Comparison of clinical scores in their ability to detect hypoxemic severe OSA patients

Background: Severe obstructive sleep apnea (sOSA) and preoperative hypoxemia are risk factors of postoperative complications. Patients exhibiting the combination of both factors are probably at higher perioperative risk. Four scores (STOP-Bang, P-SAP, OSA50, and DES-OSA) are currently used to detect OSA patients preoperatively. This study compared their ability to specifically detect hypoxemic sOSA patients. Methods: One hundred and fifty-nine patients scheduled for an overnight polysomnography (PSG) were prospectively enrolled. The ability of the four scores to predict the occurrence of hypoxemic episodes in sOSA patients was compared using sensitivity (Se), specificity (Sp), Youden Index, Cohen kappa coefficient, and the area under ROC curve (AUROC) analyses. Results: OSA50 elicited the highest Se [95% CI] at detecting hypoxemic sOSA patients (1 [0.89–1]) and was significantly more sensitive than STOP-Bang in that respect. DES-OSA was significantly more specific (0.58 [0.49–0.66]) than the three other scores. The Youden Index of DES-OSA (1.45 [1.33–1.58]) was significantly higher than those of STOP-Bang, P-SAP, and OSA50. The AUROC of DES-OSA (0.8 [0.71–0.89]) was significantly the largest. The highest Kappa value was obtained for DES-OSA (0.33 [0.21–0.45]) and was significantly higher than those of STOP-Bang, and OSA50. Conclusions: In our population, DES-OSA appears to be more effective than the three other scores to specifically detect hypoxemic sOSA patients. However prospective studies are needed to confirm these findings in a perioperative setting. Clinical trial registration: ClinicalTrials.gov: NCT02050685.