| Summary: | Introduction/Aim: Trajectory modelling approaches have been developed to determine subgroups within a given population and are increasingly used to better understand pain outcomes. With the purpose of enabling pain researchers to choose the technique that best suits their research questions, the objective of this narrative review was to explore various trajectory modelling approaches used in health research and discuss about their applications. Methods: To establish and identify relevant peer-reviewed literature, PubMed, Psych-Info and Google Scholar were used with no date of restriction. Approaches were compared in terms of definitions, rationale of use, assumptions, concrete clinical applications, and availability of statistical software programs. Results: Three common approaches of trajectory modelling were identified: Latent class modelling (LCM) approaches (e.g. Growth mixture modelling-GMM, Group-based trajectory modelling-GBTM, Latent class analysis-LCA), cluster analysis (CA) and sequence analysis (SA). LCM are based on a probabilistic modelling approach with a finite mixture distribution that describes an observed life course sequence of categorical values as resulting from the conditional probabilities that define membership of a latent class. LCM provides the contribution of every observed variable on the definition of classes. CA is a tool to explore groups within a data set. When variables under study are continuous, CA is sometimes called latent profile analysis. When the variables are categorical, CA is sometimes called LCA. SA is a fully nonparametric approach based on algorithmic, approaches aimed at making use of measures of distance between individual trajectories. Discussion/Conclusions: Depending on the research question and the available data particularities, one or another of these approaches can be used for trajectory modelling.
|
|---|