The No-Apnea score for early obstructive sleep apnea detection in a sleep clinic: a study of diagnostic accuracy and comparative performance of three screening instruments

This cross-sectional study investigated the No-Apnea tool's diagnostic ability to detect obstructive sleep apnea (OSA), across different severities (respiratory disturbance index [RDI] ≥ 5, ≥ 15, ≥ 30) compared to polysomnography (gold standard) in 2,087 consecutive patients attending a sleep clinic (age: 45.92 ± 13.46, male: 70.8%). We further compared the diagnostic performance of No-Apnea, NoSAS, and STOP-Bang tools. Across all RDI levels, No-Apnea showed superior sensitivity (96.9-99.0%, based on RDI thresholds), negative predictive value (NPV) (47.8-91.2%), negative likelihood ratio (0.10-0.22), and odds ratio (10.6-11.8). Across all RDI levels, NoSAS demonstrated the highest specificity (72.2-87.8%), positive predictive value (PPV) (67.5-97.4%), and positive likelihood ratio (2.21-3.87). No-Apnea achieved the highest accuracy and F1 scores at RDI ≥ 5 (90.37 and 94.8%) and ≥ 15 (74.75 and 84.8%). At RDI ≥ 5, NoSAS and STOP-Bang showed comparable areas under the ROC curve (AUC) (0.675 vs. 0.685, P > 0.05), exhibiting a significant advantage over No-Apnea (0.622, P = 0.004 and 0.002, respectively). At RDI ≥ 15 and ≥ 30, NoSAS had the highest (0.676 and 0.668), and No-Apnea had the lowest AUCs (0.566 and 0.543). STOP-Bang performance measures fell between No-Apnea and NoSAS across RDI levels. In our sleep clinic population, No-Apnea demonstrated high sensitivity, NPV, accuracy, and F1 score for detecting OSA, but low specificity, PPV, and AUC.