143EMF Chest Radiograph Findings Suggestive of Pneumonia Following Cardiac Arrest

Abstract

Over 347,000 adults suffer an out-of-hospital cardiac arrest (OOHCA) annually. Pneumonia is a common complication following cardiac arrest, the estimated incidence ranges from 21 to 67%. Diagnostic uncertainty for pneumonia arises from overlap with clinical findings of the post-cardiac arrest syndrome (eg, elevated white blood cell count, hemodynamic and respiratory compromise, tachycardia), as well as targeted temperature management which masks fever. Additionally, chest radiograph abnormalities result not only from infection but also from cardiopulmonary resuscitation and contusion, edema, aspiration, and acute respiratory distress syndrome (ARDS). Non-specific chest radiograph abnormalities result in antibiotic administration for over 60% of post-arrest patients at our institution. Untargeted antibiotic therapy can be harmful to patients by causing further organ dysfunction, antibiotic resistance and resistant infections, and increased hospital length of stay and cost. As a first step to develop a multivariable logistic regression model for diagnosing pneumonia and appropriately allocating antibiotics, we tested whether inter-rater agreement on specific chest radiograph findings in post-arrest patients was reliable. Two board certified body-imaging radiologists, blinded to the clinical data and subject identifiers, independently retrospectively reviewed initial chest radiograph from subjects 18 years of age or older admitted within 12 hours of OOHCA to a single quaternary care center in Western Pennsylvania from 2010-2015. They characterized findings according to a priori definitions consistent with standard radiology classifications and their expertise. A pre-specified abstraction form denoting abnormality, location of abnormality, and classification of abnormality as compatible with pneumonia, ARDS, aspiration, contusion, or heart failure was used. Gwet’s AC1 was used to assess inter-rater agreement. Two radiologists reviewed 917 OOHCA subjects. Mean age 58.7 (SD 16.4), 59% were male, 32% had a shockable initial rhythm, and 39% survived to hospital discharge. Pneumonia was suspected in up to 78% of subjects. Up to 78% of subjects had a suspected abnormality. Inter-rater agreement was moderate for less specific findings including: presence of abnormality, side of the thorax the abnormality was located, and distribution (cephalad to caudal direction) of the abnormality (Table 1). Specific findings compatible with pneumonia, ARDS, and heart failure demonstrated poor agreement. There was fair disagreement for findings compatible with aspiration or contusion. Chest radiograph abnormalities are common. There is moderate inter-rater agreement for the presence of an abnormality on initial chest radiograph, but poor agreement for more specific findings. A combination of more advanced imaging (chest computed tomography) along with clinical variables (microbiologic pulmonary specimens) may improve diagnosis of pneumonia in post-arrest patients.Table 1Inter-rater Agreement for Specific Chest Radiography FindingsRange (%)Gwet’s AC1Standard Error95% Confidence IntervalP-valueAbnormality642-719 (70.01-78.41)0.04330.03180.3705-0.4952<0.0001Side of abnormalityN/A0.56730.02300.5222-0.6124<0.0001Distribution of abnormalityN/A0.50910.02390.4623-0.5560<0.0001Pneumonia446-715 (48.64-77.97)0.10990.03570.0398-0.18000.002Aspiration249-711 (27.15-77.54)-0.23610.0323-0.2995 –0.1727<0.0001ARDS168-518 (18.32-56.49)0.06070.0359-0.0098-0.13120.091Heart Failure367-520 (40.02-56.71)0.10890.03290.0443-0.17350.001Contusion87-673 (9.49-73.39)-0.35190.0341-0.4187-0.2850<0.0001 Open table in a new tab