Development and validation of a model for diagnosis of obstructive sleep apnoea in primary care.

Abstract

Use of in-laboratory polysomnography (PSG) to diagnose obstructive sleep apnoea (OSA) is cost and resource intensive. Questionnaires, physical measurements and home monitors have been studied as potential simpler alternatives. This study aimed to develop a diagnostic model for OSA for use in primary care. Primary care practitioners were trained to recognize symptoms of sleep apnoea and recruited patients based on the clinical need to investigate OSA. Assessment was by symptom questionnaires, anthropomorphic measurements, digital facial photography, and a single-channel nasal flow monitor (Flow Wizard©, DiagnoseIT, Sydney, Australia) worn at home for 3 nights. The in-laboratory PSG was the reference test, with OSA defined as apnoea-hypopnoea index (AHI) ≥10 events/h. In the model development phase, 25 primary care practitioners studied 315 patients in whom they suspected OSA, of which 57% had AHI≥10 and 22% had AHI≥30. Published OSA questionnaires provided low to moderate prediction of OSA (area under the curve [AUC] 0.53-0.73). The nasal flow monitor alone yielded high accuracy for predicting OSA with AUC of 0.87. Sensitivity was 0.87 and specificity 0.77 at a threshold respiratory event index (REI) of 18 events/h. A model adding age, gender, symptoms and BMI to the nasal flow monitor REI only modestly improved OSA prediction (AUC 0.89), with similar AUC (0.88) confirmed in the validation population of 114 patients. Sleep apnoea can be diagnosed in the primary care setting with a combination of clinical judgement and portable monitor test outcomes.