Ali A El Solh1, Zaher Aldik, Moutaz Alnabhan, Brydon Grant. 1. Western New York Respiratory Research Center, Department of Medicine, Division of Pulmonary, Critical Care, and Sleep Medicine, University at Buffalo, School of Medicine and Biomedical Sciences, 462 Grider Street, Buffalo, NY, USA. solh@buffalo.edu
Abstract
BACKGROUND: Mathematical formulas have been less than adequate in assessing the optimal continuous positive airway pressure (CPAP) level in patients with obstructive sleep apnea (OSA). The objectives of the study were (1) to develop an artificial neural network (ANN) using demographic and anthropometric information to predict optimal CPAP level based on an overnight titration study and (2) to compare the predicted pressures derived from the ANN to the pressures computed from a previously described regression equation. METHODS: A general regression neural network was used to develop the predictive model. The derivation cohort included 311 consecutive patients who underwent CPAP titration at a University-affiliated Sleep Center. The model was validated subsequently on 98 participants from a private sleep laboratory. RESULTS: The correlation coefficients between the optimal pressure determined by the titration study and the predicted pressure by the ANN were 0.86 (95% confidence interval [CI] 0.83-0.88; p<0.001) for the derivation cohort and 0.85 (95% CI 0.78-0.9; p<0.001) for the validation cohort, respectively. Whereas there was no significant difference between the optimal pressure obtained during overnight polysomnography and the predicted pressure estimated by the ANN (p=0.4), the estimated pressure derived from the regression equation underestimated the optimal pressure in both the derivation and the validation group, respectively. CONCLUSION: The optimal CPAP level predicted by the ANN provides a more accurate assessment of the pressure derived from the historic regression equation.
BACKGROUND: Mathematical formulas have been less than adequate in assessing the optimal continuous positive airway pressure (CPAP) level in patients with obstructive sleep apnea (OSA). The objectives of the study were (1) to develop an artificial neural network (ANN) using demographic and anthropometric information to predict optimal CPAP level based on an overnight titration study and (2) to compare the predicted pressures derived from the ANN to the pressures computed from a previously described regression equation. METHODS: A general regression neural network was used to develop the predictive model. The derivation cohort included 311 consecutive patients who underwent CPAP titration at a University-affiliated Sleep Center. The model was validated subsequently on 98 participants from a private sleep laboratory. RESULTS: The correlation coefficients between the optimal pressure determined by the titration study and the predicted pressure by the ANN were 0.86 (95% confidence interval [CI] 0.83-0.88; p<0.001) for the derivation cohort and 0.85 (95% CI 0.78-0.9; p<0.001) for the validation cohort, respectively. Whereas there was no significant difference between the optimal pressure obtained during overnight polysomnography and the predicted pressure estimated by the ANN (p=0.4), the estimated pressure derived from the regression equation underestimated the optimal pressure in both the derivation and the validation group, respectively. CONCLUSION: The optimal CPAP level predicted by the ANN provides a more accurate assessment of the pressure derived from the historic regression equation.
Authors: Thays Crosara Abrahão Cunha; Thais Moura Guimarães; Fernanda R Almeida; Fernanda L M Haddad; Luciana B M Godoy; Thulio M Cunha; Luciana O Silva; Sergio Tufik; Lia Bittencourt Journal: Braz J Otorhinolaryngol Date: 2018-12-12