Clinical prediction of the sleep apnea syndrome

Polysomnography, the standard diagnostic test for people suspected of having sleep apnea, is a limited resource due to its expense. Decisions about which patients to refer to a sleep center and which require polysomnography can be made based on an estimate of the probability that they have sleep apnea. Clinical features that are associated with the severity of sleep apnea, as judged by the apnea-hypopnea index, can be combined together using statistical modeling into a clinical prediction rule, whose diagnostic performance can be summarized by its sensitivity and specificity or by likelihood ratios. To date, at least seven different sleep apnea clinical prediction rules have been developed, most incorporate anthropomorphic variables such as the body mass index, waist circumference, and/or neck circumference, and some type of abnormal respiration during sleep (snoring, apneas, choking and/or gasping) witnessed by a bed partner. In general these rules have reasonably high sensitivities but only intermediate specificities, thus they can be useful in excluding the diagnosis but do not usually raise the probability of sleep apnea high enough to warrant initiating therapy without at least some type of additional testing to confirm the diagnosis. In isolation the apnea-hypopnea index is not an optimal indicator of disease severity, thus ultimately clinical decisions about the need for polysomnography and/or the need for treatment must take into account other important clinical information such as symptom severity, quality of life, and the presence or absence of comorbid illness.