Xingyu Chen, a masters student in the BIDS Health Sciences Informatics-Research program, recently had a paper (co-authored with Drs. Christopher Kitchen and Jadi Kharrazi) published in the American Medical Informatics Association journal, JAMIA Open. Xingyu’s paper is entitled, “Assessing the Advantages and Disadvantages of Dimensionality Reduction Methods in Summarizing Housing Determinants of Health in the US.”
Housing conditions significantly affect health outcomes, but measuring housing quality remains challenging with no consistent approach across research studies. This study compared three mathematical techniques for creating housing quality scores using census data across counties, ZIP codes, and Census tracts throughout the United States. They analyzed 15 housing characteristics from the American Community Survey data (2010-2019) to determine which method best captures housing conditions. They found that the Principal Component Analysis (PCA) methodology produces the most consistent and understandable results compared to other machine learning approaches. PCA showed superior stability over time and across different geographic levels, while maintaining acceptable interpretability of how different housing factors contribute to the overall score.
This research provides public health researchers and policymakers with a reliable, data-driven method that can quantitatively assess housing conditions and compare housing disparities. The findings of this study contribute to the methodological foundation required to develop a robust summarized housing score that can inform public health policies and address health disparities.