Latent class mediator for multiple indicators of mediation

Area of Science:

• Psychiatry
• Statistics
• Health Services Research

Background:

• Evaluating intervention effectiveness often involves complex mediation pathways.
• Multiple indicators of mediation can present challenges in traditional statistical modeling.
• Latent class analysis offers a novel approach to handle multiple mediation indicators.

Purpose of the Study:

• To demonstrate the utility of latent class mediators for analyzing interventions with multiple mediation indicators.
• To decompose total mediating effects into additive effects from distinct mediating pathways.
• To apply this method to evaluate schizophrenia treatment services.

Main Methods:

• Utilized latent class analysis to identify underlying mediating pathways from multiple observed indicators.
• Applied the method to a 2-year clustered randomized trial dataset for first-episode schizophrenia.
• Considered four indicators: resiliency training, family psychoeducation, supported education/employment, and medication assessment.

Main Results:

• The latent class mediator, derived from four service indicators, significantly mediated symptom improvement in schizophrenia patients.
• Simulation studies indicated minimal bias in maximum likelihood estimation when indicator entropy was high.
• The model successfully decomposed total mediating effects into additive pathway contributions.

Conclusions:

• Latent class mediators provide a robust framework for evaluating interventions with multiple mediation indicators.
• This approach simplifies the analysis of complex mediation structures, offering interpretable additive effects.
• The findings support the use of latent class analysis in psychiatric research for treatment outcome evaluation.