Objective: To determine whether presenting clinical history, pharyngeal examination, and the overall subjective impression of the clinician could serve as a sensitive screening test for sleep apnea.
Design: Blinded comparison of history and physical examination with results of nocturnal polysomnography.
Setting: Sleep clinic of a tertiary referral center.
Patients: A total of 410 patients referred for suspected sleep apnea syndrome. Most patients reported snoring.
Measurements: All patients were asked standard questions and given an examination relevant to the diagnosis of the sleep apnea syndrome, and all had full nocturnal polysomnography. Patients with more than ten episodes of apnea or hypopnea per hour of sleep were classified as having sleep apnea. Stepwise linear logistic regression was used to develop two predictive models of sleep apnea: one based on the presence of characteristic clinical features, age, sex, and body mass index; and one based on subjective clinical impression.
Results: The prevalence of sleep apnea in our patients was 46%. Only age, body mass index, male sex, and snoring were found to be predictors of sleep apnea. The logistic rule discriminated between patients with and without sleep apnea (receiver operating characteristic [ROC] area, 0.77 [95% Cl, 0.73 to 0.82]). For patients with a predicted probability of apnea of less than 20%, the clinical model had 94% sensitivity and 28% specificity. Subjective impression alone identified correctly only 52% of patients with sleep apnea and had a specificity of 70%.
Conclusions: In patients with a high predicted probability of the sleep apnea syndrome, subjective impression alone or any combination of clinical features cannot serve as a reliable screening test. However, in patients with a low predicted probability of sleep apnea, the model based on clinical data was sufficiently sensitive to permit about a 30% reduction in the number of unnecessary sleep studies.