Integrated Use of Conventional Chest Radiography Cannot Rule Out Acute Aortic Syndromes in Emergency Department Patients at Low Clinical Probability.

Abstract

Guidelines recommend chest radiography (CR) in the workup of suspected acute aortic syndromes (AASs) if the pretest clinical probability is low. However, the diagnostic impact of CR integration for the rule-in and rule-out of AASs is unknown. We performed a secondary analysis of the ADvISED multicenter study. Emergency department outpatients were eligible if an AAS was clinically suspected. Clinical probability was defined with the aortic dissection detection risk score (ADD-RS). CR was evaluated blindly by a radiologist, who judged on mediastinum enlargement (ME) and other signs. In 2014 through 2016, a total of 1,129 patients were enrolled and 1,030 were analyzed, including 48 (4.7%) with AASs. ADD-RS/ME and ADD-RS/any CR sign (aCRs) integration were more accurate than ADD-RS alone (area under the curve= 0.8 and 0.78 vs. 0.66, p<0.001). The sensitivity and specificity of the integrated strategies were 66.7% (95% confidence interval [CI]= 51.5% to 79.9%) and 82.5% (95% CI= 79.9% to 84.8%) for ADD-RS/ME and 68.8% (95% CI= 53.6% to 80.9%) and 76.5% (95% CI= 73.7% to 79.1%) for ADD-RS/aCRs, respectively. The sensitivity and specificity of CR per se were 54.2% (95% CI= 39.2% to 68.6%) and 92.4% (95% CI= 90.5% to 93.9%) for ME and 60.4% (95% CI= 45.3% to 74.2%) and 85.2% (95% CI= 82.9% to 87.4%) for aCRs. The agreement (κ) between attending physicians and radiologists for ME was 0.44 (95% CI= 0.35 to 0.54). ADD-RS/ME rule-in (ADD-RS ≤ 1 and ME-present, or ADD-RS>1) applied to 204 versus 130 patients with ADD-RS>1, including 14 with AAS and 60 false-positives (FP). ADD-RS/aCRs rule-in (ADD-RS ≤ 1 and aCRs-present, or ADD-RS>1) applied to 264 patients, including 15 with AAS and 119 FP. ADD-RS/ME rule-out (ADD-RS ≤ 1 and ME-absent) applied to 826 (80.2%) patients, including 16 with AAS (33.3% of cases). ADD-RS/aCRs rule-out (ADD-RS ≤ 1 and aCRs-absent) applied to 766 patients (74.4%), including 15 with AAS (31.3% of cases). CR integration with clinical probability assessment showed modest rule-in efficiency and insufficient sensitivity for conclusive rule-out.