Psychosocial-Behavioral Phenotyping: A Novel Precision Health Approach to Modeling Behavioral, Psychological, and...

Area of Science:

• Computational social science
• Precision medicine
• Behavioral science

Background:

• Context significantly impacts behavioral intervention efficacy, necessitating systematic approaches to identify contextual factors.
• Machine learning (ML)-based phenotyping offers a precision health approach by enabling personalized behavior interventions.
• Psychosocial-behavioral phenotypes can identify population subgroups and intervention targets, integrating social determinants of health (SDOH) with individual factors.

Purpose of the Study:

• To demonstrate a novel ML application for psychosocial-behavioral phenotyping.
• To identify subgroups with similar psychosocial characteristics using ML.

Main Methods:

• Secondary analysis of community cohort data (n=5,883).
• Optimized a multichannel mixed membership model (MC3M) via Bayesian inference for phenotyping.
• Used logistic regression to associate phenotypes with elevated weight status (BMI ≥ 25kg/m²).

Main Results:

• Identified 20 distinct psychosocial-behavioral phenotypes.
• Phenotypes were consistent and discriminative, with most participants having one primary phenotype.
• Two phenotypes were positively associated, and four were negatively associated with elevated weight status, suggesting tailored intervention needs.

Conclusions:

• Psychosocial-behavioral phenotypes illuminate the complexity of SDOH and individual factors.
• These phenotypes provide actionable insights into community member similarities and differences.
• Phenotypes can identify context-specific intervention targets for precision health.