Joint mixed-effects modelling of longitudinal health data for m

Abstract

Cystic fibrosis (CF) is a genetic disease that affects several organs of the body such as the liver, pancreas, or the lungs. Cystic fibrosis is monitored using hospital and home spirometry, where the difference between the two is what type of technique that is used. This study evaluates joint mixed-effects models as a statistical framework for analysing home and hospital spirometry measurements. Statistical methods, such as simple linear regression and univariate mixed-effects models can be used to analyse spirometry. However, both simple linear regression and univariate linear mixed effects models considers only hospital spirometry and do not accommodate the rich patient-generated data obtained by home spirometry. By jointly modeling home and hospital spirometry, we aim to investigate prediction accuracy of hospital spirometry. We compare the performance of joint and univariate linear mixed-effects models with that of simple linear regression, evaluating their ability to predict: (i) the current (i.e., most recent) hospital spirometry value for a patient, (ii) the current mean value, (iii) the subject-specific trend, and (iv) the population-average trend. The results showed that joint linear mixed-effects models and univariate linear mixed effects models should be used over simple linear regression to estimate trends and specific values in a CF-setting. Moreover, the results supported the use of joint linear mixed-effects models when jointly modeling home and hospital spirometry data.