Purpose: To assess interobserver agreement (IOA) in the diagnosis of pulmonary infiltrates on chest x-rays for patients with community-acquired pneumonia (CAP).
Materials and methods: From 7/2002 to 12/2005, 806 adults with CAP were included in the multicenter study "CAPNETZ" (7 hospitals). Inclusion criteria were clinical signs of pneumonia and pulmonary opacification on chest x-rays. Each x-ray was reevaluated by two radiologists from the university hospital in consensus reading against the interpreter at the referring hospital in regard to: presence of infiltrate (yes/no/equivocal), transparency (<or=/>50%), localization, and pattern of infiltrates (alveolar/interstitial). The following parameters were documented: digital or film radiography, hospitalization, fever, findings of auscultation, microbiological findings.
Results: The overall IOA concerning the detection of infiltrates was 77.7% (n=626; CI 0.75-0.81), the infiltrates were not verified in 16.4% (n=132) by the referring radiologist with equivocal findings in 5.9% (n=48). The IOA of the different clinical centers varied between 63.2% (n=38, CI 0.48-0.78) and 92.3% (n=65, CI 0.86-0.99). The IOA for the diagnosis of infiltrates was significantly higher for inpatients with 82.6% (n=546; CI 0.80-0.85) than for outpatients with 55.2 % (n=80; CI 0.47-0.63), p<0.0001. The IOA of infiltrates with a transparency >50% was 95.1% (n=215; CI 0.92-0.98) versus 80.4% (n=403; CI 0.77-0.84) for infiltrates with a transparency >50% (p<0.0001). In patients with positive auscultation, the IOA was higher (p=0,034). Chest x-rays of patients with antibiotic therapy or an alveolar infiltrate showed more equivocal findings compared to patients without these features.
Conclusion: There is considerable interobserver variability in the diagnosis of pulmonary infiltrates on chest radiographs. The IOA is higher in more opaque infiltrates, positive auscultation and inpatients.