The Clinical Utility of Chest Radiography for Identifying Pneumonia: Accounting for Diagnostic Uncertainty in Radiology Reports

Abstract

OBJECTIVE. Currently, chest radiography is the first-line imaging test for identifying pneumonia; chest CT is considered the reference standard. The purpose of this study was to calculate the statistical measures of performance of chest radiography for identifying pneumonia when taking into account uncertain results of both chest radiography and CT examinations. MATERIALS AND METHODS. Statistical measures of performance of chest radiography, using CT as the reference standard, were calculated with 95% CIs by varying uncertain radiology report impressions of both chest radiography and CT to all negative or all positive. The resulting scenarios were as follows: scenario 1, uncertain chest radiography and CT impressions are considered positive for pneumonia; scenario 2, uncertain chest radiography impressions are positive but uncertain CT impressions are negative; scenario 3, uncertain chest radiography impressions are negative and uncertain CT impressions are positive; scenario 4, uncertain chest radiography and CT impressions are negative; and scenario 5, uncertain chest radiography and CT impressions are excluded. RESULTS. A retrospective analysis of 2411 patient visits revealed the prevalence of uncertain radiology report impressions to be 31.8% for chest radiography and 21.7% for CT. Scenario 1 yielded the following performance values: sensitivity, 51.9%; specificity, 71.3%; PPV, 59.4%; and NPV, 64.5%. Scenario 2 produced the following performance values: sensitivity, 59.6%; specificity, 67.1%; PPV, 59.6%; and NPV, 67.1%. Scenario 3 showed the following performance values: sensitivity, 13.4%; specificity, 97.7%; PPV, 82.6%; and NPV, 58.1%. Scenario 4 yielded the following performance values: sensitivity, 19.6%; specificity, 96.4%; PPV, 81.6%; and NPV, 59.5%. Scenario 5 produced the following performance values: sensitivity, 32.7%; specificity, 96.8%; PPV, 89.2%; and NPV, 63.8%. CONCLUSION. Uncertain chest radiography results for the evaluation of pneumonia are prevalent. A chest radiography impression using the strongest language in support of a pneumonia diagnosis is useful to rule in pneumonia radiographically, but a negative result performs poorly at ruling out disease.

Keywords: chest radiography; performance measures; pneumonia; uncertain or intermediate results.