The objective of this retrospective study was to identify factors affecting the accuracy of pulse oximetry in the ED. Over a 3-year period, 664 consecutive emergency department (ED) patients had simultaneous arterial blood gas (ABG) and pulse oximeter readings taken. Pulse oximeter saturations (SpO₂) were compared with ABG CO-oximeter saturations (SaO₂) for accuracy. Multiple variables including age, sex, hemoglobin, bicarbonate, pH, and carboxyhemoglobin (COHb) were analyzed to see if they affected SpO₂ accuracy. ROC curves were used to determine the best pulse oximeter threshold for detecting hypoxia. Using multivariate analysis, COHb was the only statistically significant factor affecting the accuracy of pulse oximetry. In patients with COHb < 2%, SpO₂ overestimated SaO₂ by more than 4% in 8.4% of cases. In patients with COHb ≥ 2%, SpO₂ overestimated SaO₂ by more than 4% in 35% of cases. The best pulse oximetry threshold for detecting hypoxia is 92%. At this threshold, if COHb is <2%, pulse oximetry has a sensitivity of 0.92 and specificity of 0.90. If COHb is ≥2%, sensitivity is 0.74 and specificity is 0.84. For patients likely to have a COHb < 2, pulse oximetry is an effective screening tool for detecting hypoxia. However, more caution must be exercised when using pulse oximetry in patients likely to have a COHb ≥ 2%. (Am J Emerg Med 2000;18:427-431. Copyright © 2000 by W.B. Saunders Company)