Background: Obstructive sleep apnea (OSA) is a common and underdiagnosed entity that favors perioperative morbidity. Several anatomical characteristics predispose to OSA. We developed a new clinical score that would detect OSA based on the patient's morphologic characteristics only.
Methods: Patients (n = 149) scheduled for an overnight polysomnography were included. Their morphologic metrics were compared, and combinations of them were tested for their ability to predict at least mild, moderate-to-severe, or severe OSA, as defined by an apnea-hypopnea index (AHI) >5, >15, or >30 events/h. This ability was calculated using Cohen κ coefficient and prediction probability.
Results: The score with best prediction abilities (DES-OSA score) considered 5 variables: Mallampati score, distance between the thyroid and the chin, body mass index, neck circumference, and sex. Those variables were weighted by 1, 2, or 3 points. DES-OSA score >5, 6, and 7 were associated with increased probability of an AHI >5, >15, or >30 events/h, respectively, and those thresholds had the best Cohen κ coefficient, sensitivities, and specificities. Receiver operating characteristic curve analysis revealed that the area under the curve was 0.832 (95% confidence interval [CI], 0.762-0.902), 0.805 (95% CI, 0.734-0.876), and 0.834 (95% CI, 0.757-0.911) for DES-OSA at predicting an AHI >5, >15, and >30 events/h, respectively. With the aforementioned thresholds, corresponding sensitivities (95% CI) were 82.7% (74.5-88.7), 77.1% (66.9-84.9), and 75% (61.0-85.1), and specificities (95% CI) were 72.4% (54.0-85.4), 73.2% (60.3-83.1), and 76.9% (67.2-84.4). Validation of DES-OSA performance in an independent sample yielded highly similar results.
Conclusions: DES-OSA is a simple score for detecting OSA patients. Its originality relies on its morphologic nature. Derived from a European population, it may prove useful in a preoperative setting, but it has still to be compared with other screening tools in a general surgical population and in other ethnic groups.