Overnight polysomnography (PSG) poses logistical and financial challenges in resource-constrained settings for the diagnosis of OSA. This study aimed to develop a simple, effective screening tool for identifying individuals at risk for OSA. Data from 1015 healthy volunteers enrolled in the BLESS cohort were analysed, all of whom underwent Level I PSG. Optimal thresholds for selected demographic, anthropometric and clinical variables were determined using a bootstrapped matrix maximisation method, with apnoea-hypopnoea index (AHI) as the reference. The dataset was split into training (n = 732) and testing (n = 244) subsets, and four predictive models were developed. A scoring system, termed SONA, was derived from coefficients of the best-performing model. Its performance was compared to existing tools. The final model incorporated four variables (AIC = 699.21, BIC = 726.79, pseudo-R2 = 0.36): waist circumference (> 93 cm in males, > 85 cm in females) and neck circumference (> 37 cm in males, > 32 cm in females) assigned 3 points each; age (> 35 years in males, > 45 years in females) assigned 2 points; and presence of snoring assigned 3 points. The total score ranged from 0 to 11. Diagnostic cut-offs were identified at scores of 5 (i.e., any two variables affirmative) in community settings and 8 (i.e., any three variables affirmative) in hospital settings. At AHI cut-off of 15, the SONA score demonstrated superior discrimination (AUC = 0.83; p value < 0.001) compared to STOPBANG (0.72), NoSAS (0.77), GOAL (0.74) and the Berlin Questionnaire (0.70). The SONA score offers a simpler, more accurate alternative for OSA screening compared to existing tools and is well-suited for implementation in primary care and resource-limited settings.
Joshi et al. (Sat,) studied this question.