OBJECTIVES: The objectives of this prospective observational study were to: (1) determine the accuracy of physician diagnosis in patients with an acutely altered mental status (AMS) within the first 20 min of emergency department (ED) presentation; and (2) access if physician confidence in early diagnosis correlates with accuracy of diagnosis. METHODS: A prospective observational convenience study was conducted of 112 adult patients who presented to an urban county ED with AMS (Glasgow Coma Scale (GCS) score ≤14) between August 2008 and July 2009. Within the first 20 min of patient presentation to the ED, treating physicians were asked to record their best diagnostic guess of the cause of the patient's AMS and their confidence in this diagnosis. Serial hourly GCS was performed and the results of all diagnostic testing were recorded. Blinded investigators determined the final consensus diagnostic cause of the patient's AMS. RESULTS: The final consensus diagnoses for AMS aetiologies were as follows: isolated alcohol intoxication 31%, other (psychotic episodes, underlying dementia) 21%, combination alcohol/other drug intoxications 18%, isolated other drug intoxications 10%, other metabolic derangements 6%, cerebrovascular accident/transient ischaemic attack 4%, seizures/post-ictal states 4%, traumatic brain injuries 3%, isolated opiate intoxications 2%, isolated benzodiazepine intoxication 1% and septic episode 1%. The emergency physician's initial diagnosis of the AMS patient correlated with the accuracy of the final diagnosis (r(2)=0.807). The quintiles of confidence of diagnosis were: 0-20% degree of confidence had a 33% diagnostic accuracy, 21-40% had 25% accuracy, 41-60% had 43% accuracy, 61-80% had 52% accuracy and those with 81-100% confidence of initial diagnosis had 78% accuracy. Of the 106 patients with an initial diagnosis, 52 (51%) had a head CT performed, with eight (8%) having an acute abnormality. DISCUSSION: Early diagnoses of AMS patients are moderately accurate. Few early misdiagnoses of AMS patients were clinically relevant. Physicians' greater degree of confidence in their diagnosis correlated with greater accuracy.