Analyzing Multiple Social Determinants of Health Using Different Clustering Methods

Area of Science:

• Health Services Research
• Biostatistics
• Public Health

Background:

• Social determinants of health (SDoH) are crucial in healthcare but challenging to analyze due to collinearity.
• The Protocol for Responding to and Assessing Patient Assets, Risks, and Experience (PRAPARE) tool collects standardized SDoH data.
• Existing research faces difficulties in analyzing complex, interconnected SDoH factors.

Purpose of the Study:

• To evaluate and compare three statistical methods for analyzing SDoH data collected via the PRAPARE tool.
• To demonstrate how different statistical approaches can address SDoH collinearity in outcomes research.
• To guide researchers in selecting appropriate methods for analyzing complex SDoH data.

Main Methods:

• Utilized exploratory factor analysis (FA), hierarchical clustering, and latent class analysis (LCA).
• Analyzed data from 2380 patients with complete PRAPARE and neighborhood-level information.
• Compared the identified SDoH clusters and latent classes from each method.

Main Results:

• Identified three composite SDoH clusters using FA.
• Discovered four distinct clusters via hierarchical clustering.
• Revealed four latent classes of patients using LCA.
• Demonstrated varying results based on the chosen statistical method.

Conclusions:

• Multiple statistical methods can effectively analyze complex SDoH data.
• The choice of method depends on the researcher's specific outcomes and goals.
• Employing diverse analytical approaches enhances understanding of SDoH in healthcare settings.