Using self-reported questionnaire data to prioritize OSA patients for polysomnography.

Abstract

Many laboratories have large numbers of patients with suspected obstructive sleep apnea (OSA) waiting to be tested. We assessed the use of simple clinical data to detect those patients with an apnea index <20 (low AI) who could be studied less emergently. Using questionnaires completed by patients prior to evaluation, we collected data on 354 consecutive patients (281 males, 73 females; mean age 48.6 years) referred for OSA and assessed with polysomnography (PSG). The questionnaires included the Epworth sleepiness scale (ESS), height, weight, age, and a history of observed apnea. Analysis of receiver operating characteristics curves revealed that both body mass index (BMI) [area under curve = 0.7258, standard error (SE) = 0.03, p < 0.01] and ESS (area under curve = 0.5581, SE = 0.03, p = 0.03) were significantly better than chance alone in detecting people with AI < 20. ESS < or =12 was found in 37.9% of the subjects but 39.6% of those expected to have a low AI using ESS had an AI > or =20. A BMI < or =28 was found in 24.9% of the subjects; 14.8% of those expected to have a low AI using BMI had an AI > or =20. Combining these variables improved accuracy but resulted in smaller groups; a cut-off of ESS < or =12 and BMI < or =28 resulted in a group of 33 (9.3% of subjects), only two (6%) of whom were falsely called low AI. Adding to this the fact that apnea had not been observed resulted in a group of nine patients (2.5% of subjects), none of whom had an AI > or =20. Thus there is a tradeoff; the more variables used, the greater the accuracy but the smaller the percent of cases selected to have low AI. However, in laboratories with hundreds of patients waiting to be tested, any procedure better than chance to help prioritize patients seems worthwhile.