Prediction of obstructive sleep apnea in patients presenting to a tertiary care center

Abstract

The objective of this prospective observational clinical study is to derive and validate a diagnostic model for prediction of obstructive sleep apnea (OSA) in subjects presenting with non-sleep-related complaints in a tertiary care center in north India. We included 102 subjects (group I, range 31-70 years) presenting to the hospital with non-sleep-related complaints. None of the subjects had any significant comorbid illness such as respiratory or congestive cardiac failure. All subjects underwent detailed evaluation including polysomnography (PSG). Various parameters were compared between the cases (apnea-hypopnea index, AHI ≥ 15/h) and controls (AHI < 15/h). Using multivariate logistic regression analysis, a diagnostic model for prediction of OSA was derived. Subsequently, using similar selection criteria, 104 subjects (group II, range 32-68 years) were included for validation of the newly derived diagnostic model. Body mass index [BMI; OR (95% CI), 1.14(1.1-1.2)], male gender 5.0(1.4-27.1), relative-reported snoring index (SI) 2.8(1.7-5.0), and choking index (ChI) 8.1(1.4-46.5) were significant, independent predictors of OSA. Diagnostic model was computed as score = [1.61 × (gender)] + [1.01 × (SI)] + [2.09 × (ChI)] + [0.1 × (BMI )] where, gender: 0=female, 1=male and SI, ChI, BMI are actual values. The diagnostic model had an area under the receiver operator characteristics curve of 89.6%. A cutoff of 4.3 for the score was associated with sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of 91.3, 68.5, 70.5, and 92.3%, respectively. Misclassification rate with the application of the diagnostic model on group II subjects was 13.5% (14/104). Sensitivity, specificity, PPV, and NPV of the model for predicting OSA in this group were 82, 90.7, 89.1, and 84.5%, respectively. BMI, male gender, SI, and ChI are independent predictors of OSA. Diagnostic model derived from these parameters is useful for predicting presence of OSA and screening subjects for PSG.