An overview of longitudinal data analyses in sleep research

Repeated measures or longitudinal data, are commonly used in sleep apnea research to assess changes in weight, symptoms or continuous positive airway pressure (CPAP) treatment adherence over time. Such data allow the examination of both between-person differences and within-person changes in outcomes. However, analyzing longitudinal data can be challenging, particularly in selecting the appropriate statistical method, as an inadequate choice may introduce bias into parameter estimates and outcome predictions.

This paper aims to provide a structured and accessible overview of longitudinal data analysis in sleep research by describing the most commonly used statistical methods and illustrating their applications through a simulated example.

A simulated dataset of 300 patients with 90 repeated measurements over time was created illustrating changes in CPAP adherence and Epworth Sleepiness Scale score over time considered as binary, categorical and continuous outcomes.

A stepwise analytical framework encompassing the main stages of longitudinal data analysis-data description, visualization, formulation of the research question, and selection of the statistical method. Within this framework, ten commonly used analytical approaches are presented, covering descriptive (ANOVA, χ²), clustering (K-means, LTA, GMM, GBTM), and modeling methods (mixed models, ARIMA, survival analysis, HMM), with emphasis on their assumptions, scope, and interpretation.

This structured overview proposes a coherent analytical pathway for longitudinal data, supported by implementation examples, to help researchers and clinicians navigate complex longitudinal-measure analyses in sleep research and related fields.