Examining moderation analyses in propensity score methods: application to depression and substance use.

Area of Science:

• Epidemiology
• Biostatistics
• Social Science Research

Background:

• Nonexperimental studies often require methods to assess causal effects within specific subgroups.
• Propensity score methods are valuable for adjusting confounding in observational research.
• Understanding effect modification is crucial for targeted interventions.

Purpose of the Study:

• To provide guidance on integrating propensity score methods with moderation analyses for subgroup causal effect estimation.
• To examine how depression influences substance use differently across genders using a longitudinal cohort.
• To compare various propensity score techniques for estimating subgroup effects.

Main Methods:

• Utilized data from a longitudinal community cohort study (N=952) of urban African Americans.
• Estimated the influence of young adult depression on midlife substance use, assessing gender as a moderator.
• Compared five propensity score techniques, including separate and joint models with different matching strategies for gender subgroups.

Main Results:

• Estimating separate propensity score models for men and women demonstrated superior balance, making it a preferred technique for subgroup analyses.
• The study found that depression has consequences for substance use.
• No significant gender differences were observed in the substance use consequences of depression.

Conclusions:

• Prespecifying subgroup effects and checking balance within these subgroups are critical before propensity score estimation.
• The findings suggest that depression affects multiple substance use outcomes in midlife for both men and women with similar impact.
• Separate propensity score models are recommended for subgroup analyses when balance is a priority.