Diagnostic accuracy of a two-stage model for detecting obstructive sleep apnoea in chronic tetraplegia

Abstract

BackgroundObstructive sleep apnoea (OSA) is highly prevalent in people with spinal cord injury (SCI). Polysomnography (PSG) is the gold-standard diagnostic test for OSA, however PSG is expensive and frequently inaccessible, especially in SCI. A two-stage model, incorporating a questionnaire followed by oximetry, has been found to accurately detect moderate to severe OSA (MS-OSA) in a non-disabled primary care population. This study investigated the accuracy of the two-stage model in chronic tetraplegia using both the original model and a modified version for tetraplegia.MethodsAn existing data set of 78 people with tetraplegia was used to modify the original two-stage model. Multivariable analysis identified significant risk factors for inclusion in a new tetraplegia-specific questionnaire. Receiver operating characteristic (ROC) curve analyses of the questionnaires and oximetry established thresholds for diagnosing MS-OSA. The accuracy of both models in diagnosing MS-OSA was prospectively evaluated in 100 participants with chronic tetraplegia across four international SCI units.ResultsInjury completeness, sleepiness, self-reported snoring and apnoeas were included in the modified questionnaire, which was highly predictive of MS-OSA (ROC area under the curve 0.87 (95% CI 0.79 to 0.95)). The 3% oxygen desaturation index was also highly predictive (0.93 (0.87–0.98)). The two-stage model with modified questionnaire had a sensitivity and specificity of 83% (66–93) and 88% (75–94) in the development group, and 77% (65–87) and 81% (68–90) in the validation group. Similar results were demonstrated with the original model.ConclusionImplementation of this simple alternative to full PSG could substantially increase the detection of OSA in patients with tetraplegia and improve access to treatments.Trial registration numberResults, ACTRN12615000896572 (The Australian and New Zealand Clinical Trials Registry) and pre-results, NCT02176928 (clinicaltrials.gov).