Overdiagnosis of urinary tract infection linked to overdiagnosis of pneumonia: a multihospital cohort study

Abstract

Urinary tract infection (UTI) and community-acquired pneumonia (CAP) are the most common infections treated in hospitals. UTI and CAP are also commonly overdiagnosed, resulting in unnecessary antibiotic use and diagnostic delays. While much is known individually about overdiagnosis of UTI and CAP, it is not known whether hospitals with higher overdiagnosis of one also have higher overdiagnosis of the other. Correlation of overdiagnosis of these two conditions may indicate underlying hospital-level contributors, which in turn may represent targets for intervention. To evaluate the association of overdiagnosis of UTI and CAP, we first determined the proportion of hospitalised patients treated for CAP or UTI at 46 hospitals in Michigan who were overdiagnosed according to national guideline definitions. Then, we used Pearson’s correlation coefficient to compare hospital proportions of overdiagnosis of CAP and UTI. Finally, we assessed for ‘diagnostic momentum’ (ie, accepting a previous diagnosis without sufficient scepticism) by determining how often overdiagnosed patients remained on antibiotics on day 3 of hospitalisation. We included 14 085 patients treated for CAP (11.4% were overdiagnosed) and 10 398 patients treated for UTI (27.8% were overdiagnosed) across 46 hospitals. Within hospitals, the proportion of patients overdiagnosed with UTI was moderately correlated with the proportion of patients overdiagnosed with CAP (r=0.53, p<0.001). Over 80% (81.8% (n=952/1164) of UTI; 89.9% (n=796/885) of CAP) of overdiagnosed patients started on antibiotics by an emergency medicine clinician remained on antibiotics on day 3 of hospitalisation. In conclusion, we found overdiagnosis of UTI and CAP to be correlated at the hospital level. Reducing overdiagnosis of these two common infections may benefit from systematic interventions.

INTRODUCTION

Urinary tract infection (UTI) and community-acquired pneumonia (CAP) are the two most common infections treated in hospitals.¹ They are also commonly overdiagnosed.^2–4 Infection overdiagnosis may lead to unnecessary antibiotic exposure and delay correct diagnosis and treatment.⁵ While we have previously published work related to overdiagnosis of UTI,⁶ we have yet to quantify overdiagnosis of pneumonia or compare hospital proportions of overdiagnosis of CAP and UTI. Correlation of overdiagnosis of these two conditions may indicate underlying hospital-level contributors, which in turn may represent targets for intervention.

METHODS

This retrospective cohort study includes patients treated for UTI or CAP at 46 hospitals in the Michigan Hospital Medicine Safety (HMS) Consortium between 1 July 2017 and 7 December 2019.⁷

Patient eligibility criteria^{2 5} have been published previously and are detailed in the online supplemental appendix. Briefly, patients were eligible for inclusion if they were adult, hospitalised patients without a concomitant infection who received antibiotic treatment for UTI or CAP.⁵ We followed established methods for identifying patients with UTI and CAP.^{2 8–10} Patients treated for UTI were defined as those with a positive urine culture who were prescribed antibiotic therapy during hospitalisation. Patients treated for CAP were defined as those with a discharge diagnostic code of pneumonia who received antibiotic treatment on day 1 or 2 of hospitalisation. Patients treated for concomitant infections (identified via case review) were excluded. For each included patient, trained abstractors (training details in online supplemental appendix) collected detailed medical record data including symptoms, diagnostic testing, antibiotic treatment and (from January 2018 on) antibiotic prescriber specialty.^{2 5}

Consistent with national guidelines, we considered patients to be overdiagnosed with UTI if they lacked signs or symptoms of UTI (ie, had asymptomatic bacteriuria)^{2 8} and to be overdiagnosed with CAP if they did not meet published diagnostic criteria for CAP (detailed definitions in online supplemental appendix).^{11 12} To ensure we were accurately defining overdiagnosis, we conducted two additional layers of review. First, we notified the treating hospital of all cases we considered overdiagnosed. Overdiagnosis cases were expected to be reviewed by treating hospital experts. In HMS’s early years, this expert review occasionally found errors in our case classification that informed modifications to overdiagnosis definitions. Second, a small subset (~1%) of cases within areas of known uncertainty (eg, pneumonia treated with <3 days of antibiotics, asymptomatic bacteriuria treated without antibiotics) underwent formal implicit review by two to three physicians, during which patient data were reviewed independently (interrater reliability assessed using ĸ statistic) and then jointly until consensus was achieved (details in online supplemental appendix). This process, again, allowed minor modifications to our operationalised definitions to the final form presented here.

The primary outcome was the proportion of patients overdiagnosed with UTI and CAP, defined as the number of patients treated for UTI/CAP who were considered overdiagnosed out of all patients treated for UTI/CAP. As emergency medicine (EM) clinicians often serve as initial diagnosticians on presentation, we also assessed overdiagnosis of UTI versus CAP for patients started on antibiotics by an EM clinician. Because national guidelines recommend reassessing antibiotic need (and diagnostic accuracy) 48–72 hours after initiation,¹³ we also include, as a secondary outcome, the proportion of patients initiated on antibiotics by an EM clinician who were still on antibiotics on day 3 of hospitalisation.

Differences between patients accurately diagnosed versus overdiagnosed were compared using Wilcoxon scores (rank-sum) for continuous or χ² test for categorical variables. Hospital proportions of patients overdiagnosed with UTI versus CAP were assessed using Pearson’s correlation coefficient (r), with correlation strength assessed using standard values (online supplemental appendix).¹⁴ SAS V.9.4.s was used for analyses. P<0.05 was statistically significant.

RESULTS

The study included 24 483 patients from 46 hospitals. Among 10 398 patients treated for UTI, 27.8% (n=2889) lacked signs or symptoms of a UTI and thus were considered overdiagnosed. Of 14 085 patients treated for CAP, 11.4% (n=1602) failed to meet diagnostic criteria and thus were considered overdiagnosed (online supplemental eTable). On implicit review, reviewers agreed with overdiagnosis classifications in 92% (n=23/25) of UTI (ĸ=0.72) and 94% (n=16/17) of CAP cases (ĸ=0.72).

Across hospitals, the proportions of patients misdiagnosed with UTI and CAP ranged from 11.0% to 44.6% and from 3.6% to 27.8%, respectively. The hospital proportion of patients overdiagnosed with UTI versus CAP was moderately correlated (r=0.53, p<0.001) (figure 1A).

Figure 1.

Overdiagnosis of urinary tract infection (UTI) and community-acquired pneumonia (CAP), by hospital. Each dot represents one hospital. The solid line represents the regression equation for the association of overdiagnosis of UTI and CAP at the hospital level. (A) There is a moderate positive correlation between the proportion of patients overdiagnosed with UTI and the proportion of patients overdiagnosed with CAP at the hospital level. (B) In a subgroup analysis of just patients started on antibiotics by an emergency medicine clinician, there is a low positive correlation between the proportion of patients overdiagnosed with UTI and CAP at the hospital level.

Prescriber data were available for 18 308 patients (7723 UTI; 10 657 CAP). Among patients overdiagnosed with UTI and CAP, antibiotics were started by EM clinicians in 55.6% (n=1164/2093) and 76.1% (n=885/1163) of cases, respectively. EM-related overdiagnosis of UTI had a low positive correlation with EM-related overdiagnosis of CAP (r=0.46, p=0.002) (figure 1B). The majority of patients started on antibiotics by an EM clinician remained on antibiotics on day 3 of hospitalisation (81.8% (n=952/1164) overdiagnosed with UTI; 89.9% overdiagnosed with CAP (n=796/885)).

DISCUSSION

In a 46-hospital, retrospective cohort study, we found overdiagnosis of UTI to have a moderate positive correlation with overdiagnosis of CAP at the hospital level. That overdiagnoses of both UTI and CAP were correlated may indicate underlying hospital-level contributors, which in turn may represent targets for intervention. Furthermore, despite national guidelines recommending clinicians reconsider antibiotic necessity and diagnosis at 48–72 hours of hospitalisation, more than 80% of overdiagnosed patients started on antibiotics by EM clinicians remained on antibiotics on hospital day 3. This persistence of unnecessary antibiotic therapy suggests ‘diagnostic momentum’ (ie, accepting a previous diagnosis without sufficient scepticism) was prominent in both diseases.

Study limitations include the observational and retrospective nature of our study, the limitations of correlational versus causal data, an inability to assess reasons for overdiagnosis correlation, and the inclusion of Michigan hospitals alone. Furthermore, our methodology for case identification differed for UTI and CAP, which may limit interpretation of relative rates of overdiagnosis. Study strengths include size, hospital diversity and ability to robustly classify overdiagnosis across diseases states.

In conclusion, overdiagnoses of UTI and CAP are correlated both at the emergency and hospital level. While cognitive contributions to overdiagnosis, including diagnostic momentum, are evident, the hospital correlation between diseases suggests important hospital-level contributors. Identification and targeting of these hospital contributors may help improve diagnosis of multiple infectious diseases.