The effect on survival of episodic hypoxemia was prospectively studied in 100 patients hospitalized on general medicine services of a large, tertiary care, university hospital. Pulse oximetry monitoring (POM) was initiated within 24 hours of admission and was maintained for approximately 24 hours independent of patient management. Hypoxemia lasting for at least 5 consecutive minutes and resulting in an arterial oxygen saturation of less than 90% occurred in 26 of the 100 patients. No clinical characteristics were found that could reliably distinguish those patients who did develop hypoxemia from those who did not, though the small number of patients in many categories precludes drawing firm conclusions. However, severe desaturation was unlikely to occur in patients with normal chest roentgenograms. During the following 4 to 7 months, 8 patients (32%) suffering episodic hypoxemia died, while only 7 individuals (10%) without hypoxemia died, an increase in mortality that remained significant after adjustment for severity of illness. The relative risk of death associated with desaturation was 3.3 (95% CI 1.41 to 8.2). The severity of hemoglobin oxygen desaturation, expressed as the saturation-time index, correlated inversely with survival time.