Machine Learning-Based Spirometry Reference Values for the Iranian Population: A Cross-Sectional Study from the Shahedieh PERSIAN Cohort

Mohammad Sadegh Loeloe ¹ Reyhane Sefidkar ¹ Seyyed Mohammad Tabatabaei ² Amir Houshang Mehrparvar ³ Sara Jambarsang ^1*

• ¹ Department of Epidemiology and Biostatistics, School of Public Health, Shahid Sadoughi University of Medical Sciences and Health Services, Yazd, Iran
• ² Clinical Research Development Unit, Imam Reza Hospital, Faculty of Medicine, Mashhad University of Medical Sciences, Mashhad, Iran., Department of Medical Informatics, Faculty of Medicine, Mashhad University of Medical Sciences, Mashhad, Iran., Mashhad, Iran
• ³ Occupational Medicine Department, Shahid Rahnemoon Hospital, Farrokhi Ave, Yazd, Iran, Occupational Medicine Department, Industrial Diseases Research Center, Shahid Sadoughi University of Medical Sciences, Yazd, Iran, Yazd, Iran

Objective: This study aimed to determine spirometric norm values for the healthy Iranian adult population and compare them with established norm equations, specifically the GLI-Caucasian and Iranian equations. Methods: During the recruitment phase of Shahedieh Prospective Epidemiological Research Studies in Iran (PERSIAN)in 2016, spirometric parameters of 998 participants were obtained. KNN regression was employed to extract reference values for spirometric parameters FEV1, FVC, FEV1/ FVC, and FEF25-75%, considering height and age as features. The performance of KNN regression was compared with conventional models used in previous studies, such as the multiple linear regression (MLR) model and Lambda-Mu-Sigma (LMS). The predicted values were compared with those obtained from GLI-Caucasian and Iranian equations. The validation criterion was mean squared error (MSE) based on 5-fold cross-validation. Results: This study included 473 females and 525 males. KNN regression provided more accurate predictions for four spirometric parameters than MLR and LMS. The MSE to predict FVC for females was 0.15988, 0.16989, and 0.16549 in KNN regression, MLR, and LMS, respectively. The present study's predictions were closer to the actual values of the reference population for four indicators compared to the prediction values using two sets of reference equations. The MSE of predicted FVC for females was 0.15977 in the present study, which was less than the Iranian equations (MSE = 0.344) and GLI-Caucasians (MSE = 0.397). Conclusion: Using a flexible machine learning approach, this study established spirometry reference values specifically for the Iranian population. Recognizing that spirometry reference values vary across different populations, the Excel calculator developed in this research can be a valuable tool in healthcare centers for assessing lung function in Iranian adults.