Decreasing Misdiagnoses of Urinary Tract Infections in a Pediatric Emergency Department

The Hospital for Sick Children in Toronto, Ontario, Canada is a 370-bed, quaternary care, freestanding academic children’s hospital caring for patients from birth to age 18 with annual ED volumes of ∼80 000 patients. Using the hospital EMR for data extraction, ED patients being evaluated for a possible UTI with urine testing were eligible for this QI study. The hospital EMR was Sunrise Clinical Manager (Allscripts, Richmond, BC) until June 2018, when the EMR transitioned to Epic (Epic Systems Corporation, Verona, WI). To minimize any unintended risk, we excluded children who (1) were aged <3 months, (2) were clinically unwell, with pyelonephritis or urosepsis or requiring hospital admission, (3) had underlying medical complexity, including a known underlying urologic or renal condition, (4) had any antibiotic exposure in the last week, or (5) had Udip or UC results that were not documented in the EMR. Exclusion criteria were applied through manual chart reviews.

A multifaceted QI strategy was used to address the quality gap of UTI misdiagnosis and unnecessary antibiotic utilization by using the Model for Improvement methodology.²²  In June 2017, a study team involving key stakeholders from Emergency Medicine, Infectious Disease, and Microbiology was assembled. The QI team reviewed baseline data and created a driver diagram to meet our aim (Fig 1).

All urine specimens in our ED were collected with 1 of 2 mechanisms: (1) clean catch or (2) urinary catheterization. A POC Udip (Clinitek Status, Siemens Healthcare, Munich, Germany) was the screening test used to assist with UTI diagnosis before sending a UC to the laboratory. Samples obtained via catheterization had an automatic UC sent to the laboratory, regardless of Udip results, whereas clean catch samples were left to provider discretion after Udip to determine if a UC should be sent.

To improve diagnostic accuracy, a guided UTI diagnostic algorithm was developed by using plan-do-study-act cycles to standardize the clinicians’ posttest interpretation of abnormal dipstick results. First, a literature review and environmental scan of other pediatric centers and national organizations was completed along with reviewing the analysis from our center’s study.^4,23  Once an initial draft of the algorithm was completed, qualitative feedback was elicited at ED Quality Council and staff meetings, leading to additional refinements. Finally, the algorithm was piloted clinically in the ED setting before finalizing.

The diagnostic algorithm (Fig 2) guided clinicians on whether to make a presumptive UTI diagnosis with empirical antibiotic prescription during the ED visit while cultures were pending versus discharging while awaiting the UC results without empirical antibiotics. If the inclusion criteria were met, the decision points in the algorithm included the presence of nitrites, the method of urine collection, and the quantity of leukocyte esterase present. To integrate the algorithm into clinical care, posters were displayed in the ED and the intranet site. In September 2019, an EMR banner link to the algorithm was created and appeared when a Udip was ordered.

At baseline, an ED callback system was well established with a nurse practitioner or physician assistant contacting patients’ families within 48 hours for all positive UC results and confirming the appropriate treatment. If a UTI diagnosis or antibiotic prescription had not been provided during the ED visit with a positive UC result, the nurse practitioner or physician assistant would assess for ongoing symptoms and, in consultation with an ED physician, determine the next steps, including whether to start antibiotics. Conversely, negative UC results were generally not reviewed and were discarded. Confirmed UTI was defined as the presence of >10 000 colony-forming units (CFU) per mL (>10 × 10⁶ CFU/L) of 1 or more uropathogens.²⁴  Negative UC results were defined at our institution as no growth, insignificant growth (≤10 000 CFU/mL), or mixed growth.

In collaboration with Microbiology, negative UC results were added to the pool of outstanding test results sent to the ED for daily follow-up in December 2017. While continuing the baseline process for positive UC results, the callback group also began contacting patients with negative UC results and an empirical UTI diagnosis, informing them of the incorrect diagnosis and to discontinue antibiotics. In September 2019, the ED developed a dedicated nursing (RN) role for all outstanding test results, including the negative UC result callback system.

A standardized antibiotic prescription duration for uncomplicated pediatric UTIs was also targeted because it is a recognized strategy to improve antimicrobial stewardship.²⁵  After completing a literature review, the consensus recommendation was 7-days’ duration. Specific antibiotics based on local antibiograms for UTI treatment along with the duration recommendation were added to the UTI algorithm, promoted in educational rounds, and EMR modifications. A reminder of a 7-day standard was also added to the original EMR and triggered whenever a discharge diagnosis of UTI was entered into the discharge summary. This intervention was lost in June 2018 until a UTI discharge order set was built in the new EMR (Supplemental Fig 6) in June 2020. If a Udip was ordered, a suggested UTI discharge order set could be selected and prepopulated with the UTI discharge diagnosis, appropriate 7-day antibiotic prescriptions, and routine UTI anticipatory guidance.

In October 2017, the UTI practice changes and rationale were discussed at an ED education session. After this, the initiative was then incorporated into biannual divisional education rounds on resource stewardship. In January 2018, the QI project was adopted as part of the hospital’s Choosing Wisely campaign. Reminders were regularly delivered in group e-mails and posted on hospital screensavers and the intranet site. In July 2019, the initiative was also added to the hospital’s corporate scorecard. Lastly, clinician adherence to the algorithm was routinely monitored and consistent outliers were targeted with e-mails to understand their decision-making while reminding them of the initiative (Supplemental Fig 7).

Outcome measures were tracked retrospectively from January to September 2017 (baseline) and prospectively from October 2017 through June 2021. The primary outcome measures included the percentage of ED UTI misdiagnoses, as well as the total antibiotic days saved (ADS). UTI misdiagnoses were determined by calculating the percentage of monthly ED patient visits with a discharge diagnosis of UTI with a subsequent negative UC result. ADS was calculated by subtracting the antibiotic days presumed completed (ED discharge date to notification date of the negative UC results) from the total prescription duration.

For process measures, the monthly clinician adherence to the diagnostic criteria in the UTI algorithm and the percentages of cases that were prescribed the recommended 7-day antibiotic duration was followed retrospectively from June to September 2017 (baseline) and prospectively from October 2017 through June 2021. Additionally, the percentage of patients with a UTI misdiagnosis who were subsequently contacted and notified of their negative UC results was monitored.

As a balancing measure, charts of ED patients with positive UC results meeting study inclusion criteria who did not have an ED discharge diagnosis of UTI were reviewed prospectively for subsequent ED visits and hospitalization. Cases were excluded if the Udip result was negative for both leukocyte esterase and nitrites as the absence of pyuria would not have met our algorithm’s criteria for a possible UTI diagnosis. Cases were also excluded if the UTI algorithm was not followed because these cases were not directly impacted by the study intervention. The remaining positive UC result cases were presumed to be a potential missed UTI.

Data were collected from January 2017 through June 2021 and tracked with statistical process control (SPC) and run chart methods. SPC charts were created by using QI Macros (QI Macros for Excel, version 2015; KnowWare International, Inc, Denver, CO) and special cause variation was monitored by using standard SPC chart methods.²⁶  Annotations were added to the charts to highlight the relationship between the interventions and significant context changes. The study was approved as a QI study by our institutional quality risk management team and met the criteria for research ethics board exemption.

From January 2017 until June 2021, 3545 children (1.1% of all ED visits) were discharged from the ED with a diagnosis of UTI. After applying exclusion criteria, 2254 (63.6%) met inclusion for analysis (Table 1).

At baseline, the percentage of UTI misdiagnoses was 54.6% of the 617 UTI diagnoses meeting inclusion criteria. Postimplementation, the primary outcome of UTI misdiagnoses decreased to 26.4% of 1729 UTI diagnoses (51.6% relative reduction) and was sustained for 19 months (Fig 3). Special cause variation was detected in both October 2017 and December 2019.

After implementation of the negative UC result callback system, an increase from 0% to a median of 89.5% of patients were contacted with 2128 ADS (Fig 4). Nonrandom variation was detected at the start of the system and in September 2019 with the implementation of a dedicated callback nursing role.

Clinician adherence to the Udip criteria in the UTI algorithm was 60.9% at baseline, improved to 83.7% with the study interventions, and was sustained at 72.7% (Fig 5A). The run chart reveals a shift in both October 2017 and 2018 coinciding with the UTI education and audit and feedback. A downward shift in adherence occurred in February 2020 coinciding with the start of the coronavirus disease 2019 (COVID-19) pandemic but still improved from baseline. The percentage of UTI prescriptions with a 7-day duration increased from 45.1% to 84.6% postintervention (Fig 5B). An early shift was detected in December 2017 and August 2020 coinciding with the new EMR discharge order set.

For balancing measures, a total of 186 patients had positive UC results without meeting our UTI diagnostic criteria over an almost 4-year period (Table 2). All (100%) patients were contacted to clarify the diagnosis and prescribed antibiotics as indicated. Fourteen (7.5%) patients returned to the ED within 72 hours, 12 of whom were discharged again with UTI diagnoses, whereas 2 (1.1%) patients (aged 4 months and 13 months) were admitted for intravenous antibiotics because of a lack of oral options due to resistant gram-negative UC results and not illness severity.

We present a QI study revealing a significant and sustained reduction in UTI misdiagnoses for children in our high-volume, quaternary care, pediatric ED while simultaneously improving antimicrobial stewardship. Implementing a 2-pronged approach targeting both the initial UTI diagnosis and the postvisit follow-up of the final UC result, an estimated 350 fewer children were misdiagnosed with a UTI annually and >2000 ADS. By applying the Agency for Healthcare Research and Quality Safety Program’s framework for improving antibiotic use to our study design,²⁷  children not only received more accurate diagnoses during their ED visit, but when the diagnoses were inaccurate, families received timely notification and discontinuation of antibiotics.

Diagnostic errors are common in medicine, particularly incorrect interpretation of test results, which contribute to an estimated 37% of the diagnostic errors in emergency medicine.²⁸  In addition, an individual test can lead to a cascade of events, including unnecessary tests and treatments, some of which have the potential to harm.^29,30  The downstream impacts of UTI misdiagnosis include unnecessary antibiotic exposure either empirically while awaiting culture results or because of the lack of follow-up on negative culture results. Harms include antibiotic side effects, potential allergic reactions, and the selection of antimicrobial-resistant organisms, all while potentially masking the underlying cause for the illness, leading to additional health care visits.²⁵  Further downstream, a UTI misdiagnosis can lead to increased UTI screening with future illnesses, renal ultrasounds, or more invasive imaging with voiding cystourethrograms.^8,9  Our improvement efforts likely reduced unnecessary health care utilization through the mitigation of immediate and potential downstream impacts of the UTI misdiagnoses.

In the planning stages of the initiative, a multifaceted, iterative intervention strategy was targeted, which included system-based changes because these interventions are generally more effective in achieving successful improvements compared with single-component, person-based interventions.^22,31  At the beginning, an ED UTI education session shared baseline data on the high percentages of UTI misdiagnoses and created a sense of urgency to address this quality gap.³²  However, recognizing that educational interventions alone do not tend to change behaviors or produce improvements,³³  a UTI diagnostic algorithm was also developed to standardize the clinical approach to empirical UTI diagnosis and treatment. The algorithm’s official launch coincided with the initiative also being adopted as part of the broader hospital’s Choosing Wisely campaign. These interventions all led to significant improvements at the outset; however, they were notably shy of achieving our intended aim. Recognizing that electronic decision support tools embedded within care environments standardize care, deliver evidence-based practice recommendations, and improve quality,^34–39  we anticipated the need to integrate the algorithm into our EMR, but our hospital’s migration to a new EMR mid-2018 created a significant delay. However, once the initiative was selected for the corporate quality scorecard, the EMR build was prioritized, and the pathway became embedded into workflows as a decision support tool whenever a Udip was ordered. The EMR algorithm integration, coupled with the additional EMR UTI discharge order set, led to the special cause variation necessary to achieve our intended aim and eventual sustainability of the improvements. Lastly, audit and feedback, a well-studied improvement strategy and powerful driver of individual practice change,⁴⁰  targeted clinicians who did not adhere to the UTI algorithm and helped address barriers. Routine reminders, biannual stewardship education, and audit and feedback all contributed to the long-term sustainability of the results.

Recognizing that the UTI decision support tool is based on an imperfect test (Udip) and would never prevent misdiagnoses entirely, a reflexive callback system for negative UC results was an equally important intervention to improve care quality. Despite implementing the negative culture result callback system in December, the highest ED volume month when many core staff members are away, the system quickly gained traction. ED leadership support along with identifying a small, core group of clinicians responsible for this added workload were key factors at the outset. Reminders highlighting the impact of their efforts were frequently used for positive feedback and incentivization. Once a dedicated callback RN role was created, whereby the main responsibility was to manage all outstanding test results without other clinical responsibilities, the system accelerated and achieved close to the target of notifying 100% of families.

Our balancing measure revealed a potential for some missed true UTIs; however, our robust callback system ensured follow-up with all patients with timely antibiotic prescriptions when indicated. Short-term delays in antibiotic treatment in low-risk children have not been associated with increased renal scarring, urosepsis, or other significant complications.¹  This was demonstrated in our study with only 2 hospitalizations from a delayed UTI diagnosis compared with 773 total hospitalizations for UTIs during this time period. Both of these hospitalizations were related to multiresistant organisms that did not have oral treatment options as opposed to illness severity. Although there is a need for better POC technology for more accurate UTI diagnoses at the time of care,^3,9  limitations in diagnostic certainty can be mitigated through risk reduction strategies like short-term follow-up plans and reviewing outstanding test results, which is an important safety principle in today’s complex health care environment.^41,42  Notably, some of the positive UC results were from patients with suspected contaminated specimens due to alternative diagnoses like gastroenteritis and balanitis in which UCs are generally not indicated and often represent false positives (Table 2).

Our study does have limitations. First, the total ADS might be overestimated as it assumed families stopped the antibiotics as recommended when they were contacted. However, the ADS could also be underestimated because parents sometimes reported that they had not yet filled the empirical prescription from the ED encounter. Second, although the new RN callback role was not created specifically for this study, it did directly benefit our callback rates and, importantly, can add a significant, ongoing cost for institutions. However, this cost can potentially be offset by freeing up others for increased clinical productivity. For the selected balancing measures, return visits were only captured for patients returning to our hospital ED and not other institutions. Patients may have returned to outside hospitals, clinics, or other health care settings, but our hospital has the only pediatric ICU in the city. Conversely, the balancing measure criteria for a missed UTI diagnosis were presumed and likely overly conservative because UTI criteria require pyuria and a positive UC result in a symptomatic child. It is unclear how many of these children were still symptomatic once the culture was received and the family was contacted. To reduce practice variation and concerns with follow-up from a busy ED, many of these children were prescribed antibiotics and told they likely had a UTI diagnosis. Depending on the setting, these patients could undergo repeat testing to better clarify a true UTI diagnosis. Finally, our institution uses more conservative criteria: >10 000 CFU/mL for a positive UC result rather than >50 000 CFU/mL, which is often the reference standard for UTIs.^8,9,24 

A multipronged and iterative QI approach can reduce misdiagnoses of UTIs in children while improving stewardship practices. Leadership support, electronic decision support tools at the time of diagnosis, and a follow-up system for outstanding results are essential to improvement success and minimizing unintended consequences. These improvement strategies not only make it easier to provide better care but support clinicians in providing children the right care.

ADS

antibiotic days saved

CFU

colony forming unit

ED

emergency department

EMR

electronic medical record

POC

point-of-care

QI

quality improvement

RN

nursing

SPC

statistical process control

UC

urine culture

Udip

urine dipstick

UTI

urinary tract infection