Methodology for generating continuous prediction equations for pulmonary function measures

Area of Science:

• Biostatistics
• Mathematical Modeling
• Pulmonary Medicine

Background:

• Existing methods for creating reference equations often result in discontinuous data across age ranges.
• Accurate, continuous reference equations are crucial for interpreting lung function in diverse populations.

Purpose of the Study:

• To develop and validate a novel mathematical procedure for fitting constrained piecewise linear equations.
• To apply this method for generating continuous lung function reference equations for healthy individuals.

Main Methods:

• A "Separable Least Squares" algorithm was employed, nesting linear parameter estimation (General Linear Model) within nonlinear optimization.
• The procedure allows for multiple, estimable breakpoints and the inclusion of additional covariates.
• The method was demonstrated by estimating continuous lung function reference equations.

Main Results:

• The developed mathematical procedure successfully fitted piecewise linear equations with constrained multiple breakpoints.
• Continuous lung function reference equations were generated for healthy normal subjects.
• Comparison revealed advantages over previously published, discontinuous equations.

Conclusions:

• The proposed separable least squares method provides a robust framework for fitting complex piecewise linear models.
• This approach yields superior continuous reference equations for lung function, improving age-specific interpretations.
• The methodology is adaptable for various biological and medical modeling applications.