Authors: Sonali D. Advani, David Ratz, Jennifer K. Horowitz, Lindsay A. Petty, Mohamad G. Fakih, Kenneth Schmader, Lona Mody, Tawny Czilok, Anurag N. Malani, Scott A. Flanders, Tejal N. Gandhi, Valerie M. Vaughn
Categories: Original Investigation, Research, Online Only, Infectious Diseases
Source: JAMA Network Open
This cohort study evaluates prevalence and factors associated with bacteremia from a presumed urinary source in inpatients with asymptomatic bacteriuria with or without altered mental status.
Urinary tract infection (UTI) is one of the most overdiagnosed infections, especially in hospitalized patients, older adults, and catheterized patients.^1^ Recent data show almost 50% of antibiotic prescriptions for outpatient UTIs are either inappropriate or unnecessary.^2^ Treatment of asymptomatic bacteriuria (ASB) is even more common in hospitalized older adults and those presenting with dementia and/or altered mental status (AMS).^3,4,5,6,7^ AMS remains one of the primary indications for ASB treatment despite guidelines from the Infectious Diseases Society of America (IDSA) recommending withholding antibiotic treatment in patients with ASB.^8^
One reason for ASB overtreatment is clinicians’ concern that poor outcomes (eg, bacteremia from UTI) may occur if antibiotics are not started early.^9,10^ Evidence to guide treatment in these situations are sparse. On the one hand, studies suggest antibiotic delays in patients with bacteremia or severe sepsis may increase mortality.^11^ In contrast, antibiotic treatment of ASB has not been shown to improve clinical outcomes and is instead associated with increased health care utilization, adverse drug events, and Clostridioides difficile infection (CDI).^6,12^ Prior studies have suggested that estimating a patient’s risk of an outcome (eg, bacteremia) and then treating with antibiotic therapy if their risk exceeds 2% could balance potential under and overtreatment.^13,14^ However, there is no validated way to estimate the risk of bacteremia in a hospitalized patient with ASB.
We sought to (1) determine the prevalence of bacteremia from a presumed urinary source in a large, multihospital cohort of hospitalized patients with ASB and a subgroup of patients with AMS; (2) examine factors associated with bacteremia from a presumed urinary source in patients with ASB; and (3) estimate antibiotics avoided if a 2% risk of bacteremia were used as a threshold for empiric antibiotic treatment of ASB.
This cohort study includes data from 68 hospitals (53 urban and 15 rural, including a critical access hospital) in the Michigan Hospital Medicine Safety (HMS) Consortium. HMS hospitals range in size from 25 to 1131 beds with a median (IQR) bed size of 275 (151-392). This project was not regulated by the University of Michigan Medical School’s institutional review board, which deemed the work to be quality improvement and thus waived the requirement of informed consent. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline was followed.
This study included a sample of adult patients admitted to nonintensive care unit settings in HMS hospitals with a positive urine culture between July 1, 2017, and June 30, 2022. Patients were consecutively reviewed with the first patient included daily. Patients were ineligible for inclusion if they met any of the following (1) age younger than 18 years; (2) pregnant and/or breastfeeding; (3) altered urinary tract anatomy, urologic surgery during hospitalization, or urinary stent or nephrostomy tube in place during hospitalization; (4) intensive care unit admission within 3 days before or after urine culture; (5) under hospice care on admission; (6) patient directed discharge; (7) concomitant infection (abstractors excluded any patient with documented antibiotic treatment for a concomitant bacterial infection during the hospital encounter unless the infection was potentially related to the UTI [eg, bacteremia, CDI]); (8) active treatment and/or prophylaxis for UTI on admission; (9) solid organ or bone marrow transplant recipient; (10) HIV with CD4 count less than 200 cells/mm^3^; (11) neutropenia (absolute neutrophil count <0.5 cells/μL on hospital day 1 or 2); or (12) within the 30 days after discharge from index hospitalization already abstracted for that patient.
For this study, patients were excluded if they had specific signs or symptoms of a UTI defined per IDSA ASB guidelines^8,16,17,18,19^ (Box). Patients with isolated candiduria and Staphylococcus aureus bacteriuria were also excluded as candiduria frequently represents colonization or contamination, and Staphylococcus aureus in urine usually represent hematogenous spread^20,21^ (eFigure 1 in Supplement 1).
As previously described,^6,7,22^ trained abstractors collected data retrospectively from patient records (eMethods in Supplement 1). Briefly, deidentified data were collected from 90 days before admission until 30 days after discharge or sooner if follow-up was terminated by a major complication (eg, death). Variables collected from the medical record include demographics, signs and symptoms, laboratory findings, vital signs, antibiotic type and duration, and outcomes. Information on signs and symptoms were collected from clinician and nursing documentation 3 days before through 3 days after urine culture collection. Demographics like race, ethnicity, age, and sex were assessed to better understand their association with bacteremia from a presumed urinary source. Thirty-day patient outcomes were collected via medical record review and prospective patient phone call 30 days following hospitalization. Patients who died or were discharged to hospice or another care facility were not eligible for postdischarge phone calls.
The primary outcome was bacteremia from a presumed urinary source (positive blood culture with matching organism within 3 days of urine culture). We assessed for possible variables associated with bacteremia including age, sex, body mass index (BMI), complicated urologic history, comorbidities, presence of dementia or AMS, nonspecific signs or symptoms (eg, foul smelling urine, fatigue) and relevant laboratory results (ie, elevated peripheral white blood cell [WBC] count, urinalysis parameters, and urine culture results). Secondary outcomes included duration of hospitalization after urine culture and 30-day CDI event, mortality, readmission, and/or emergency department visit.
We calculated the prevalence of bacteremia from a presumed urinary source in all patients and in a subgroup of patients with AMS with or without dementia. We assessed variables associated with bacteremia initially in a bivariable analysis using χ^2^ tests or Wilcoxon rank-sum tests, as appropriate. We treated age as a linear variable and BMI as a categorical variable. Missing variables (urine WBC, leukocyte esterase, serum leukocytosis, BMI) were imputed using 10-fold multiple imputation. For those with a urinalysis, we assessed variables associated with bacteremia in multivariable logistic regression analysis accounting for clustering within hospitals with a random intercept. WBC counts from urine were provided as ranges rather than exact values, so we categorized this variable and dichotomized it for the final model as less than 25 vs 25 or higher. Serum WBC count was log-transformed to improve model fit. Age was modeled linearly. We retained variables in the final multivariable model that were significant or that were observed to have a confounding effect on the association between another variable and risk of bacteremia. We expressed results as adjusted odds ratios (aORs) with 95% CIs, using a 2-sided P value less than .05 to indicate significance, although log odds ratios were used in the figure to preserve spatial relationships between variables. All analyses were performed in SAS, version 9.4 (SAS Institute Inc). Data were analyzed from August 2022 to January 2023.
Based on the multivariable model, we next calculated mean estimated probability of bacteremia for each patient using various combinations of factors including sex, symptoms such as urinary retention and fatigue, clinical signs such as tachycardia, and laboratory markers such as WBC thresholds and presence of pyuria on urinalysis. Using a 2% risk of bacteremia as a cutoff for whether a patient should receive empiric antibiotics or not, we first assessed the estimated risk of bacteremia for each patient in our cohort. For patients below 2% risk, we then calculated—compared with their actual antibiotic treatment—how many patients could have avoided antibiotic therapy for ASB. Here, empiric therapy is defined as any antibiotic therapy on the day the urine culture was sent or the day after.
Of 11 590 hospitalized patients with ASB (median [IQR] age, 78.2 [67.7-86.6] years); 8595 female patients [74.2%]; 2235 African American/Black patients [19.3%], 184 Hispanic patients [1.6%], and 8897 White patients [76.8%]), 8364 (72.2%) received antimicrobial treatment for UTI, while only 161 (1.4%) developed bacteremia from a presumed urinary source. Demographics of the entire cohort are described in Table 1. Nearly half (5059 patients [43.6%]) had AMS, 3210 (27.7%) had dementia, 1761 (15.2%) had an indwelling urinary catheter, 6323 (54.6%) had complicated urologic history, 11 039 (95.2%) had a urinalysis test (individual urinalysis parameters further described in Table 1), and 3589 (31.0%) had blood cultures obtained within 3 days of urine cultures. Of blood cultures, 401 of 3589 (11.2%) were obtained before the urine culture, 2836 of 3589 (79%) were obtained the same day, and 487 of 3589 (13.6%) were obtained after the urine culture. Patients who did not receive blood cultures had a significantly lower prevalence of systemic inflammatory response syndrome (SIRS), hypotension, tachycardia, and leukocytosis than those who had blood cultures (eTable 1 in Supplement 1).
Among patients with ASB, 5059 (43.6%) had AMS (Figure). Of these, 89 (1.8%) were found to have bacteremia from a presumed urinary source (vs 72 [1.1%] in patients without AMS). Risk of bacteremia in patients with AMS differed 4-fold based on whether they had systemic signs of infection (ie, evidence of SIRS or leukocytosis): 0.7% (17 of 2610) for patients without systemic signs of infection (number needed to treat, 154) vs 2.9% (72 of 2449) for patients with systemic signs of infection (number needed to treat, 34). Differences based on whether patients had dementia are shown in the Figure.
Figure. Bacteremia From a Presumed Urinary Source Among Hospitalized Patients With Bacteriuria and Altered Mental Status With or Without DementiaSBP indicates systolic blood pressure; SIRS, systemic inflammatory response syndrome; UTI, urinary tract infection.
In unadjusted comparisons, patients with ASB who developed bacteremia from a presumed urinary source were older (median [IQR] age, 79.5 [72.2-88.2] years) than patients who did not develop bacteremia (median [IQR] age, 78.2 [67.7-86.6] years; z score, 2.32; P = .02). Symptoms or comorbidities associated with developing bacteremia from a presumed urinary source included presentation with AMS with or without dementia, complicated urologic history, indwelling catheter, change in urine characteristics, fatigue, functional decline, and urinary retention (see Table 1 for details). Diagnostic findings associated with development of bacteremia from a presumed urinary source serum leukocytosis, elevated urinalysis parameters (ie, pyuria or leukocyte esterase), and growth of Escherichia coli (compared with other pathogens) (Table 1).
Patients with ASB who developed bacteremia from a presumed urinary source had higher unadjusted mortality (9.3% vs 4.5%; χ^2^1 = 8.24; P = .004) and median (IQR) duration of hospitalization after urine culture (6 [4-7] vs 4 [3-6] days; z score, 7.08; P < .001) than those without bacteremia. Other outcomes including 30-day readmission, 30-day ED visit, and CDI event at 30 days were similar between the 2 groups (Table 1).
On multivariable analyses accounting for hospital clustering, we found that male sex (aOR, 1.45; 95% CI, 1.02-2.05), hypotension (aOR, 1.86; 95% CI, 1.18-2.93), 2 or more SIRS criteria (aOR, 1.72; 95% CI, 1.21-2.46), urinary retention (aOR, 1.87; 95% CI, 1.18-2.96), fatigue (aOR, 1.53; 95% CI, 1.08-2.17), log of serum leukocytosis (aOR, 3.38; 95% CI, 2.48-4.61), and pyuria with more than 25 WBC/hpf on urinalysis (aOR, 3.31; 95% CI, 2.10-5.21) were associated with bacteremia from a presumed urinary source (Table 2; eFigure 2 in Supplement 1). In contrast, older age, AMS, dementia, and change in urine were not associated with a higher risk for bacteremia from a presumed urinary source (Table 2; eFigure 2 in Supplement 1).
In the absence of other findings, no single nonspecific sign or symptom (eg, AMS) or comorbidity conferred a 2% or greater risk of bacteremia from a presumed urinary source. The mean estimated probability of having bacteremia from a presumed urinary source ranged from 0.09% to 16.18% across combinations of variables (see eTable 2 in Supplement 1). For example, mean (SD) estimated probability of bacteremia in a male patient with ASB presenting with urinary retention (no fatigue or tachycardia), serum leukocytosis (>20 000/mL), and less than 25 WBCs on urinalysis was 16.18% (6.4%). In contrast, the mean (SD) estimated probability of bacteremia in a female patient with ASB presenting without tachycardia, urinary retention, or fatigue, and with serum WBC less than 5000/mL and pyuria of 25 or lower WBCs on urinalysis was 0.09% (0.004%).
Mean estimated probabilities of bacteremia from a presumed urinary source that were 2% or higher (our prespecified cut off for considering empiric antibiotic treatment) have been highlighted in eTable 2 in Supplement 1. Patients’ actual empiric treatment compared with their recommended treatment if a 2% risk of bacteremia were used to determine empiric treatment is shown in Table 3. Based on these results, using 2% as a cutoff to inform empiric antibiotic use in ASB would have avoided treatment in 6323 patients with very low risk of bacteremia (of whom 0.7% [44 of 6323] had bacteremia) and empirically treated an additional 206 patients with higher risk for bacteremia (of whom 1 [0.5%] developed bacteremia).
In this 68-hospital study of 11 590 hospitalized patients with ASB, 1.4% had bacteremia from a presumed urinary source and only 0.7% of patients with AMS and no systemic signs of infection did. No single risk factor conferred a 2% or greater risk of bacteremia. Specifically, older age, AMS, dementia, and change in urine character were not associated with bacteremia. These data reinforce prior evidence highlighting the poor yield of urine and blood cultures among hospitalized patients without systemic signs of infection and support not empirically treating patients with AMS and no systemic signs of infection.^23,24,25^
Our study highlights that bacteremia in adult inpatients with ASB is rare, compared with estimates as high as in 24% to 56% in symptomatic patients.^26,27^ The probability of bacteremia varies widely based on individual risk factors, clinical presentation, and laboratory findings. We found that the highest risk group had a 16.2% mean estimated probability of bacteremia as compared with 0.09% in the lowest risk group. Our study also adds data on laboratory markers that help identify patients at risk for bacteremia. Specifically, patients without specific signs or symptoms of UTI who developed bacteremia were twice as likely to have pyuria with 25 or greater WBCs on urinalysis and also more likely to have serum leukocytosis greater than 10 000/μL. Notably, no single characteristic conferred 2% or greater risk of bacteremia; rather, patients at elevated risk generally had multiple diagnostic findings, comorbidities, or symptoms.
IDSA guidelines for ASB highlight the dearth of evidence on whether antimicrobial therapy is beneficial for bacteriuria in patients with delirium or AMS in the absence of specific signs or symptoms of UTI.^8^ The guidelines suggest a strategy of watchful waiting in patients with AMS and no systemic signs of infection while recommending empiric antibiotic therapy in patients with systemic signs of infection. Our findings support this strategy, as patients with systemic signs of possible infection were significantly more likely than those without systemic signs of infection (2.9% vs 0.7%) to develop bacteremia. To obtain a thorough picture of a patient’s clinical condition, we used a 3-day infection window for assessing signs and symptoms and capturing blood culture data consistent with national guidance.^28^ Our data also highlight negligible rates of bacteremia in patients with AMS in the absence of SIRS, hypotension, or leukocytosis, supporting the safety of deferring empiric antibiotic therapy in this group. Moving forward, interventions that provide absolute risks personalized to a patient’s presenting signs and symptoms could be one way to inform evidence-based empiric antibiotic use—or avoidance. In our analyses, a risk-based approach would reduce unnecessary antibiotic exposure in almost 70% of patients with low risk of bacteremia.
Little scientific evidence exists on risk factors and epidemiology of bacteremia from a urinary source, especially in older adults.^29^ Prior studies have shown that diabetes, immunosuppression, catheterization, and shaking chills are independent risk factors for bacteremia from a urinary source.^30,31,32^ In our analysis, patients with diabetes, immunosuppression, and catheterization were not at higher risk for bacteremia in the absence of specific signs or symptoms of UTI. This discrepancy could indicate those risk factors are more prominent in symptomatic patients—a group we excluded. Additionally, none of these studies investigated specific urinalysis pyuria thresholds.
Our study has important implications for risk stratifying inpatients with ASB. First, these data provide assurance that history of dementia alone is not a risk factor for bacteremia, likely because ASB is so common in this group. Second, if patients have altered mentation and cannot attest to having specific signs or symptoms of UTI, clinicians should assess for SIRS, leukocytosis, and pyuria when deciding who may possibly benefit from empiric antibiotic treatment. If the patient with ASB does not have systemic signs of infection, they have a very low risk of bacteremia from a urinary source. Finally, moving forward, a validated UTI risk calculator may help determine the need for empiric antibiotic therapy in patients with positive urine cultures and could improve the precision of stewardship interventions. Using these personalized risk estimates would allow us to decrease unnecessary ASB treatment without substantially delaying early empiric therapy in those at highest risk of bacteremia.
Our study has limitations. First, this was an observational study dependent on presence of positive urine culture and documentation of signs and symptoms in the medical record. Second, we excluded concomitant infections, and hence, may not have captured the full scope of ASB in hospitalized patients. Third, we do not have data related to varying methods of diagnosing CDI events or nuances of urine culture reporting across hospitals. Fourth, our study only captures bacteremia in patients with blood and urine cultures growing the same organism within the 3-day infection window, so it may miss patients for whom blood or urine cultures were not obtained or were drawn after antibiotic initiation. Without systematically drawing blood cultures on all patients, it is not possible to determine the true prevalence of bacteremia in this population, but our data reflect clinical experience. Similarly, although we exclude patients with documented concomitant infections, we cannot be assured that all bacteremia was from a urinary source. Additionally, severely immunocompromised patients and those in intensive care units were excluded, where the risk or benefit calculation may be different. In real life, decisions are often made before urine culture results are known (eg, based on urinalyses); thus, for a population of all-comers, we over-estimated the risk of bacteremia.
In our multihospital cohort of 11 590 hospitalized patients with ASB, bacteremia from a presumed urinary source was rare. The risk of bacteremia in patients presenting with AMS was negligible in the absence of systemic signs of infection (eg, leukocytosis or SIRS). A personalized, risk-based approach to empiric antibiotic therapy in patients without specific signs or symptoms of a UTI could decrease unnecessary ASB treatment without delaying early empiric therapy in those at highest risk of bacteremia.