Low accuracy and high inter-assessor variability in the interpretation of chest radiographs have been well described for decades [[1], [2], [3], [4]]. Portable chest radiographs are often the first radiological investigations used to help identify the cause of respiratory deterioration in patients in the intensive care unit (ICU) because they are readily available, practical and low risk [[5], [6], [7]]. The quality of chest radiographs in the ICU is often compromised by poor patient positioning, the presence of overlying lines or tubes, and the timing of mechanical ventilation. In many hospitals, chest radiographs are requested and immediately interpreted by ICU clinicians, with formal reporting by a radiologist occurring at a later stage [8]. Often, ICU clinicians without extensive radiology training will be asked to interpret chest radiographs of relatively low technical quality. Inaccurate interpretation of chest radiographs has the potential to delay therapies or to direct treatments that patients do not require.

Thoracic computerised tomography (CT) provides higher definition imaging but is less readily available than chest radiography, involves higher doses of radiation and requires critically ill patients to be transferred to radiology departments.

Inter-observer agreement on the presence of bilateral opacities on chest radiographs from patients in the ICU has been described as moderate when comparing intensivists with other intensivists, and when comparing interpretations from intensivists and radiologists [9,10].

However, there remains a paucity of evidence around inter-observer agreement in chest radiograph evaluation in ICU. Additionally, how chest radiograph interpretation varies by specialty and training experience remains to be fully determined.

The current study sought to address this gap in the evidence by assessing agreement between chest radiograph and CT interpretation, and by evaluating inter-observer variation in chest radiograph interpretation, in ICUs. The primary objectives were to assess the accuracy of interpretation of chest radiographs from critically ill patients in ICUs, using CT scans as a reference standard, and to understand how interpretation of chest radiographs varies with experience and specialisation of the interpreter. The secondary objective was to assess the inter-observer variation in chest radiograph interpretation by experience and specialisation of interpreter.