<sec id="s1">Introduction:Acute respiratory failure (ARF) is a common medical emergency. Pre-hospital management includes controlled oxygen therapy, supplemented by specific management options directed at the underlying disease. The aim of the current study was to characterise the accuracy of paramedic diagnostic assessment in acute respiratory failure. </sec> <sec id="s2">Methods:A nested diagnostic accuracy and agreement study comparing pre-hospital clinical impression to the final hospital discharge diagnosis was conducted as part of the ACUTE (Ambulance CPAP: Use, Treatment effect and Economics) trial. Adults with suspected ARF were recruited from the UK West Midlands Ambulance Service. The pre-hospital clinical impression of the recruiting ambulance service clinician was prospectively recorded and compared to the final hospital diagnosis at 30 days. Agreement between pre-hospital and hospital diagnostic assessments was evaluated using raw agreement and Gwets AC1 coefficient. </sec> <sec id="s3">Results:77 participants were included. Chronic obstructive pulmonary disease (32.9%) and lower respiratory tract infection (32.9%) were the most frequently suspected primary pre-hospital diagnoses for ARF, with secondary contributory conditions recorded in 36 patients (46.8%). There was moderate agreement between the primary pre-hospital and hospital diagnoses, with raw agreement of 58.5% and a Gwets AC1 coefficient of 0.56 (95% CI 0.43 to 0.69). In five cases, a non-respiratory final diagnosis was present, including: myocardial infarction, ruptured abdominal aortic aneurysm, liver failure and sepsis. </sec> <sec id="s4">Conclusions:Pre-hospital assessment of ARF is challenging, with limited accuracy compared to the final hospital diagnosis. A syndromic approach, providing general supportive care, rather than a specifically disease-orientated treatment strategy, is likely to be most appropriate for the pre-hospital environment. </sec>