BACKGROUND: Overnight polysomnography is expensive and time consuming. An approach based on a logistic regression model and overnight pulse oximetry has been developed to determine which of the snorers referred to our sleep laboratory need polysomnography. METHODS: The variables entered in the regression model were derived from questionnaires completed by 95 habitual snorers and 89 patients with obstructive sleep apnoea. The resulting regression equation included weight, height, sex, witnessed episodes of apnoea, and reports of falling asleep when reading. This prediction equation was applied to a sample of 116 consecutive patients referred for investigation of heavy snoring. Pulse oximetry data on the 116 test subjects were obtained during polysomnography and analysed separately. Pulse oximetry was judged to indicate obstructive sleep apnoea when it showed cyclic oscillations of oxyhaemoglobin saturation or heart rate, or both, for more than 30 minutes during the study night. RESULTS: A cut off probability of 0.31 gave the prediction model a sensitivity of 94% to predict an apnoea-hypopnoea index above 10, with a specificity of 45%. When this cutoff point was used to predict an apnoea-hypopnoea index of over 20 sensitivity was 95% and specificity 41%. Combined with oximetry our regression model had a sensitivity of 100% for predicting an apnoea-hypopnoea index of more than 10. On the other hand, all patients with negative results from oximetry and a probability value below 0.31 had an apnoea-hypopnoea index lower than 10 according to polysomnography. CONCLUSIONS: It is concluded that snorers with negative results from oximetry classified as not having obstructive sleep apnoea according to this model do not need polysomnography.
BACKGROUND: Overnight polysomnography is expensive and time consuming. An approach based on a logistic regression model and overnight pulse oximetry has been developed to determine which of the snorers referred to our sleep laboratory need polysomnography. METHODS: The variables entered in the regression model were derived from questionnaires completed by 95 habitual snorers and 89 patients with obstructive sleep apnoea. The resulting regression equation included weight, height, sex, witnessed episodes of apnoea, and reports of falling asleep when reading. This prediction equation was applied to a sample of 116 consecutive patients referred for investigation of heavy snoring. Pulse oximetry data on the 116 test subjects were obtained during polysomnography and analysed separately. Pulse oximetry was judged to indicate obstructive sleep apnoea when it showed cyclic oscillations of oxyhaemoglobin saturation or heart rate, or both, for more than 30 minutes during the study night. RESULTS: A cut off probability of 0.31 gave the prediction model a sensitivity of 94% to predict an apnoea-hypopnoea index above 10, with a specificity of 45%. When this cutoff point was used to predict an apnoea-hypopnoea index of over 20 sensitivity was 95% and specificity 41%. Combined with oximetry our regression model had a sensitivity of 100% for predicting an apnoea-hypopnoea index of more than 10. On the other hand, all patients with negative results from oximetry and a probability value below 0.31 had an apnoea-hypopnoea index lower than 10 according to polysomnography. CONCLUSIONS: It is concluded that snorers with negative results from oximetry classified as not having obstructive sleep apnoea according to this model do not need polysomnography.
Authors: Maria Claudia Mattos Soares; Lia Rita de Azeredo Bittencourt; Adriane Iurck Zonato; Luis Carlos Gregório Journal: Braz J Otorhinolaryngol Date: 2006 Jul-Aug