Routine pulse oximetry has been studied to detect children with otherwise undiagnosed congenital heart disease prior to nursery discharge. The reported sensitivities in asymptomatic patients have been less than expected and vary widely, bringing into question the reliability of the test. The purpose of this study was to assess whether routine pulse oximetry contributes to identifying patients with critical congenital heart disease and to determine the reliability of a single pulse oximeter reading in screening asymptomatic newborn infants. Between December 26, 2003, and December 31, 2005, three hospitals in west central Florida performed a pulse oximetry routinely on all newborns at the time of discharge. Patients diagnosed with critical congenital heart disease during the study period were identified to assess whether the pulse oximetry reading initiated their diagnosis. In one hospital, the pulse oximeter data were evaluated for reliability. Downloaded data were compared to a log compiled by the nursery personnel, first without (phase 1) and then with (phase 2) their knowledge and additional training. Results were characterized as reliable, probe placed but reading not verifiable, or no evidence of probe placement. Of the 7962 infants who received oximetry testing, there were 12 postnatal diagnoses of critical congenital heart disease. None was initially identified by routine pulse oximetry. Pulse oximetry reliability improved substantially between phase 1 and phase 2 (38 v. 60%, p < 0.0001). Optimal reliability (>95%) was obtained by a nurse with a degree of LPN or higher performing an assessment of at least 360 seconds. Routine pulse oximetry was neither reliable nor an important diagnostic tool in our cohort. Important human factors (probe placement time, oximetry training, and nursing degree) impact single determination pulse oximetry reliability. With routine surveillance and quality improvement, the reliability of this test can be increased. Future studies to determine the effectiveness of pulse oximetry screening for the diagnosis of congenital heart disease in the asymptomatic newborn population must address these factors. Until such a study demonstrates acceptable sensitivity and clinical value, universal screening should not be instituted.