Patients with obesity hypoventilation syndrome (OHS) have a lower quality of life, more healthcare expenses, a greater risk of pulmonary hypertension, and a higher mortality compared to eucapnic patients with obstructive sleep apnea (OSA). Despite significant morbidity and mortality associated with OHS, it is often unrecognized and treatment is frequently delayed. The objective of this observational study was to determine the prevalence of OHS in patients with OSA seen at the sleep disorders clinic of a large public urban hospital serving predominantly minority population and to identify clinical--not mechanistic--predictors that should prompt clinicians to measure arterial blood gases. In the first stage, we randomly selected 180 patients referred to our sleep disorders clinic between 2000 and 2004 for suspicion of OSA. From this retrospective random sample we calculated the prevalence of OHS in patients with OSA and identified independent clinical predictors using logistic regression. In the second stage, we prospectively validated these predictors in a sample of 410 consecutive patients referred to the sleep disorders clinic for suspicion of OSA between 2005 and 2006. The prevalence of OHS in patients with OSA was 30% in the retrospective random sample and 20% in the prospective sample. Three variables independently predicted OHS in both samples: serum bicarbonate level (p < 0.001), apnea-hypopnea index (p = 0.006), and lowest oxygen saturation during sleep (p < 0.001). Due to the serious morbidity associated with OHS, we selected a highly sensitive threshold of serum bicarbonate level. A threshold of 27 mEq/l had a sensitivity of 92% and a specificity of 50%. Only 3% of patients with a serum bicarbonate level <27 mEq/l had hypercapnia compared to 50% with a serum bicarbonate > or =27 mEq/l. In conclusion, OHS is common in severe OSA. A normal serum bicarbonate level excludes hypercapnia and an elevated serum bicarbonate level should prompt clinicians to measure arterial blood gases.