Background

Expert consensus and stewardship interventions emphasize treatment of bacteriuria based on urinary symptoms [1,2,3,4]. However, clinicians often diagnose and treat elderly patients for urinary tract infection (UTI) when they have only non-specific symptoms such as delirium based on the belief that elderly patients with UTI may present without localizing symptoms [5,6,7]. This raises uncertainty as to what constitutes symptoms of UTI, what is asymptomatic bacteriuria, and whether it warrants treatment. Hereafter, we use the term asymptomatic bacteriuria to signify bacteriuria without any urinary symptoms.

Bacteriuria with bacteremia is a true infection requiring treatment, so it can be used to guide diagnostic criteria for UTI. Diagnostic criteria for UTI should capture all bacteremic UTI (bUTI), because it is associated with a higher mortality rate [8, 9]. Prior studies suggested that elderly patients with bUTI often do not have urinary symptoms [10,11,12], which is recognized by Centre of Disease Control as “asymptomatic bacteremic urinary tract infection” [13]. However, prior studies do not provide alternative clinical features to reliably capture bUTI [10,11,12].

Approximately 20–30% of patients presenting to emergency department with febrile or complicated UTI or pyelonephritis have bacteremia [14,15,16]. Escherichia coli is the most important pathogen for UTI and accounts for over 70% of all cases [17, 18]. In patients admitted to hospital with E coli bacteriuria who had a blood culture done, approximately 15% of patients had E coli bacteremia [19].

We conducted a study on E. coli bUTI patients to characterize the proportion of and risk factors for bUTI without urinary symptoms. We also aimed to find clinical features that would be sensitive enough to capture bUTI cases without urinary symptoms.

Methods

Study design

We conducted a retrospective cohort study at 4 acute care academic and community hospitals in the Greater Toronto Area. Research ethics board approval was obtained from each institution.

The study included consecutive adult patients admitted to the hospital for community acquired E. coli bacteremia from January 1, 2015 to December 31, 2016. Community acquired bacteremia was defined by positive blood culture collected at admission or within 48 h of hospital admission. E. coli bacteremia was defined as at least 1 positive blood culture for E. coli. Patients were excluded if they had an unclear infectious focus. This group was likely composed of patients with urinary source as well as patients with non-urinary source. Analysis and comparison to this mixed patient population would be difficult to interpret.

Data collection

Data were obtained from electronic and paper medical records at each hospital site and entered into a standardized case report form. Data on demographics, comorbidities, clinical presentation, investigations, microbiological data, investigations, surgical interventions, antibiotic therapy and clinical outcomes were collected. A second auditor performed sample reliability checks on 10% of the population.

Variable definitions

Comorbidities were entered as per Charlson comorbidity index [20].

Urinary source (i.e. bUTI) required a urine culture with significant monomicrobial growth of E. coli > =10 × 10^6 CFU/L and any of the following criteria:

  1. 1)

    Clinical urinary tract infection as per diagnostic criteria for treatment by Loeb et al. [21]. Patients were diagnosed with UTI based on dysuria or > =2 of the following: fever, urgency, flank pain, urinary incontinence, shaking chills, frequency, gross hematuria or suprapubic pain [21]. Patients with urinary catheter were diagnosed with UTI based on > = 1 of the following: new costovertebral tenderness, rigors, new onset delirium or fever [21].

  2. 2)

    Imaging findings suggestive of pyelonephritis including perinephric stranding or hydronephrosis with flank pain or costovertebral angle (CVA) tenderness

  3. 3)

    Recent urologic procedure (including ureteric stenting, cystoscopy, prostate biopsy) with no other clear source

  4. 4)

    Monomicrobial growth of E. coli in blood and urine culture with the same susceptibility pattern and no other obvious source on clinical assessment

Biliary source was defined as any of the following:

  1. 1)

    Evidence of cholecystitis, choledocholithiasis, or cholangitis on imaging

  2. 2)

    Known cholecystostomy tube or biliary malignancy with no other obvious source

  3. 3)

    Recent manipulation of biliary tree including ERCP with no other obvious source

Intra-abdominal source was defined as any of the following:

  1. 1)

    Evidence of intra-abdominal abscess, appendicitis, diverticulitis, pancreatitis or mass on abdominal imaging

  2. 2)

    Known intra-abdominal drain with no other obvious source

  3. 3)

    Recent intra-abdominal surgery with no other obvious source

The above criteria were also checked for agreement with the main responsible physician’s diagnosis for accuracy. Other infectious foci including pneumonia were based on the clinician’s final diagnosis.

Urinary symptoms and signs were based on the clinician’s documentation on presentation. Urinary symptoms included dysuria, urinary urgency, urinary frequency, gross hematuria, flank pain, suprapubic pain and urinary retention. Urinary signs included suprapubic tenderness and costovertebral or flank tenderness. Hereafter, bUTI without urinary symptoms refer to bUTI without any of the aforementioned urinary symptoms or signs.

Patients were screened regularly by nurses and assessed by physicians for delirium based on the Confusion Assessment Method (CAM) criteria [22]. Fever was defined as > 37.8C in elderly patients age > 65 [23] and > =38C in all other patients. Sepsis as per the SIRS [24] and qSOFA [25] score were calculated for all patients.

Urinalysis was done using urine test strips.

History of prior UTI was based on patient reporting, as documented in the patient chart by the main responsible physician.

Outcomes

Patients were followed until death in hospital or discharge. Length of stay was calculated from time of blood culture collection to discharge or death in hospital.

Statistical analysis

Comparisons between two groups were done with Wilcoxon rank-sum test for non-normally distributed continuous variables and Fisher’s exact test for categorical variables.

Diagnostic properties were determined for clinical factors as the test and urinary source as criterion standard. For example, if dysuria was the test, then a true positive was a patient who had dysuria and bUTI. A true negative was a patient who did not have dysuria and had a non-urinary source for the E. coli bacteremia. A false positive was a patient who had dysuria and a non-urinary source for the E. coli bacteremia. A false negative was a patient who did not have dysuria, but had bUTI. We calculated sensitivity and specificity with 95% confidence intervals (CI) using the Wilson method. For likelihood ratios, we calculated the 95% CI according to the method described by Simel et al. [26].

In patients with bUTI, a univariate logistic regression model was done predicting bUTI without urinary symptoms. Potential predictors were selected a priori, which included age, gender, stroke, dementia, urinary risk factors, delirium, and severity of infection. Significant predictors were selected based on p < 0.2 from univariate analyses. A final multivariable logistic regression model of significant predictors was selected based on clinical judgment, p-value, full model with all predictors, as well as both forward and backward stepwise regression based on Akaike information criterion. Hospital site was forced as a predictor into this model.

There are < 5% missing data for all variables, so listwise deletion was done for analyses such as modeling.

All reported CI were 2-sided 95% intervals and all tests were 2-sided with a P < 0.05 significance level. All analyses were done with R version 3.6.3 (R Foundation for Statistical Computing, Vienna, Austria).

Results

In total, 462 patients with E. coli bacteremia and a known infectious source were included in the analysis (Fig. 1). Of the 462 patients, 284 (61.5%) patients had a bUTI and 178 (38.5%) patients had a non-urinary source (Table 1). Baseline characteristics of patients with E coli bacteremia and an unclear source are described in Additional file 1: Table S1.

Fig. 1
figure 1

Flow diagram of patients included in this study

Full size image
Table 1 Baseline characteristics of E. coli bacteremic patients with bUTI versus non-urinary source
Full size table

Urinary frequency had the highest positive likelihood ratio (PLR) of 7.6 (95% CI 3.1–18.7) for ruling in urinary source (Table 2). Other urinary symptoms and signs have PLR ranging from 1.4 to 4.4 and negative likelihood ratio (NLR) ranging from 0.81 to 0.99. On clinical assessment, no urinary symptoms and signs had a NLR of 0.56 (95% CI 0.48–0.66). Negative leukocytes and nitrites on urinalysis had the lowest NLR of 0.18 (95% CI 0.11–0.28) for ruling out a urinary source.

Table 2 Diagnostic utility of symptoms and signs for bacteremic UTI
Full size table

Of the 284 patients with bUTI, 123 (43.3%) had no urinary symptoms (Table 3). Fever and/or urinary symptoms were present in 244 (85.9%) patients with bUTI. In the univariate analysis, potential significant predictors of bUTI without urinary symptoms included age, dementia, benign prostate hypertrophy, prior UTI, nephrolithiasis, delirium, sepsis by qSOFA criteria, and hypotensive shock (Additional file 1: Table S2). In the final multivariable model, significant predictors included only age, delirium and prior UTI (Table 4, Additional file 1: Table S3). With increasing age, the proportion of patients with delirium and bUTI without urinary symptoms also increased (Additional file 1: Table S4).

Table 3 Characteristics and outcomes of E. coli bUTI patients with and without urinary symptoms or signs
Full size table
Table 4 Multivariable logistic regression model predicting bUTI without urinary symptoms
Full size table

Discussion

This retrospective cohort study of patients with E. coli bacteremia showed that the majority of patients (61.5%) had a urinary source. A significant proportion of patients (43.3%) with bUTI did not have any urinary symptoms or signs elicited on history or physical exam especially in context of older age and delirium. Even the inclusion of fever and/or urinary symptoms would miss approximately 1 in 7 bUTIs. Sepsis by SIRS criteria was present in almost all of the bUTI cases without urinary symptoms.

We used a comparison group of non-urinary source E. coli bacteremia patients. The majority of these patients had intra-abdominal or hepato-biliary infection, similar to prior studies [27, 28]. Patients with a non-urinary source had similar outcomes including mortality compared to patients with urinary source. We believe this is an appropriate comparison group as it simulates a common scenario whereby a patient presents with sepsis or E. coli bacteremia of unknown source, and the clinician must evaluate and determine the source of infection.

Our finding of a significant proportion of bUTI presenting without urinary symptoms in elderly patients is consistent with prior studies on bUTI [11, 12] and non-bacteremic UTI [29]. Although our study found urinary frequency to be useful for ruling in UTI, it demonstrated little diagnostic utility in a systematic review of observational studies [29]. This may be due to the wide variation in the case definition of UTI that includes asymptomatic bacteriuria without true infection. The usefulness of urinalysis in ruling out UTI in our study was similarly found in prior studies with a reported high sensitivity [30] and negative predictive value [31].

There are several reasons why an elderly patient with delirium and a UTI may have no urinary symptoms. Frail adults may have atypical presentation of infections [6, 32]. Even if such patients experience UTI symptoms, they may not be able to describe or demonstrate them due to cognitive dysfunction [6]. While clinical evaluation for typical urinary symptoms and signs were not sensitive to capture all bUTI, sepsis by SIRS criteria came close. The qSOFA components such as altered mental status and tachypnea tend to be non-specific for infection in elderly patients [33, 34]. This may lead to overtreatment and missed diagnosis of delirium causes other than infection [8]. In our study, qSOFA was not sensitive for bUTI, being positive in only 38.7% of all bUTI patients and 46.3% of bUTI patients without urinary symptoms. In contrast, SIRS criteria have been shown previously to be sensitive in identifying bacteremia in all patients [35] including elderly patients in particular [33]. Therefore, physicians may continue to use SIRS criteria for decisions on when to draw blood cultures, especially in elderly patients.

This study has several strengths. It included a large population of 284 bUTI cases compared to prior studies that ranged from 61 to 191 patients [10,11,12]. It was conducted across academic and community hospitals, which increases its generalizability. Also, by limiting inclusion to those patients with bacteremia, it allows characterization of UTI without urinary symptoms and largely eliminates the inclusion of asymptomatic bacteriuria without true infection.

Our study has limitations that merit discussion. First, there are the inherent limitations from a retrospective chart review. However, the data collection was rigorous and a second auditor had performed sample data checking to ensure completeness and quality. The clinical assessment of patients for urinary symptoms and signs were not standardized and may vary between clinicians. Almost all patients in our study had sepsis as per SIRS or qSOFA criteria, and had a blood culture drawn at presentation. Therefore, all patients should have undergone a systematic evaluation for possible infectious foci including UTI. As well, the non-standardized approach reflects real world settings. Second, there may be ascertainment bias. The study could only capture patients in whom blood cultures were drawn. Physicians may be more likely to order blood cultures for patients in whom the symptoms are non-specific or the infectious source is uncertain. The study also likely selected sicker patients with higher suspicion for bacteremia. This could lead to overestimation of the diagnostic properties but does not affect our most important findings: that UTI is the most common cause of E. coli bacteremia, and most elderly or delirious patients had bUTI without urinary symptoms. Third, our study captured bUTI by E. coli only, so it would be an extrapolation to all bUTI. Nonetheless, E coli is the most common and important pathogen responsible for approximately 70% of pyelonephritis [17]. Fourth, UTI presentation may be different and unique in patients with spinal cord injury or disease. Unfortunately, our study did not collect information on spinal cord injury or disease. It is likely that many of these patients would have been captured in the variable of chronic indwelling Foley catheter.

These findings offer insight in the clinical presentation of E coli bUTI. Urinary symptoms are useful to diagnose UTI. However, elderly and/or delirious patients may have a bUTI despite having no urinary symptoms or signs elicited on history or exam. For these cases, in addition to symptoms and signs, SIRS criteria and positive urinalysis without any other clear infectious source may be important clues to a bUTI that require antibiotic therapy. Our study should be interpreted with caution and not be extrapolated to E coli bacteriuria without bacteremia. This is outside the scope of our study. Thus, our study does not address which patients are at risk for bUTI. This should be explored in future studies.

Our study adds to the evidence that UTI without urinary symptoms is common and important in elderly and/or delirious patients. We hope that this study can contribute towards a meaningful update of the concept of UTI symptoms. Emphasis on urinary symptoms without consideration of other aspects of the patient’s presentation is potentially harmful and may miss bUTI. A more holistic approach would consider other clinical factors including age, delirium and sepsis by SIRS criteria. This approach may help increase the chance of antibiotics being given to those who truly need it.

Conclusions

Most of the patients with E. coli bacteremia had a urinary source. A significant proportion of bUTI cases had no urinary symptoms. Elderly and delirious patients were more likely to have bUTI without urinary symptoms. In these elderly and delirious patients who satisfied SIRS criteria but without a clear infectious source, clinicians should suspect, investigate, and treat for bUTI.