Schedule of Events | Symposia

Beyond Traditional Factor Analysis: Exploring Latent Variable Modeling Strategies to Capture BrainHealth Index Trajectory.

Poster Session F - Tuesday, April 1, 2025, 8:00 – 10:00 am EDT, Back Bay Ballroom/Republic Ballroom

Quentin Coppola¹ (coppola.q@northeastern.edu), Marc Yanguez¹, Jeffrey Spence², Lori Cook², Mark D'Esposito³, Sandra Chapman², Susanne Jaeggi¹, Aaron Seitz¹; ¹Northeastern University, ²University of Texas, Dallas, ³University of California, Berkeley

Brain health is a multifaceted construct encompassing cognition, mood, physical and mental well-being, and social engagement. Comprehensive batteries capturing these outcomes are crucial for evaluating the efficacy of interventions aimed to improve cognition and overall brain health. Integrating these subdomains in a unified model is challenging. Traditional exploratory and confirmatory factor analysis (EFA; CFA) methods often omit critical cross-loadings that better capture complex structures. Alternative methods leveraging flexible specifications (e.g., cross-loadings) may better capture this complex interplay, providing a more nuanced understanding and measurement of brain health. This study systematically evaluates latent variable modeling techniques to create a unified model of a brain health battery that includes measures of cognition, well-being, social interaction, and daily functioning. Data were collected from 4000 adults aged 18 to 92 (Mean [SD] = 61.4 [13.0]) through The BrainHealth Project led by the Center for BrainHealth (University of Texas, Dallas, USA). We compare traditional techniques, such as EFA and CFA, to alternative approaches, including exploratory structural equation modeling (ESEM), Bayesian confirmatory factor analysis (B-CFA), and latent network analyses (LNA). These methods provide greater flexibility in specifying interrelationships among variables, allowing for more nuanced representations of brain health. Results demonstrate the utility of these alternative models in characterizing underlying latent structures across a battery of over 20 measures while preserving key interactions between domains. Future work will leverage latent factors with complementary outcomes (e.g., brain imaging) to evaluate intervention efficacy and understand brain health across diverse populations, ultimately revealing the multidimensional contributors to brain health improvements.

Topic Area: METHODS: Other