AAPI-Recruitment-Optimization-DCE

Optimizing AAPI Recruitment for Alzheimer's Research: A Discrete Choice Experiment (DCE)

Author: Isaac Nguyen Status: Completed

---

Executive Summary

This is an independent Collab Project on DCE Simulation + Demographic Weighting for AANHPI Recruitment

GOAL: By the end of this, I am making a recommendation for optimal recruitment strategy for Vietnamese populations.

To accomplish this, I am creating a dataset that mimics the structure of the CARE registry survey (N=1000 AANHPI seniors) and replicates the specific utility weights/preferences found in the study—including unique deviations for Vietnamese participants.

Technical Approach:

This project demonstrates the following data science competencies:

• Uses a Monte Carlo Simulation: Generated a synthetic dataset mirroring the specific probability distributions and utility weights derived from the original literature.
• Behavioral Modeling: Implemented Conditional Logistic Regression using the "Difference Method" to isolate preference utilities for recruitment attributes (Who, Why, How).
• Heterogeneity Analysis: Statistically isolated subgroup deviations, specifically quantifying the preference for "Ethnic Radio/TV" among Vietnamese participants (OR > 1.0) vs. other groups.
• Business ROI Modeling: Translated statistical odds ratios into a predictive "Recruitment Yield" calculator to guide budget allocation.

Key Findings

1. Digital Divide: While Social Media is the dominant channel for younger demographics (OR: 1.50), it underperforms for specific older subgroups.
2. The "Hidden" Channel: Vietnamese participants showed a statistically significant preference for Ethnic Radio/TV.
3. Strategic Recommendation: A segmented campaign allocating specific resources to radio for Vietnamese targets is projected to increase recruitment yield by ~XX% (see model outputs) compared to a generic flyer campaign.

How to Run This Code

This project is written in Python and designed to run in a Jupyter Notebook environment (Google Colab or Kaggle).

Dependencies

• numpy & pandas: Data manipulation and synthetic generation.
• statsmodels: Advanced statistical modeling (Conditional Logit).
• matplotlib & seaborn: Visualization of forest plots and ROI heatmaps.

Quick Start

1. Clone this repository.
2. Install requirements: pip install pandas numpy statsmodels seaborn
3. Run the notebook AAPI_Recruitment_Optimization.ipynb sequentially.

Reference Literature

Original Study: Ta Park, V. M., et al. (2023). [cite_start]Asian Americans' and Pacific Islanders' preferences in recruitment strategies and messaging for participation in the CARE registry: A discrete choice experiment. Alzheimer's & Dementia, 19, 5198–5208. [cite: 7, 30]

---

This analysis was conducted as an independent replication to demonstrate statistical modeling and recruitment optimization techniques.