310 – Statistical Methodology for Business and Economics
Semiparametric Mixed Model for Detection of Rapid Disease Progression
Leo Li Duan
Cincinnati Children's Hospital Medical Center
John P. Clancy
Cincinnati Children's Hospital Medical Center
Rhonda D. Szczesniak
Cincinnati Children's Hospital Medical Center
Semiparametric regression provides a more flexible mean structure to analyze longitudinal data through penalized regression splines and mixed modeling. This representation provides smooth estimates of nonlinear mean functions and derivatives while preventing overfitting. Meanwhile, covariance structures must be included to account for repeated measurements collected on multiple subjects. An additional complication arises in the case of unequally spaced measurements, which are common in observational settings. We propose and implement a semiparametric mixed effects model that fits a complex mean function while incorporating subject-specific random effects and an exponential correlation function to account for irregularly observed measurements from multiple subjects. We illustrate the model using serial lung function measurements from the United States Cystic Fibrosis Foundation Patient Registry. This database has clinical encounter information that has been collected over the lifetimes of more than 30,000 individuals. We present our longitudinal model fit and its derivatives to show how this method can provide further insights into the degree and timing of disease progression.