Development of longitudinal data analysis in mental health research for military service members

Abstract

Introduction: Since the 1991 Gulf War, mental health conditions of military Service members have received increasing public attention and are a major focus for the U.S. government. A substantial proportion of military health research has been devoted to examining the pattern of change over time in mental health symptoms and diagnostic status among Service members. Unfortunately, many researchers continue to use somewhat obsolete methods to analyze trends and transitions in mental health, despite advances in statistical methodology that permit attention to the unique features inherent in longitudinal data. Materials and Methods: This article defines and describes data features and structures, and basic specifications of longitudinal data analysis to military health researchers. In particular, we highlight the respective impacts of missing data and intra-individual correlation on longitudinal data analysis. Based on the descriptions of the basic features in longitudinal data, we introduce several popular techniques to analyze a variety of longitudinal data types. Results: We demonstrate that traditional analytic techniques do not properly account for missing data and intra-individual correlation inherent in longitudinal data. Failure to use correct, appropriate models and methods can result in major bias in analytic results and mental health predictions. Conclusions: Failure to use correct, appropriate models and methods in longitudinal data analysis can have unfortunate repercussions on a military health system that needs accurate findings to support valid policy decisions. By applying adequate models and methods, military health researchers will be able to better understand the complex interactions of biological, psychological, and social factors on mental health trends and transitions among military Service members.