time to first recurrent IEI after the index admission was estimated by infection type. A hazard function plot was created using PROC LIFETEST and Epanechnikov kernel methods to show the probability of recurring urosepsis on a particular day and its 95% confidence intervals for at-risk patients for both overall

The frequency distribution of source of IEI (ie, HA-IEI, HCA- IEI, and NHCA-IEI) was estimated for overall IEI sample and by infection type. The statistical significance level was set at .05. All tests were 2 sided. The Cochran-Armitage test was used for trend analysis, and all analyses were performed using SAS software (version 9.4).

RESULTS

Study Population

In this analysis, 144 944 725 unique hospital visits (10% inpatient; 90% outpatient) and 37 207 510 unique patients from 236 hospitals were included. Of these, 71 909 visits (0.19%) among 67 583 patients (0.18%) met the IEI definition. Among patients with IEI identified, 58 168 (86%) were inpatients and 9415 (14%) outpatients (Table 1). Among all patients with IEI, 39.8% were male; 55.6% were >65 years of age; 68.2% were white, and 15.3% were black. Among inpatients, 81.3% were admitted through the emergency department; 60.5% were discharged home and 26.5% to skilled nursing, rehabilitation facility, or another hospital; and 8.4% died during the index admission (n = 4900).

More than 75% of inpatients were classified with major or extreme APR-DRG severity of illness. Nearly one-third of inpatients had an ICU admission; the median length of stay was 4 days (IQR, 2–6). The mean CCI (SD) was 4.07 (3.51). Urosepsis was the most common IEI, accounting for 66.0% of IEI cases. Other BSIs (22.1%) ranked second, followed by wound infection (6.8%), complicated pneumonia (2.8%), intra-abdominal infection (1.8%), neonatal or infant sepsis (1.7%), neutropenic fever (1.6%), other IEI (0.8%), and meningitis (0.2%).

Characteristics of the 44 529 patients urosepsis resembled those of the overall sample, with a similar percentage having an ICU admission (29.9%) but with a shorter median ICU stay (3.5 days; IQR, 2–5 days). Compared with patients with urosepsis overall, a higher percentage of patients with recurrent urosepsis were male (41.4% vs 30.7%, respectively), ≥65 years of age (62.8% vs 57.6%), discharged to skilled-nursing or rehabilitation facilities (33.8% vs 22.3%), and in major and extreme APR-DRG severity categories (84.0% vs 74.2%) (all P < .05).

Prevalence of IEI

IEI prevalence was estimated to be 0.50 events/1000 Premier facility visits and 1.82 patients with IEI per 1000 Premier facility patients from the overall study sample (Table 2). A higher proportion of inpatients than outpatients experienced an IEI event (10.01 vs 0.30/1000 patients, respectively). At both event and patient levels, prevalence among adults increased with age, ranging from 0.23/1000 among patients aged 18–49 years to 1.3/1000 for those aged >85 years. Among children, infants had higher prevalence than those 2–17 years old.

IEI Among Patients With Prostate Biopsy Procedures

A total of 17 833 patients underwent prostate biopsy procedures during the study period, with 17 686 undergoing TRUS-PB and 147 having another procedure type. The 30-day IEI risk was 5.03/1000 patients undergoing TRUS-PB and 6.80/1000 patients undergoing other procedures.

Risk of Recurrent IEI

Overall, 26.9 patients per 1000 patients with IEI had another event in the following 12 months (Table 3). Some patients had ≥1 repeat recurrences, resulting in 30.2 recurrent IEI events per 1000 patients with IEI (no temporal trends in recurrent risk observed). Recurrence risk varied between different IEI types from 136.4 events per 1000 patients with IEI in patients with post–TRUS-PB infection to 4.3 events per 1000 patients after neonatal/infant sepsis. The 12-month risk of recurrent urosepsis was 16.7/1000 patients with any IEI at the index visit and 21.3/1000 patients with E. coli urosepsis at the index visit.

Among 1764 patients with recurrent IEI (Table 4), the mean time to first recurrence (SD) was 117.2 days (96.1) after the index discharge date, with a median (IQR) of 81.5 (140) days. The median (IQR) time to recurrent IEI varied: 86 (147) days for urosepsis, 82 (195) days for complicated pneumonia, and 74 (140) days for other BSI, with a low of 23.5 (29) days for post–prostate biopsy–related infection.

Supplementary Figure 2 shows the hazard rate of recurrent urosepsis among overall patients with IEI and patients with urosepsis. The highest recurrent risk occurred approximately 30 days after index discharge and then decreased significantly thereafter.

Antibiotic Susceptibility Status

Of 71 909 IEI visits, 41 696 (57.8%) had antibiotic susceptibility test results available in PHD. Of these, 9.18% of cases were ESC4 resistant, 28.22% fluoroquinolone resistant, and 0.14% carbapenem resistant (Table 5). ESC4 antibiotic resistance increased during the study from 5.46% (2009) to 12.97% (2016; trend P < .05), but fluoroquinolone resistance did not increase. ESC4 resistance occurred in 20.68% of complicated pneumonia cases, in 9.09% of urosepsis visits, and in 13.04% of TRUS-PB visits. Fluoroquinolone resistance was most common among TRUS-PB–related IEIs (78.26%) and complicated pneumonia events (51.55%) (Table 5).

Source of Infection

Among all hospital visits with IEI (inpatient and outpatient), 8.7% were considered HA-IEI, 24.9% HCA-IEI, and 66.4% NHCA-IEI. Most urosepsis cases (71.6%) were community acquired, with only 4.1% hospital acquired. For intra-abdominal infections, wound infections, meningitis, complicated pneumonia, neonatal/infant sepsis, and neutropenic fever, between 20% and 40% of cases were hospital acquired (Table 6).

DISCUSSION

Using microbiology laboratory data from 236 geographically diverse hospitals across the United States, this study identified 71 909 IEI events in the study population corresponding with a high burden of IEI: 1.8 per 1000 Premier patients experienced IEI during the study period and 10 per 1000 inpatients experienced IEI. These invasive events were associated with a high acuity of care; 86% involved an inpatient stay, and among those, 33% had an ICU stay and 8% died. This high burden and acuity indicate that additional strategies for prevention are needed for IEI.

Urosepsis was the most common IEI clinical syndrome and accounted for two-thirds of all IEI cases followed by other BSIs and wound infections. This preponderance of invasive infections originating from urinary tract is consistent with research involving BSIs, where about half of E. coli bacteremia is from a genitourinary source [6, 7, 29].

The high proportion of IEI events involving BSIs (from urine or other sources) is consistent with E. coli being a leading cause of BSIs in developed countries, with incidence estimates ranging from 28 to 74 per 100 000 population; [6, 7, 30–33] a recent metanalysis calculated a pooled estimate of 48 per 100 000 person years, comprising 11%–27% of events in a 2020 systematic review [6]. The prevalence of IEIs seen in our study (1.8 per 1000 patients, equivalent to 180 per 100 000) is higher, likely because of the inclusion of all IEIs, not just BSIs, and because our denominator was not the total population but rather people with ≥1 outpatient visit in the study period. Events associated with biliary disease and intra-abdominal infections are another important contributor to IEI, comprising 11%–27% and 4%–48% of E. coli bacteremia events in the systematic review [6]. In the current analysis, these events were captured in the following IEI categories: “other BSI” (22.1% of IEI events), “intrabdominal infection”/peritonitis (1.8%), or “other IEI” (0.8%). Future research should focus on developing and validating case definitions for this subset of IEI, as well as further characterizing these groups by underlying diagnosis (eg, hepatic cirrhosis).

As mentioned, IEI events had a high level of acuity with one-third of inpatient stays involving an ICU stay (median duration 4 days). Most admissions (80%) were on an emergency basis, 75% had a major or extreme APR-DRG severity of illness level, and only 60% of ended with a discharge to home. Overall, 8% of inpatient IEI events were associated with in hospital death. This high rate of mortality has been seen in studies of E. coli BSI—7-day mortality rates were 5% and 11% in multicenter studies from Australia and England, respectively [30, 32].

Clinicians working in inpatient settings are familiar with the high frequency of recurrence of urosepsis—especially among the elderly—but formal studies of the frequency of this recurrence are few. More than one-fifth of patients with urosepsis had recurrent urosepsis during the 12 months after index discharge. The risk of urosepsis recurrence peaked at 30 days after index discharge. To our knowledge, this is the first study reporting the risk of recurrent urosepsis. The high recurrence risk during the 12 months after index urosepsis provides a smaller population to potentially focus intervention efforts to reduce the risk of such potentially deadly reinfections among these patients.

Clinicians working in this area are aware that an important cause of recurrent of IEI is structural abnormalities of the urinary tract or biliary tract and related indwelling medical devices. In these patients, IEI events can be associated operative manipulation (eg, urologic procedures for urosepsis and biliary procedures for sepsis arising from the biliary tract). To better characterize these recurrent infections, future investigation should determine the proportion of recurrence patients who had such structural abnormalities and/or indwelling medical devices, and determine whether events were associated with related surgical procedures.

Resistance to fluoroquinolones was common among all types of infections especially among patients with TRUS-PB procedures (78.3%). The level of resistance to fluoroquinolones remained stable during the study period, which is different from what was reported by the Centers for Disease Control and Prevention report on an increase in fluoroquinolone resistance among HA-IEIs during 2011–2014 [15]. However, we did find a slight increase in fluoroquinolone resistance among patients with HA-IEI from 2012 to 2014. Resistance to extended-spectrum cephalosporins was less common, but the percentage of patients with ESC4 resistance more than doubled during the study period. Resistance to carbapenem remained low (0.14%) among all types of IEI, which is consistent with what was reported for E. coli bacteremia infections in England [7].

The level of resistance to fluoroquinolones among patients undergoing TRUS-PB was comparable to levels in prior reports from individual hospitals [29, 30]. In a large multicenter Veterans Affairs study, hospitals with more quinolone resistance had higher rates of post–TRUS-PB E. coli BSI because many patients receive quinolones for prophylaxis at the time of the procedure and these infections are breaking through the prophylaxis and thus more likely to be resistant.

In terms of source of infection, findings from this study indicated that NHCA-IEI accounted for two-thirds of all IEI cases, HCA-IEI accounted for a quarter of all cases, and only 8.7% of cases were hospital acquired. In the report on E. coli bacteremia from England, 76.1% of cases were classified as community onset, which is comparable to the current study findings [7].

The study findings has at least the following 3 limitations. First, outcome estimates are based on hospitals contributing microbiology data and may not be representative of the entire PHD or other hospitals in the United States. Nonetheless, this study was at the population level, spanning 8 years and representing tens of thousands of inpatients and outpatients with IEI diagnoses confirmed with positive cultures from microbiology laboratories. Moreover, the information gleaned about prevalence, recurrence, and antibiotic resistance addresses existing knowledge gaps about extraintestinal IEI in the United States. Second, PHD only captures readmissions or outpatient visits to the same hospitals where the index visit occurs, which may have resulted in an underestimation of recurrent E. coli infection. Finally, owing to the lack of detailed clinical information and specific specimen source for some patients, we could not determine the infection type for 0.04% of patients with IEI, which did not affect overall prevalence estimates but may slightly affect the prevalence of specific types of IEI.

In conclusion, IEI affected 1.82/1000 patients in the overall study sample, and the prevalence was higher among inpatients and outpatients >65 years of age. Urosepsis and other BSIs were the most common type of infections, and recurrent IEI was common. The highest risk of recurrent infection occurred about 30 days after the index discharge date, and the majority of IEIs were community acquired. Resistance to fluoroquinolones was most common, but resistance to ESC4 increased during the study period. More research is needed to find effective ways to prevent these potentially deadly infections.

Supplementary Data

Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Financial support. This work was supported by Pfizer.

Potential conflicts of interest. E. B., A. G., and S. P. L. are employees and stockholders of Pfizer. N. A. R. and A. K. are employees of Premier Healthcare Solutions. R. G. K. D. reports no potential conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

References