Evaluation of Indian Prediction Models for Lung Function Parameters: A Statistical Approach

Area of Science:

• Pulmonary Medicine
• Biostatistics
• Medical Informatics

Background:

• Lung function test interpretation relies on reference values from healthy populations.
• Predicted values are derived using regression models based on anthropometric and ethnic characteristics.
• Current Indian prediction models for lung function are predominantly based on traditional linear regression.

Purpose of the Study:

• To statistically evaluate existing Indian prediction models for lung function parameters.
• To critically assess the suitability of reference values for lung function in the Indian population.
• To identify limitations in current Indian lung function prediction models.

Main Methods:

• Systematic review and meta-analysis following Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines.
• Evaluation of prediction models based on modeling approach, regression diagnostics, and methodology.
• Suitability assessment using an 8-criterion checklist derived from American Thoracic Society (ATS) guidelines.

Main Results:

• 32 articles and 25,289 subjects were included. Multiple linear regression was common (27 articles).
• Regression diagnostics were reported in 22 articles, but homoskedasticity of residuals was often unexamined.
• Only 5 articles met over 7 of 8 suitability criteria; 8 met fewer than 3 criteria.

Conclusions:

• Traditional linear regression models are prevalent in Indian lung function prediction.
• There is a need for more robust prediction models utilizing advanced statistical techniques in India.
• Enhancements in computational power necessitate the development of sophisticated models for accurate lung function assessment.