Disagreement in the interpretation of chest radiographs among specialists and clinical outcomes of patients hospitalized with suspected pneumonia

Abstract

Background

Pneumonia is diagnosed by a combination of clinical symptoms and findings on chest X-ray (CXR). However, there is often disagreement, even among experts, upon the interpretation of the CXR. The purpose of this study was to compare the agreement rates in CXR interpretation of suspected community-acquired pneumonia (CAP) between a radiologist, a pulmonologist, an infectious disease specialist, and an internal medicine staff and to establish the correlation of such an agreement with the length of hospitalization and 30-day mortality rate.

Methods

We prospectively enrolled in our study all patients admitted to the Department of Medicine with suspected CAP, as defined by the admitting physician, within the first 24 h of hospitalization. A radiologist, pulmonologist, and infectious disease specialist who were aware of the suspected diagnosis independently interpreted the CXR. The final diagnosis was obtained from the discharge notes.

Results

A total of 262 patients participated in the study, 214 of whom (81.7%) were eventually discharged with a diagnosis of CAP. The agreement rates between the readers of the CXR ranged from a kappa of 0.09 to 0.44. There were no differences in terms of background illness, PORT (Pneumonia Patients Outcomes Research Team) score, length of hospitalization, or mortality rates between patients discharged with or without a diagnosis of CAP. In multivariate analysis, only the PORT score was a significant predictor of length of hospitalization and mortality rate.

Conclusion

We found a low to moderate agreement rate of the diagnosis of CAP between CXR readers. Identification of an infiltrate on CXR, either by specialists or by the attending physician, did not impact the clinical outcomes.

Introduction

Community-acquired pneumonia (CAP) in immunocompetent patients is so common that several professional societies have issued clinical guidelines with recommendations on how to manage these patients [1], [2], [3], [4], [5]. These guidelines are not very explicit with regard to how to diagnose CAP or to the potential pitfalls in diagnosing CAP [3], [4], [5].

The gold standard for diagnosing CAP is to demonstrate an infectious pathogen from lung tissue obtained through a sterile pathway [6]. However, this is rarely done unless the patient is suspected of suffering from a complicated para-pneumonic effusion mandating evacuation. Even then, obtaining an etiological microbiologic agent may be hindered by prior antibiotic therapy [6]. Instead, clinicians diagnose CAP based on a combination of systemic and respiratory complaints, usually accompanied by fever and confirmed by an infiltrate on CXR that resolves 6–10 weeks later [1], [2], [6], [7], [8], [9]. Computed tomography (CT) has also been shown to be more sensitive for the diagnosis of CAP than CXR [10].

Physical examination is an integral part of the clinical evaluation that every clinician performs; however, when experienced physicians (a general internist, an infectious and a pulmonary sub-specialist) compared their objective findings of CXRs of patients with and without CAP, the correlation between their findings was only in moderate agreement (60–72%) [7], [11]. In clinical practice, after obtaining a history from the patient and performing a physical examination, the physician decides that the acute presentation of the patient represents a suspected CAP; he then requests a CXR for confirmation [6], [7]. Thus, the CXR is, in fact, the clinical standard that takes the place of the above-mentioned microbiological gold standard in diagnosing CAP [6]. Moreover, the clinical condition of the patient may influence the physician's ability to interpret the findings on the CXR. For example, if the patient is dehydrated, the infiltrate may only show up after rehydration [6]. In addition, many elderly patients have findings on the CXR from old tuberculosis, congestive heart failure, or prior surgery, making it difficult to interpret the findings without a prior CXR, which may not always be available. This leads to discordance between the clinical diagnosis of CAP and the radiological diagnosis.

With these limitations in mind, we decided to compare the agreement between a radiologist with a special interest in pulmonary radiology, an infectious disease specialist, and a pulmonary disease specialist in interpreting the CXR of the patients admitted to the medical wards with suspected CAP. Moreover, we assessed whether such an agreement was related to the patients' treatment and outcomes, as one could postulate that increased agreement among the specialists could predict a worse outcome.