If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Escherichia coli causes more than one-third of the bacteraemia cases in England each year, and the incidence of these infections is increasing.
To determine the underlying risk factors associated with E. coli bacteraemia.
A three-month enhanced sentinel surveillance study involving 35 National Health Service hospitals was undertaken in the winter of 2012/13 to collect risk factor information and further details on the underlying source of infection to augment data already collected by the English national surveillance programme. Antimicrobial susceptibility results for E. coli isolated from blood and urine were also collected.
A total of 1731 cases of E. coli bacteraemia were included. The urogenital tract was the most frequently reported source of infection (51.2% of cases) with previous treatment for a urinary tract infection being the largest independent effect associated with this infection source. Half of all patients had previous healthcare exposure in the month prior to the bacteraemia with antimicrobial therapy and urinary catheterization being reported in one-third and one-fifth of these patients, respectively. Previous healthcare exposure was associated with a higher proportion of antibiotic non-susceptibility in the blood culture isolates (P=0.001).
Analysis of risk factors suggests the potential benefit of community- and hospital-related interventions, especially the better use of urinary catheters and improved antibiotic management of urinary tract infections. As part of the latter strategy, antibiotic resistance profiles need to be closely monitored to ensure that treatment guidelines are up to date to limit inappropriate empiric therapy.
Thus appropriately targeted interventions are required to reduce morbidity and mortality associated with E. coli bacteraemia. Whereas English mandatory surveillance of E. coli bacteraemia (initiated in 2011) allows better estimation of E. coli bacteraemia incidence than was previously possible, detailed epidemiological information is needed to elucidate the reasons behind observed trends.
Following a recommendation from the UK government's Advisory Committee on Antimicrobial Resistance and Healthcare Associated Infection (ARHAI), a sentinel surveillance scheme was initiated to augment existing mandatory surveillance.
The sentinel programme aimed to gather more detailed risk factor information for patients in the hospital and community settings. Specific data collected included antibiotic consumption, use of urinary catheters, indwelling vascular access and other devices, and invasive procedures prior to the bacteraemia. Additionally we aimed to gather detailed information regarding the clinically identified focus or cause of bacteraemia and antibiotic susceptibility data for the E. coli blood and urine cultures.
The sentinel study ran in participating English National Health Service (NHS) Trusts (hospitals under the same management board) over the winter of 2012/13. The study was powered to detect with 95% confidence the prevalence of an underlying infection focus or risk factor with a true frequency of at least 10%, based on estimates of the second most prevalent focus (hepatobiliary) reported via mandatory surveillance.
As the sampling strategy was clustered (by NHS Trust, not by patient), a design effect was included. This equated to a sample size of 2625 E. coli bacteraemia cases; we therefore aimed to recruit 40 Trusts with data collection running over three months. Participating Trusts were selected using simple random sampling. Additionally, two specialist cancer Trusts and six interested Trusts were included post hoc.
Cases from the sentinel study were linked to mandatory E. coli bacteraemia reports (linked using NHS number, a unique personal identifier), to obtain further patient and specimen information, and to voluntary laboratory reports (linked using a combination of available personal identifiers (NHS number, hospital number, date of birth, gender, encrypted surname) to obtain antibiotic susceptibility data for both E. coli blood culture (if not reported by sentinel sites) and urine cultures, one month and one year before the date of the blood culture.
Duplicate entries for the same patient within the same episode (considered as 14 days) were removed from analysis with data relating to the earliest specimen retained. This study focused on E. coli bacteraemia and for cases of polymicrobial bacteraemia only information pertaining to the E. coli isolate was retained.
Data items relating to healthcare exposure prior to the bacteraemia were collected. Specifically, in the previous three days: indwelling vascular access devices (in situ or removed), including the type of intravascular device; in the previous seven days: urinary catheterization (in situ, inserted, removed, or manipulated), including the type of catheter, insertion method and primary indication for catheterization; in the prior four weeks: other devices (in situ or removed), including the type of device and date of insertion; other procedures, including the type and date of procedure; and antimicrobial chemotherapy, including antibiotic name(s), indication, and the treatment area. Patients with more than one healthcare exposure reported or with more than one occurrence of a specific healthcare exposure were included in the study. Previous healthcare exposure in either the community or hospital in the four weeks and one week before bacteraemia was categorized as ‘Yes’ if at least one of the above data items were selected as ‘Yes’ or as ‘No’ if all of the above items were selected as ‘No’; otherwise it was coded as ‘Not known’. In addition to these data items, the primary focus or reason for the bacteraemia was collected. More detail on data collection can be found in Supplementary Appendix A.
Time of bacteraemia onset based on the days between hospital admission and the taking of a positive blood culture was categorized as follows: on or day after admission (a proxy for community-onset infection); two to six days after admission (proxy for early healthcare-onset infection); seven days or more after admission (proxy for late healthcare-onset infection); from a non-admitted patient.
Information on susceptibility of the E. coli blood culture and urine isolates to ciprofloxacin, trimethoprim, co-amoxiclav, third-generation cephalosporins (cefotaxime/ceftazidime), carbapenems (imipenem/meropenem), gentamicin, nitrofurantoin and piperacillin/tazobactam was ascertained using laboratory data reported via the national voluntary surveillance system. The presence of a urine culture in the linked laboratory dataset was taken as a proxy for urinary tract infection (UTI) in the month or year prior to the E. coli bacteraemia. Combined susceptibility for blood cultures, using the above antibiotics, was categorized as: ‘non-susceptible’ if the blood culture was recorded as non-susceptible to at least one of the aforementioned antibiotics; ‘susceptible’ if the blood culture was recorded as susceptible to all of the aforementioned antibiotics.
General associations between two variables were examined using the χ2-test. Multivariate logistic regression was used to estimate the independent effects of previous UTI, previous treatment of a urogenital tract infection and urinary catheterization, on having a urogenital tract focus of bacteraemia, controlled for the effects of age, gender (both of which were considered a-priori confounders, thus automatically included in the model), treatment specialty, and timing of bacteraemia onset. Only the presence of UTI in the month prior to the bacteraemia was used in the model as it was considered more directly relevant than UTI one year previous. A random effects term was used to control for clustering by reporting Trust. Variables that were not significant in the final model were omitted, even if the crude odds ratios showed a significant effect. All data management and analyses were performed using Stata 13.0.
Thirty-five NHS acute Trusts submitted data to the sentinel programme, representing 1731 cases of E. coli bacteraemia. Thirty-seven cases could not be linked to the mandatory surveillance dataset and were excluded from analysis and six cases were removed as within-episode duplicates. We achieved 90% power in our estimate of a risk factor or focus with a frequency of 10%. Distributions of cases from this sentinel study compared to the national mandatory data by patient sex or age and by time of onset or focus of bacteraemia were not significantly different (P>0.05; data not shown).
Description of the study participants
Half of the bacteraemias (N=833; 49.3%) were in patients aged ≥75 years, and around half were in women (N=901; 53.4%) (Table I). More than two-thirds of patients had a positive blood culture taken 0–1 day after admission (N=1153; 68.3%). The underlying infection focus was reported as the ‘urogenital tract’ in 51.2% (N=865) of patients. The next most frequent foci were ‘hepatobiliary’ (15.6%; N=264), and ‘unknown’ (14.9%; N=252).
Table IStudy participants (N=1688)
Patient and specimen details
Age group (years)
0 to <1
Timing of E. coli bacteraemia onset
0–1 day after admission
Underlying focus of infection
Bone and joint infection
Central nervous system
Indwelling intravascular device
Skin/soft tissue infection
Any exposure in the week before bacteraemia
Any exposure in the month before bacteraemia
Antibiotics in the 4 weeks before bacteraemia
Urinary catheter in situ, inserted, removed, manipulated in the 7 days before the bacteraemia
Indwelling vascular access device in situ, or removed in the 3 days before the bacteraemia
Other devices in situ or removed in the 4 weeks before the bacteraemia
Other procedures in the 4 weeks before the bacteraemia
Half (N=930) of the patients had a healthcare exposure, as defined in Methods, in the four weeks before the bacteraemia and one-third (N=584) had a healthcare exposure in the week before the bacteraemia. The percentage of patients with previous healthcare exposure increased as the time of bacteraemia onset after hospital admission increased (Figure 1). Of the patients diagnosed 0–1 day after admission, 46.7% (N=538) had a healthcare exposure in the month before the bacteraemia, increasing to 85.4% (N=251) in patients in hospital for ≥7 days before the bacteraemia, reflecting the increased opportunity of healthcare exposure for admitted patients as well as potentially more comorbidities. Stratification by age showed some variation in prior healthcare exposure four weeks prior to the bacteraemia, with 26.9% (N=7) of patients aged 0–1 years, 54.3% (N=95) of patients aged 1–44 years, 57.5% (N=651) of patients aged 45–84 years, and 49.9% (N=177) of patients aged ≥85 years having had a previous healthcare exposure (details not shown).
Antimicrobial therapy was the most frequently reported prior healthcare exposure in one-third (N=546) of patients (Table I); of these, 58.4% (N=319) received one antibiotic, 23.1% (N=126) received two, and the remaining 101 received three or more. Seventy-nine percent (721/907) of the antibiotic prescriptions were for treatment of infection. Treatment for infection of the urogenital tract was most frequently reported (N=229; 31.8%), followed by treatment of the respiratory tract (N=123; 17.1%). For patients treated for urogenital tract infection, trimethoprim and co-amoxiclav were most frequently prescribed [N=50 (21.8%) and N=47 (20.5%), respectively]. Fourteen percent (125/907) of prescriptions were for medical or surgical prophylaxis; where the site was reported (N=115), 26.1% (N=30) of prescriptions were for a genito-urinary site. Twelve of these patients were catheterized in the three days before the bacteraemia. It was not possible to determine whether these patients were also treated for UTI as treatment and prophylaxis were mutually exclusive options.
Twenty-two percent of patients (N=373) had an indwelling intravascular device that either remained in situ at the onset of the bacteraemia, or had been removed within the three days before the bacteraemia (Table I). Of these patients, 88.2% (N=329) had one indwelling intravascular device and 8.0% (N=30) had two, with the remainder having between three and six. Where reported (N=444), the most frequently used catheter types were peripheral (41.6%; N=184) and central (11.8%; N=52) venous catheters.
Twenty-one percent of patients (N=354) either had a urinary catheter in place at the time of the bacteraemia, or had one inserted, removed, or manipulated in the seven days before the bacteraemia (Table I); of these, 96.9% (N=343) were catheterized just once. Where reported (92.7%; N=328), urinary retention (27.1%; N=89) and fluid balance (21.6%; N=71) were the primary reasons for catheterization. Eleven percent (N=36) of patients had a catheter inserted for incontinence. The reason for catheterization was unknown in 19.2% (N=63) of instances. The primary insertion type was urethral (91.8%; N=302). Long-term (in situ ≥28 days) and short-term (in situ <28 days) catheters predominated (41.3%; N=152; and 38.3%; N=141, respectively); the remaining 20% were reported as temporary catheters. Among the patients whose bacteraemia was detected ≥7 days after hospital admission (N=294), 40.1% (N=118) had been subject to urinary catheterization in the seven days before bacteraemia. The equivalent proportion for patients with bacteraemia detected two to six days after admission was 36.4% (N=47/129) and for those detected on admission it was 15.4% (N=178/1153).
Twelve and seven percent (N=209 and N=123, respectively) of patients had another procedure or device in the four weeks before bacteraemia. Of the available procedure categories, ‘other’ was most frequently selected (64.3%), of which approximately half were surgical procedures.
Patients with an underlying urinary focus of bacteraemia
Six-hundred and ninety (79.8%) of the 865 patients in whom the underlying infection focus was reported as the ‘urogenital tract’ were recorded as having UTI. Among those patients with a urogenital tract focus where the date of infection onset was recorded (N=510), 48.4% (N=248) had the blood culture taken on the day of onset, whereas for a further 214 patients (41.8%) the onset of infection occurred up to seven days before the positive blood culture was taken. Where the urogenital infection was reported as catheter-, procedure-, or device-related (N=171), 84.2% (N=144) were related to a urinary catheter, 12.3% (N=21) to a procedure, and the remainder (3.5%; N=6) to a device. Where information on prior UTIs was reported, two-thirds of patients (62.4% N=176/282) had had at least one previous UTI. Of patients with a urogenital tract focus with at least one antibiotic prescribed in the four weeks prior to bacteraemia, 51.6% (145/281) of antibiotics were prescribed for treatment of urogenital system-associated infection. Where reported, co-amoxiclav (23.1%; N=29/212) and trimethoprim (22.2%; N=47) were most frequently prescribed, whereas 9.9% (N=21) were prescribed nitrofurantoin.
Regression analysis of factors associated with urinary tract focus of infection
The largest independent risk factor for a bacteraemia's underlying focus being the urogenital tract was previous treatment for UTI within four weeks of the bacteraemia onset: [adjusted odds ratio (aOR): 10.7; 95% confidence interval (CI): 6.3–18.1] (Table II). Having had a UTI in the month before bacteraemia increased the odds of urogenital tract-related bacteraemia five-fold (aOR: 5.4; 95% CI: 3.6–8.1). Having a catheter inserted for incontinence (versus ‘other’) and surgical specialty (versus medical specialty) also increased the odds of a urogenital tract focus (aOR: 5.2; 95% CI: 1.5–18.1; and aOR: 4.3; 95% CI: 2.0–9.3, respectively). Several factors were associated with a reduction in the odds of a urogenital tract focus of infection, including male gender, unknown presence of catheter, general specialty, and bacteraemia detected 2–6 days or ≥7 days post admission.
Table IICrude and adjusted odds ratios (ORs) for risk factors for genito-urinary focus of E. coli bacteraemia
Final model includes risk factors identified from the literature, and age and gender as a-priori risk factors. Other variables were only included if they changed the effect of the main risk factors and were associated in the crude analysis. Adjusted ORs are adjusted for all other risk factors in the final model.
Age group (years)
UTI in the month before bacteraemia
Urinary catheterization in the 7 days before bacteraemia
Duration of catheterization
Omitted due to collinearity with ‘Urinary catheterization in the 7 days before the bacteraemia’
Insertion method of urinary catheter
Reason for urinary catheterization
Treatment for UTI in the month before the E. coli bacteraemia
Timing of E. coli bacteraemia onset
0–1 day after admission
CI, confidence interval.
a Final model includes risk factors identified from the literature, and age and gender as a-priori risk factors. Other variables were only included if they changed the effect of the main risk factors and were associated in the crude analysis. Adjusted ORs are adjusted for all other risk factors in the final model.
Where tested, the highest levels of antibiotic non-susceptibility among isolates from blood were to co-amoxiclav (43.0%; N=511/1188) or trimethoprim (40.5%; N=317/783) (Figure 2). Ciprofloxacin non-susceptibility was seen in 17% of tested isolates (N=187/1100). Carbapenem non-susceptibility was only observed in two isolates out of 1060 tested. There was an association between timing of bacteraemia onset and co-amoxiclav (P=0.027) and piperacillin/tazobactam (P=0.008) susceptibilities, both of which showed a greater proportion of non-susceptible isolates from patients with bacteraemia detected two or more days after hospital admission (Table III). Ciprofloxacin (P=0.02) and trimethoprim (P<0.001) non-susceptibility were associated with different foci of infection with a greater percentage of isolates non-susceptible to ciprofloxacin found in patients with ‘pneumonia’ reported as the underlying infection focus (23.7%; 9/38), whereas trimethoprim non-susceptibility was most frequently observed in patients with an underlying urogenital tract focus (48.1%; 198/412) (Table IV). There was an association between antibiotic non-susceptibility of blood culture isolates and healthcare exposure in the four weeks before the bacteraemia (P=0.001), with 60.9% (N=406/667) of isolates from patients with previous healthcare exposure showing non-susceptibility to at least one of the antibiotics tested versus 50.7% (N=165/330) of isolates without previous healthcare exposure (data not shown). Antibiotic non-susceptibility was also associated with antibiotic exposure in the four weeks before bacteraemia (P<0.001), the rates of non-susceptibility being 66.5% (N=262/394) and 51.4% (N=244/475), in those with and without previous antibiotic exposure respectively (data not shown).
Table IIIAntibiotic susceptibilities of the E. coli blood culture by timing of bacteraemia onset in relation to hospital admission
Timing of E. coli bacteraemia onset in relation to hospital admission
Urine cultures with antibiotic susceptibilities one year and four weeks before the E. coli bacteraemia were identified for 340 (20.1%) and 230 (13.6%) patients, respectively. The highest levels of non-susceptibility in isolates from urine at both time-points were for trimethoprim (47.7%; N=162/340) at one year and 46.3% (N=106/229) at four weeks (not shown). Co-amoxiclav non-susceptibility was around 30% at both time-points. Ciprofloxacin non-susceptibility was seen in 20.4% (N=47/231) and 15.5% (N=23/148) of urinary isolates in the year and four weeks before the bacteraemia. The levels of non-susceptibility to third-generation cephalosporins at the same time-points were 21.3% (29/136) and 13.8% (12/87), respectively. Non-susceptibility to piperacillin/tazobactam was 22.1% (28/127) at one year and 24.4% (20/82) at four weeks before the bacteraemia. Non-susceptibility to nitrofurantoin was low, being reported in 6.9% (22/321) of isolates at one year and 4.6% (10/216) at one month. Only one patient's infection was caused by E. coli non-susceptible to carbapenems at both time-points.
Where antibiograms were available for both urine and blood isolates from the same patient, there was a significant association between non-susceptibility in urine and the subsequent blood culture for each of the antibiotics examined, the association being stronger for isolates taken up to four weeks before the bacteraemia compared with the association at one year (data not shown).
Whereas large declines in infections such as meticillin-resistant Staphylococcus aureus bacteraemia occurred concomitant with healthcare-based interventions (e.g. intravascular device-related care bundles; screening/decolonization of high-risk patients), E. coli bacteraemia is frequently considered a community-associated infection with lesser scope for reducing incidence.
All potential areas for reducing E. coli bacteraemia incidence do, however, need to be investigated given the high burden of this infection (37,273 cases reported for 2015), as even small reductions in incidence could equate to thousands of patients per year.
The data presented here highlight potential interventions for reducing E. coli bacteraemia incidence, such as improved urinary catheter care and UTI diagnosis and management.
Several of our findings suggest that treatment failure in UTIs is an important risk factor for the development of E. coli bacteraemia. Hence prompt diagnosis and appropriate treatment of UTIs – the most frequent underyling focus of E. coli bacteraemia identified here and in the literature – with antibiotics to which the organism is susceptible are key in limiting progression from UTI to bacteraemia and severe sepsis.
Antibiotic therapy in the four weeks before bacteraemia was the most frequently reported healthcare exposure (one-third of patients); of these patients, almost one-third were prescribed antibiotics for treatment of a genito-urinary infection. Notably the most frequently prescribed antibiotics for these patients were trimethoprim and co-amoxiclav, where non-susceptibility in urine isolates was around 47% and 30%, respectively. Furthermore, trimethoprim non-susceptibility was 40.5% in E. coli blood isolates. This may reflect previous trimethoprim exposure in the treatment of the patient's UTI with subsequent selection of resistant strains leading to treatment failure and progression to bacteraemia, or UTI caused by uropathogenic strains already resistant to trimethoprim. These findings are concerning in relation to trimethoprim, as this antibiotic typically dominated first-line treatment recommendations for uncomplicated UTI in primary care, and thus is widely prescribed, as supported by our findings. However, it is reassuring that recent guidelines advocate nitrofurantoin therapy with trimethoprim use, dependent on local resistance patterns.
This reiterates that guidelines for empiric UTI therapy need constant review and updating. Whereas nitrofurantoin non-susceptibility was lower at around 5–7%, trends need to be monitored given its increasing importance in empiric UTI therapy. Non-susceptibility to third-generation cephalosporins in urine was higher than typically seen in E. coli blood isolates, especially in patients with a history of UTIs within a year before bacteraemia; these patients may represent UTI treatment failures or those carrying multidrug-resistant, but not especially virulent, isolates. Furthermore, the association between resistance in urine before bacteraemia is consistent with a recent meta-analysis showing that prior antibiotic exposure increased the odds of antibiotic non-susceptibility.
Whereas ciprofloxacin, third-generation cephalosporins, gentamicin, and carbapenems are less frequent first-line therapies for UTIs, they are important for treating more severe infections; thus non-susceptibility trends should be monitored.
Previous healthcare exposure, regardless of whether it was one week or month before the bacteraemia case, primarily equated to antibiotic prescribing and was perhaps an important factor for subsequent non-susceptible E. coli bacteraemia. This highlights patient groups with E. coli bacteraemia who may be more likely to have a non-susceptible infection and justifies the current focus on antibiotic prescribing and antimicrobial resistance. We noted a general trend for increased non-susceptibility with healthcare-onset bacteraemia cases; however, this was only significant for co-amoxiclav and piperacillin/tazobactam. This is most likely due to these being widely used first-line therapy for patients presenting at hospital with an infection and where these infections subsequently progress to bacteraemia for which resistant isolates will be selected. A larger study may also have identified significant associations for other antibiotics.
Regression analysis somewhat predictably identified previous urogenital tract infection treatment and having a UTI in the month before the bacteraemia as the most important risk factors for the development of a (presumed) urogenital tract focus bacteraemia, increasing the odds of a urogenital tract focus 10.7- and 5.4-fold, respectively. We hypothesize that patients with urine samples sent for antimicrobial testing would more likely have experienced treatment failure or had a complicated UTI, as most uncomplicated UTIs are typically treated empirically in primary care.
Furthermore, three-quarters of patients with a urogenital tract focus had their bacteraemia detected on admission. This implies either a failure in diagnosis and treatment of UTI in the community or patients presenting directly to the hospital with bacteraemia who have not visited their primary care physician for treatment of their symptoms. Greater awareness of the patient groups at risk of UTIs developing into bacteraemia in the community and hospital could reduce treatment failure or unrecognized complicated UTIs progressing to bacteraemia, through, for example, enhanced monitoring of patients with suspected UTI and prompt intervention when empirical treatment fails. Mid-stream urine sampling of all patients with symptomatic UTI, or where otherwise indicated in the guidance for susceptibility screening, would also allow appropriate antibiotic therapy to be prescribed.
Further studies are warranted to gain a better understanding of who these patients are and what the potential intervention opportunities may be. Near-patient testing for antimicrobial resistance may be a future option to improve management. Furthermore, reduction of UTI incidence would limit the largest underlying focus of E. coli bacteraemia; studies are required to understand how this can be achieved.
Urinary catheter use was also identified as an important risk factor. We found that 21% of patients had a urinary catheter inserted, removed, or manipulated in the week before bacteraemia; 144 reported bacteraemias were likely related to a urinary catheter in the study. Urinary catheters inserted for incontinence were associated with a 5.2-fold increase in the risk of a urinary focus of infection. Whereas incontinence and the other main reasons for catheterization (urinary retention and fluid balance) could be considered appropriate indications, without more detailed patient medical history it is not possible to determine whether each catheter was appropriately used.
Additionally, it is concerning that for one in five catheterized patients the indication for catheterization was not known/recorded, indicating a lack of optimized patient management, possibly increasing the risk of complications. Furthermore, almost half of the catheterized patients had a long-term catheter. Although this may have been appropriately indicated, it nonetheless increases the risk of infection compared to short-term catheters, as the risk of infection developing from catheters increases with catheterization duration.
We identified that the percentage of patients with a urinary catheter increased from 15.4% in those diagnosed around admission to 40.1% in those diagnosed ≥7 days after admission. This identifies the importance of appropriate catheter care in the community as well as close monitoring in hospitalized patients. The latter represent a patient group where device-related hospital-based interventions could be targeted to reduce E. coli bacteraemia incidence, although it is worth noting that the present data are unable to identify which patients may be catheterized as part of the sepsis pathway for monitoring of urinary output.
Urinary catheter use is associated with an increased risk of complications, notably catheter-associated UTI.
Since the urinary tract is the predominant underlying focus of E. coli bacteraemia, appropriate catheter use and management are obvious interventions, particularly when the catheter is used solely for incontinence where non-invasive treatment options exist. As with the management of UTIs, any such intervention must be targeted at the community as well as hospital setting, as the data indicate that many catheterized patients reside in the former.
These sentinel data are representative of the national picture of E. coli bacteraemia, based on comparisons of age, gender, timing of bacteraemia onset, and underlying focus of infection. Despite a smaller sample size than intended, the power of the sentinel study was 90% to detect a risk factor or focus with a frequency of 10%. There are, however, several limitations to the present study. Our study has only collected information on patients with E. coli bacteraemia, so it is not possible to compare the population with and without infection, nor to estimate the impact of interventions on reducing E. coli bacteraemia incidence. The susceptibility results relate to antimicrobials tested which were not necessarily those received by the patient. Thus, it is not possible to determine the effects of antimicrobial exposure for the treatment of UTI on subsequent non-susceptibility in blood cultures. Additionally antimicrobial testing practices vary by laboratory, and some antibiotics are less frequently tested, which may impact on the linked urine antimicrobial results. Robust information on the management and treatment of previous UTIs and catheter use would provide further resolution on specific risk factors.
It is clear from this study that E. coli bacteraemia is often secondary to an earlier UTI. Therefore, prevention of UTIs, especially in the elderly, may reduce bacteraemia developing. We have highlighted potential interventions and further research to reduce the incidence of E. coli bacteraemia in England: (i) close monitoring and effective treatment of patients with suspected UTI; (ii) awareness of local antibiotic resistance profiles with early recognition of UTI treatment failure with prompt initiation of effective antimicrobial therapy; (iii) early identification of suspected bacteraemia secondary to UTI, primarily in the community; (iv) appropriate use and management of urinary catheters in the hospital and community with monitoring for signs of infection; (v) further research on UTI incidence in England, infection rates in suprapubic versus long-term catheters, and antibiotic consumption patterns.
Whereas interventions targeted at the community setting may be harder to implement and monitor than hospital-based interventions, this should not be a reason for not trying, given the high burden of E. coli bacteraemia.
The E. coli bacteraemia sentinel surveillance group comprised the following NHS acute Trusts and collaborators whom we thank for participating in the study: Aintree University Hospitals, Dr R.A. Sen; Barts & the London, Dr A. Mifsud; Buckinghamshire Healthcare, Dr J. O'Driscoll; Cambridge University Hospitals, Dr N. Brown, Cheryl Trundle; County Durham & Darlington, Dr D. Allison; Croydon Health Services, Dr M. Twagira; Derby Hospitals, Dr Gnanarajah; East & North Hertfordshire, Dr F. Awad-El Kariem; East Cheshire, Dr R. Rajendran; East Sussex Healthcare, Dr S. Umashankar; Gateshead Health, Dr G. Horne; Homerton University Hospital, Dr A. Claxton; Lancashire Teaching Hospitals, Dr J. Cheesbrough; Leeds Teaching Hospitals, Dr A. Kirby; Luton & Dunstable Hospital, Dr R. Mulla; Mid Essex Hospital Services, Dr L. Teare; Newham University Hospital, Dr C. Rosmarin; North West London Hospitals, Dr G. Gopal Rao; Northern Devon Healthcare, Dr D. Richards; Nottingham University Hospitals, Dr T. Boswell; Oxford University Hospitals, Dr I. Bowler, L. O'Connor; Plymouth Hospitals, Dr P. Jenks; Portsmouth Hospitals, Dr S. Wyllie; Royal Berkshire, Dr N. Virgincar; Royal Free Hampstead, Dr S. Hopkins; South London Healthcare, Dr M. Dallantonia; South Tyneside, Dr A. Rodgers, Dr R. Ellis; Southport & Ormskirk Hospital, Dr J. Bowley, M. Kiernan; Surrey & Sussex Healthcare, Dr K. Knox; The Royal Marsden, Dr U. Riley; The Whittington Hospital, Dr M. Kelsey; University College London Hospitals, Dr P. Wilson, Dr N. Shetty; University Hospital of North Staffordshire, Dr J. Orendi; University Hospitals of Morecambe Bay, Dr M. Pasztor.
We would also like to acknowledge: the UK government's Advisory Committee on Antimicrobial Resistance and Healthcare Associated Infection E. coli sub-group (M. Kiernan, Dr C. McNulty, Professor P. Hawkey, Professor H. Loveday, Dr C. Pellowe, Dr R. Hope, A. Moore, Dr S. Bruce, C. Williams, Z. Head, Dr R. Milton, J. Wilson) for their support for the sentinel study; D. Ironmonger for technical support in the development of the software; and S. Wasti and A. Falola for assisting in recruitment of the sentinel sites.
Appendix A. Supplementary data
The following is the supplementary data related to this article:
M.H.W. has received: consulting fees from Actelion , Astellas , Astra-Zeneca , Bayer , Cerexa , Cubist , Durata , The Medicines Company , MedImmune , Merck , Motif Biosciences , Nabriva , Optimer , Paratek , Pfizer , Roche , Sanofi-Pasteur , Seres , Summit , and Synthetic Biologics ; lecture fees from Abbott , Alere , Astellas , Astra-Zeneca , Pfizer & Roche ; and grant support from Abbott , Actelion , Astellas , bioMérieux , Cubist , Da Volterra , Merck , Paratek , Pfizer , Sanofi-Pasteur and Summit . S.H. is affiliated with the National Institute for Health Research Health Protection Research Units (NIHR HPRU) in Healthcare Associated Infection and Antimicrobial Resistance at Imperial College London and University of Oxford in partnership with Public Health England (PHE). The other authors have nothing to disclose. The views expressed in this article are those of the authors and not necessarily shared by their organizations or affiliates.
This work was funded in its entirety by Public Health England .
Trends among pathogens reported as causing bacteraemia in England, 2004–2008.