No single model is sufficient
A systematic review and meta-analysis of 249 studies of sleep apnea diagnosis from 1980 through November 1997 concluded that studies of focused questionnaires were too heterogeneous to be combined in a meta-analysis. Sensitivity and specificity were good for clinical prediction rules in general, but evidence was insufficient to justify adoption of any single model.7
A 2000 prospective study of 4 previously developed clinical prediction models for OSAHS in sleep center populations found sensitivities of 75% to 96% and specificities of 13% to 54% for identifying patients, using a criterion of AHI >9.8
A 2004 prospective evaluation examined a clinical decision rule on 837 patients referred for polysomnogram that used age, sex, BMI, snoring, and cessation of breathing during sleep to stratify patients into low-, moderate-, or high-risk groups. The study found OSAHS prevalences of 8%, 51%, and 82%, respectively.9
Recommendations
The Institute for Clinical Systems Improvement recommends polysomnography for patients with symptoms of OSAHS and 1 or more of the following: cardiovascular disease, hypertension, coronary artery disease, obesity, sleep complaint, type 2 diabetes mellitus, recurrent atrial fibrillation, and large neck circumference.10
The Canadian Thoracic Society states that clinical prediction formulas assess the pretest probability of sleep-disordered breathing and prioritize patients for evaluation, but are insufficient for diagnosis. The guidelines recommend that all patients with suspected sleep-disordered breathing complete an assessment of daytime sleepiness such as the Epworth Sleepiness Scale.11
The American Academy of Sleep Medicine recommends no single clinical model to predict the severity of OSAHS. Polysomnography is indicated for diagnosis of any sleep-disordered breathing.12