Trajectory analyses in insurance medicine studies: Examples and key methodological aspects and pitfalls

Area of Science:

• Epidemiology
• Biostatistics
• Longitudinal Data Analysis

Background:

• Trajectory analyses are vital for understanding variations in longitudinal outcomes like sickness absence and medication use.
• Methodological and interpretational challenges exist in applying trajectory analyses to time-varying data.
• This study compares two software packages for identifying health trajectories.

Purpose of the Study:

• To compare results from Group-Based Trajectory Models (GBTM) and Latent Class Growth Models (LCGM) using SAS and Mplus.
• To discuss methodological aspects and interpretation challenges of trajectory analyses.
• To highlight the strengths and pitfalls of person-oriented methods in longitudinal research.

Main Methods:

• Fitted GBTM using SAS and LCGM using Mplus on sickness absence data from 166,192 Spanish workers.
• Stratified analyses by sex and birth cohorts (1949-1969, 1970-1990).
• Utilized repeated measures of sickness absence spells per trimester.

Main Results:

• Both software identified four groups as optimal, but differed in trajectory starting values and shapes.
• These differences can lead to divergent conclusions depending on the software used.
• Key aspects discussed include model fit, group selection, covariate investigation, and result interpretation.

Conclusions:

• No single software was found superior, emphasizing the need for careful methodological consideration.
• Differences in trajectory shapes necessitate a thorough understanding of software-specific outputs.
• Further methodological research is recommended to support longitudinal epidemiological studies.