Statistical primer: an introduction to the application of linear mixed-effects models in cardiothoracic surgery outcomes research—a case study using homograft pulmonary valve replacement data

Abstract

Summary OBJECTIVES The emergence of big cardio-thoracic surgery datasets that include not only short-term and long-term discrete outcomes but also repeated measurements over time offers the opportunity to apply more advanced modelling of outcomes. This article presents a detailed introduction to developing and interpreting linear mixed-effects models for repeated measurements in the setting of cardiothoracic surgery outcomes research. METHODS A retrospective dataset containing serial echocardiographic measurements in patients undergoing surgical pulmonary valve replacement from 1986 to 2017 in Erasmus MC was used to illustrate the steps of developing a linear mixed-effects model for clinician researchers. RESULTS Essential aspects of constructing the model are illustrated with the dataset including theories of linear mixed-effects models, missing values, collinearity, interaction, nonlinearity, model specification, results interpretation and assumptions evaluation. A comparison between linear regression models and linear mixed-effects models is done to elaborate on the strengths of linear mixed-effects models. An R script is provided for the implementation of the linear mixed-effects model. CONCLUSIONS Linear mixed-effects models can provide evolutional details of repeated measurements and give more valid estimates compared to linear regression models in the setting of cardio-thoracic surgery outcomes research.