Rate and yield of imaging for acute pyelonephritis in the emergency A retrospective cohort study

Authors: Jessica Yu, Christine Koolstra, De Villiers Smit, Biswadev Mitra

Abstract

Objectives

The role of imaging in acute pyelonephritis (APN) in the ED is poorly understood, with variability among clinical guidelines for when patients should be imaged, and the modality of imaging. The objective of this study was to identify the proportion of patients with APN being imaged, the proportion abnormal findings, and the association between abnormal imaging and discharge disposition.

Methods

A single‐centre retrospective review of patients with a discharge diagnosis of APN at an adult tertiary referral hospital over a 5‐year period (2018–2022) was conducted. The proportion of patients with APN investigated with imaging, and abnormalities on imaging were reported. Logistic regression analyses were performed to assess whether imaging was associated with change in disposition from the ED.

Results

There were 778 patients included for analysis. Among these, 210 (27%) were investigated with ultrasound (US) and/or computed tomography (CT) in the ED. Of the 214 imaging reports available, 112 (52%) were abnormal. Imaging was associated with hospital ward admission (adjusted odds ratio [aOR] 5.28; 95% confidence interval [CI] 3.35–8.31) as was abnormal imaging (aOR 4.51; 95% CI 2.62–7.75). Other variables associated with hospital ward admission were higher temperature and heart rate, higher C‐reactive protein levels and poorer renal function.

Conclusions

Among patients with APN, abnormalities on imaging were common and both imaging and abnormalities on imaging were associated with hospital ward admission. This suggests that there is possible utility of early and routine imaging for patients with APN to allow clinicians to efficiently make decisions about patient disposition.

Key findings Among patients with acute pyelonephritis, 27% of patients had imaging performed in the ED.Imaging for acute pyelonephritis was abnormal in 52% of cases.Imaging was associated with admission to hospital, thus imaging may be considered among patients with clinical features of higher severity of disease (those with higher temperature, heart rate, C‐reactive protein levels and poorer renal function).

Introduction

Acute pyelonephritis (APN) – derived from Greek, meaning ‘kidney pelvis inflammation’ – is defined as an infection of the upper urinary tract, consisting of the renal parenchyma, calyces and pelvis. ^1^ It is estimated that the global annual incidence of APN is 10.5–25.9 million cases, and the lifetime risk of APN in women is 20%–35%. ^2^ , ^3^

In the ED, the three key roles for imaging in APN are to (i) confirm the diagnosis of APN in unclear cases; (ii) identify the presence of complications that may require modification in management, with key complications that may be identified on imaging being obstruction, abscesses, necrotising infection and focal infection; and (iii) identify risk factors that may predict treatment failure. ^4^ , ^5^ , ^6^

There is variability in clinical guidelines and current practice regarding imaging among patients with APN in the ED setting – in particular, the use of imaging, the abnormalities identified on imaging, as well as the implications of imaging abnormalities on clinical management. ^7^ Therefore, the aims of the present study were (i) identify the proportion of patients with APN who are investigated with imaging in the ED; (ii) identify the abnormalities present on imaging in APN; and (iii) identify the clinical variables associated with hospital ward admission (compared with discharge from the ED, including discharge from the emergency short stay unit).

Methods

Design and setting

This was a single‐centre retrospective cohort study at The Alfred Hospital Emergency and Trauma Centre, a major metropolitan adult tertiary public hospital in Melbourne, Australia. The Alfred Hospital Emergency and Trauma Centre has an annual census of approximately 65 000 presentations per annum. At this hospital, both US and CT scanning is available at all hours, with a sonographer being on‐call after‐hours.

Inclusion and exclusion criteria

Patients were eligible for inclusion if pyelonephritis was coded as a discharge diagnosis. Electronic medical records were queried for patients that had presented to ED between a 5‐year period from 1 January 2018 to 31 December 2022 with a primary or additional diagnosis of pyelonephritis, as per the International Classification of Diseases (ICD‐10). Only patients ≥18 years at the time of presentation were included.

Patients that did not have a diagnosis of APN, or a more likely alternative diagnosis than APN were excluded. Patients that discharged against medical advice, absconded or transferred to a different hospital network were excluded. Indications for imaging in paediatric populations differ from adult populations, with imaging primarily used for investigating vesicoureteral reflux and its sequelae in APN. ^8^ As such, we excluded patients <18 years of age. Patients with conditions predisposing them to complicated infection were excluded because of a lower threshold for imaging. These conditions were the presence of a transplanted kidney, a single kidney and anatomic renal malformations. Imaging indications and their associated safety profile differ in pregnancy, particularly for modalities utilising ionising radiation such as CT. ^9^ As a result of this, pregnant patients were excluded from analysis.

Data collection

An explicit chart review was conducted for variables, which were likely to have an association with severe illness, and included demographics, vital signs, imaging, clinical features and disposition. Medical record reviews were conducted by a single investigator (JY), with 10% of records reviewed by a second investigator (CK). Abnormalities were categorised based on the most common features reported in radiologist reports and included features of inflammation, acute complications of APN and chronic abnormalities. Chronic abnormalities included scarring, angiomyolipoma, cysts and non‐obstructing calculi. These were reported and included for data analysis, as chronic abnormalities could have had clinical significance.

Statistical analysis

Continuous near‐normally distributed data were summarised using means and standard deviations (SDs), while skewed and ordinal variables were summarised using medians and interquartile ranges (IQR). Categorical variables were summarised using counts and proportions. The statistical significance of differences between means were assessed using Student's t test, differences between medians assessed using the Wilcoxon rank sum test and differences between proportions assessed using the χ ^2^ test. If values in a cell were less than 5, Fisher's exact test was used. Missing data were handled by listwise deletion of variables.

For our analyses, we defined ‘hospital ward admission’ as patients admitted to a hospital ward within the hospital and cared for by an inpatient team, and ‘no hospital ward admission’ as patients who were discharged home directly from the ED as well as patients who were admitted to the emergency short‐stay unit before being discharged home.

The association between variables and hospital ward admission were initially assessed using univariable logistic regression analysis. Variables that demonstrated some association with the outcome (P < 0.10) were entered into a multivariable model, and a parsimonious model developed using backward elimination. Post‐estimation assessments were performed using Hosmer–Lemeshow goodness of fit and multicollinearity assessed using variance inflation factors (VIF).

The effect of imaging on hospital admission was likely to have been modified by the result of the test. It was hypothesised that in the event that imaging was associated with admission to hospital, an abnormal result would be associated with admission, while a normal result would be associated with discharge from the ED. To test this effect, we repeated the regression analysis according to imaging findings. For the purpose of this analysis, imaging results were categorised to not done, normal and abnormal.

All analyses were performed using stata version 18 (StataCorp LLC, College Station, TX, USA). A P‐value of <0.05 was defined to be statistically significant. This study was approved by The Alfred Hospital Human Research and Ethics Committee (Project ID 216/23). The requirement to seek informed consent from patients was waived.

Results

There were 778 patients available for analysis who were diagnosed with pyelonephritis after presentation to the ED (Fig. 1). The mean age of these patients was 35 years (SD 15.4), and the sex distribution was 93% female and 7% male.

Of these, 210 (27.0%; 95% confidence interval [CI] 23.9–30.3) were assessed with imaging of the renal tract. Among the 210 patients, four patients were investigated with both a CT and US, with all other patients investigated with one modality of imaging. In total, 68 CT scans without contrast (9% of total cohort), 49 CT scans with contrast (6% of total cohort) and 97 US scans were performed (12% of total cohort). There were 97 US scans formally reported that included four scans performed by emergency physicians as point‐of‐care ultrasound (POCUS).

Among patients who underwent imaging, the association of imaging modality (CT vs US) with demographics and clinical characteristics are presented in Table S1. CT scanning was more frequent in older patients, those of male sex and less frequent among patients with high eGFR. The median time to first imaging was 4.9 h (IQR 2.9–11.9; Fig. 2). Patients who had imaging spent 3.4 h (IQR 2.1–5.9) in the ED, while those without imaging spent 2.7 h (1.7–3.9) in the ED (P < 0.001). Disposition destinations of patients are listed in Table S2.

Among the 214 imaging reports, 102 were normal and 112 were abnormal (inclusive of acute and chronic abnormalities). The abnormalities on imaging are summarised in Table 1. The outcome of hospital ward admission was observed in 193 (24.8%; 95% CI 21.8–28.0) patients. Imaging was associated with hospital ward admission (odds ratio [OR] 6.85; 95% CI 4.64–9.51). Among patients who did not have imaging, 82 of 568 (14%) were admitted. Among patients with normal imaging, 37 of 100 (37%) were admitted, and when imaging was abnormal, 74 of 100 (67%) patients were admitted. Univariable associations of other variables with the primary outcome are listed in Table 2.

Following backward elimination for variables that did not exhibit independent association with the primary outcome, the variables that remained are listed in Table 3 and illustrated in Figure 3. Factors associated with higher odds of admission were imaging, higher temperature, higher heart rate, higher white cell count, higher CRP and lower eGFR. The P‐value for Hosmer–Lemeshow goodness of fit was 0.76 and the mean VIF was 1.22, with maximum VIF of 1.53.

The adjusted association between abnormal imaging (inclusive of all acute and chronic imaging abnormalities) and hospital ward admission are listed in Table 4. Abnormal imaging was independently associated with hospital ward admission (adjusted odds ratio [aOR] 4.32; 95% CI 2.54–7.35). Clinical variables association with severity of infection (higher temperature, heart rate, white cell count and the CRP level) and renal function were also associated with hospital ward admission.

Discussion

Imaging for APN was common and associated with longer ED length of stay (LOS) and also hospital ward admission. It generates the hypothesis that early imaging may expedite decisions to disposition in a sub‐group of patients. This sub‐group appears to be those with vital signs consistent with infection (raised temperature and heart rate), poorer kidney function and raised inflammatory markers. This retrospective study cannot determine causality of imaging on hospital ward admission. Thus, the role of imaging to impact disposition requires prospective assessment.

In the present study, imaging by US, and/or CT with or without contrast, was performed in 27% of patients. This is consistent with previous reports from the ED setting, where Chen et al., ^10^ Esteban‐Zubero et al., ^11^ Gauthier et al. ^12^ and Khoo et al. ^13^ identified high rates of imaging with US in their patient cohorts, reporting rates of 20%–77%. CT scans, with or without contrast, were performed in 15% of patients in our study. Studies including patients from the ED reporting on the rate of imaging with CT have reported variable results. Khoo et al. ^13^ reported a rate of 10%, while Gauthier et al. ^12^ reported a rate of 63%. A study by Probst et al. investigated the factors contributing to a physician's decision making for ordering imaging – citing system factors, patient factors and physician factors as key inputs into the decision‐making process. ^14^ These likely account for the differences in imaging rates between the existing literature and our study. Four patients were investigated with both US and CT in the ED. While we cannot infer clinician decision‐making retrospectively, the assumption was that this was done to further characterise abnormalities on the initial test.

Acute abnormalities were variably detected across different imaging modalities. Among the few existing studies that have examined specific abnormalities identified on imaging for APN, Chen et al. reported hydronephrosis in 23%, perinephric fat stranding in 10% and renal abscess in 3%. ^10^ Hydronephrosis and renal abscesses were detected at a similar rate in our study, but we detected much higher rates for perinephric fat stranding.

We utilised hospital ward admission as our surrogate measure for a change in management. We identified that patients imaged with normal findings still had a high rate of admission than patients not imaged, pointing to the possibility of hospital wards requesting imaging as a prerequisite for admission. However, the admission records did not provide reliable information on decisions to imaging. This is a major confounding factor that we could not control because of the retrospective nature of the study.

We found that patients who were investigated with imaging spent longer in the ED compared to patients who did not receive imaging, likely because of the time required for imaging scans to be requested and performed in the hospital. Patients being imaged incur a larger expense to the healthcare system. However, imaging in the ED often provides earlier definitive diagnoses and may be the more efficient assessment strategy. Initially, it may seem that patients should not be imaged to mitigate additional costs incurred by an increased LOS in the ED. However, given the association between imaging results and need for hospital ward admission, early and routine imaging can enable decisions about patient disposition to be made more quickly and efficiently. Ideally, universal and early imaging would encourage the early discharge of patients that can be safely managed in an outpatient setting, which could reduce healthcare expenditure.

POCUS is a diagnostic tool that is performed and interpreted by a physician as a bedside test, commonly used in the ED. A study performed by Hashim et al. found that POCUS was equivalent to CT in the diagnosis of nephrolithiasis, and led to reduced ED LOS. ^15^ With adequate training, emergency clinicians have moderate to high sensitivity for identifying hydronephrosis on POCUS when compared to radiologists. ^16^ The training of ED physicians in independent use of POCUS can therefore lead to imaging being performed more quickly in the ED, allowing for decisions about patient disposition to be expedited.

Limitations of this study include the retrospective nature and single centre population. We found that we could not reliably determine the reasons for why a patient was referred for investigation by imaging because of poor documentation. We were also unable to infer reasons for admission under a specific hospital unit retrospectively. Imaging would have been performed for a range of reasons, such as resolving diagnoses in equivocal cases, ruling out diagnoses apart from APN, or for the purpose of satisfying hospital ward admission requirements. From the results of our study, we identified imaging requirements of hospital wards as a significant potential confounding factor, and the inability to control for this could largely be attributed to the fundamental drawbacks associated with a retrospective study design. We did not consider variables that may change the threshold for admission to hospital, such as comorbid medical conditions and other relevant past medical history. Furthermore, whether patients had a diagnosis of APN could only be determined by manual review of the medical records. The list of records was initially extracted based on an ICD‐10 primary or additional diagnosis code of pyelonephritis, in which we would expect records coded as APN to be the correct diagnosis. While the ICD‐10 diagnosis query generated a valuable list of patients potentially meeting our inclusion criteria, we ultimately primarily relied on the free text field in discharge summaries to determine whether a patient had APN, and excluded any records in which the free text field did not state APN as the most likely diagnosis. Similarly, it is likely that some patients with a diagnosis of APN were not included in this initial query because of the mistakes in diagnosis (ICD‐10) coding, and therefore would not have been included in our analysis. However, we expected that patients coded with APN had a high likelihood of the diagnosis. Involvement of only a single tertiary hospital also limits the generalisability of our results. It would be important to conduct this research in other settings to draw more comprehensive conclusions on the role of imaging in APN (Tables S3–S5).

Conclusion

Among patients with APN, abnormalities on imaging were common and both imaging and abnormalities were associated with hospital ward admission. There is potential utility of early and routine imaging to allow clinicians to efficiently make decisions about patient disposition.

Supporting information