Multivariate adaptive regression splines based prediction of peak expiratory flow with spirometric data

Area of Science:

• Pulmonary Medicine
• Biostatistics
• Medical Informatics

Background:

• Spirometry is crucial for assessing lung function and diagnosing respiratory disorders.
• Patient cooperation and effort can impact spirometry test accuracy, potentially leading to misdiagnosis.
• Accurate estimation of spirometric parameters is vital for effective clinical decision-making.

Purpose of the Study:

• To develop a predictive model for estimating Peak Expiratory Flow (PEF) using Multivariate Adaptive Regression Splines (MARS).
• To identify key spirometric and demographic features influencing PEF.
• To enhance the reliability of spirometric investigations in clinical practice.

Main Methods:

• Utilized flow-volume data from 220 subjects.
• Applied the Multivariate Adaptive Regression Splines (MARS) technique for model development.
• Evaluated model performance using coefficient of determination (R2) and Root Mean Squared Error (RMSE).

Main Results:

• The MARS model successfully predicted PEF, incorporating significant features like FEV1, FEF50, FEF25, and weight.
• Statistical evaluation showed a good model fit with a low bias, as indicated by Bland-Altman plots.
• The model demonstrated the ability to capture essential spirometric features for accurate PEF estimation.

Conclusions:

• The developed MARS model provides a reliable method for estimating PEF, addressing limitations of standard spirometry.
• Key spirometric parameters (FEV1, FEF50, FEF25) and weight are significant predictors of PEF.
• This approach can support clinicians in improving the diagnostic accuracy of respiratory disorder investigations.