Community-acquired pneumonia (CAP) is a very common diagnostic consideration. Early diagnosis and administration of antibiotics can save lives. However, the clinical diagnosis is often uncertain and misdiagnosis is frequent. This leads to inappropriate treatment with unnecessary antibiotics and may obscure the real underlying diagnosis. Even chest radiographs (CXRs) demonstrating abnormalities can be misleading, and the concordance of interpretations of these infiltrates is poor, regardless of practitioner experience.1 Thus, reliance on clinical factors and CXRs may lead to misdiagnosis and mistreatment of many patients presenting with respiratory disease.
This study hypothesized that the use of early computed tomography (CT) of the chest would improve the diagnosis and subsequent management of CAP. This was a prospective, interventional study in four tertiary teaching hospitals between November 2011 and January 2013. Enrolled in this study were 319 adults > 18 years of age who presented to the emergency department (ED) with suspicion of CAP. The criteria for CAP included: new onset of systemic symptoms (sweats, chills, aches, temperature > 38°C or < 36°C) and symptoms of lower respiratory tract infection (cough, sputum, dyspnea, chest pain, or altered breath sounds). Exclusions included: pregnancy, hospice patients, inability to complete the study, CURB-65 score of 3 or higher, or the need for ICU admission. A local radiologist performed CXRs and reported findings in a standardized fashion. Multi-detector chest CT using a low-dose protocol was performed as soon as possible afterward and was similarly interpreted. At this point, the ED physician completed a clinical report assigning a pneumonia probability and treatment plan.
Three independent evaluators who were experts in pulmonary medicine, infectious disease, or radiology subsequently adjudicated each assessment and defined the likelihood of CAP based on clinical and radiographic data and assigned a probability of CAP (definite, probable, possible, or excluded). Each case was then re-evaluated using data from the time of clinical discharge up to day 28 and assigned a final probability category.
After the initial clinical assessment and chest radiograph, the percentages of patients assigned to the CAP probability categories of definite, probable, possible, or excluded were 44.9%, 36.9%, 16.9%, and 1.2%, respectively. After chest CT, these categories shifted to 50.9%, 10.9%, 9.4%, and 28.8%, respectively. Adjudicating committee assignments were 47%, 8.7%, 11.3%, and 32.9%, respectively. At the 28-day final adjudication, the distribution was 47%, 4.1%, 10.7%, and 38.2%, respectively. Interestingly, of the 120 patients without parenchymal infiltrates on CXR, 40 had infiltrates on CT that conventional CXR missed. Conversely, of the 188 patients with parenchymal infiltrates on CXR, CT scans excluded CAP in 56 patients. Based on these CT findings, the ED physician modified the probability of CAP diagnosis in 187 of the patients (58.6%; 95% confidence interval, 53.2-64.0). Of these patients, 59 were upgraded and 128 downgraded based on CT, including 11 of 36 patients previously considered as definite CAP by CXR.
This was an intriguing study that clearly showed the limitations of clinical factors and CXRs to diagnose CAP. More than half (58.6%) of the pre-CT probabilities of CAP were altered after chest CT. Prior to chest CT, 64.7% of patients were intended to start antibiotic therapy and after chest CT, researchers ended the administration of antibiotics in 29 patients. Furthermore, 51 patients who did not receive antibiotics after CXRs were then administered therapy after receiving a CT scan. Three pulmonary emboli were discovered, and cardiac failure was diagnosed in 11 patients. Furthermore, 45 patients had a change in level of care, including 22 outpatients being admitted and 23 admissions changed to discharges. Overall, modifications of antibiotics or site of care occurred in 60.8% of patients.
It appears that most of the changes in diagnostic probability were in marginal cases. The percentage of “probable” CAP cases decreased with progressive assessments from 36.9% with CXR and clinical suspicion alone, to 10.9% after CT, to 8.7% after committee adjudication, and to 4.1% at 28 days. The number of “possible” cases decreased from 16.9% to 9.4% with chest CT. In a univariate secondary analysis, among 188 out of 308 patients with an infiltrate on CXR, CT excluded CAP in 56 patients. These patients, compared to the 132 with infiltrates confirmed on CT, tended to be older (71.1 vs 63.2 years of age; P = 0.0131), have lower white blood cell (WBC) counts (10.2 vs 12.6 x 103/mm3; P = 0.283) and lower C-reactive protein (CRP) levels (78 vs 163.3 mg/L; P = 0.0074). Conversely, among 120 patients without infiltrates on CXR who also had a CT, 40 patients had CT infiltrates compatible with CAP. Compared to the 80 patients without CT infiltrates, those 40 patients with CT infiltrates were more likely to have crackles on exam (48.7% vs 26.6%; P = 0.0169), higher WBC counts (12.3 vs 10.2 x 103/mm3; P = 0.0387), and higher CRP levels (138.1 vs 59.9 mg/L; P = 0.0037). Thus, clinical factors such as lung auscultation and CRP still seem to have a good predictive value for CAP.
Ultimately, the decision to use CT scanning for the diagnosis of CAP will require a thorough analysis of the cost of care and the outcomes data. It may very well be that the improved diagnostic accuracy of CT scanning will reduce the cost of care enough to offset the cost of additional CT scans. Furthermore, earlier administration of antibiotics with CT-confirmed CAP and prevention of unnecessary antibiotics in those without CAP might also improve health outcomes. Both these factors should be prospectively examined before entertaining the widespread adoption of routine CT for the diagnosis of CAP.
- Hopstaken RM, et al. Inter-observer variation in the interpretation of chest radiographs for pneumonia in community-acquired lower respiratory tract infections. Clin Radiol 2004;59:743-752.