01. Problem Context & Epidemiological Disparities
Chicago exhibits significant life-expectancy and health outcome disparities across neighborhoods. Historically, municipal health interventions have frequently focused on individual behavioral factors, often without statistical evaluation of underlying structural contributors.
Quantifying Structural vs. Behavioral Factors
The objective of this modeling effort was to measure the relative predictive power of socioeconomic structural indicators versus behavioral metrics in forecasting adult diabetes prevalence across Chicago's 77 designated community areas.
77 Zones
Chicago community areas
R² ~ 0.46
Predictive model variance
Structural Factors
Quantified socioeconomic weight
02. Analytical Pipeline & Full-Stack Implementation
Developed a full-stack data pipeline from data cleaning and normalization to multivariate regression, clustering, and an interactive web interface.
Multivariate Regression
Modeled variance across community areas (R² ~ 0.46, error margin <5%), isolating the relative statistical weights of median income, uninsured rates, and clinical access.
K-Means Clustering
Clustered community areas into distinct vulnerability profiles, identifying localized regions where low median income (<$40k) intersected with diabetes rates exceeding 18%.
Interactive Dashboard
Built a Next.js web dashboard with interactive charts and an integrated LLM interface to allow users to query model coefficients and localized health data.
PII Constraints
Engineered strict system prompts and input filters to prevent PII ingestion and prevent medical diagnostic claims.
03. Statistical & Modeling Insights
Predictive Weight of Structural Inequities
Standardized regression coefficients indicated that median household income and health insurance coverage status mathematically outweighed individual behavioral variables (such as smoking rates) in explaining diabetes variance across Chicago community areas.
Detection Bias in Primary Care Access
Initial correlation analysis revealed a positive relationship between primary care provider density and recorded diabetes rates. Closer evaluation indicated this was a detection artifact: higher clinical access yields higher screening rates, while lack of insurance in under-resourced areas suppresses formal diagnoses, artificially deflating reported figures.
04. Analytical Conclusions & Spatial Targeting
Structural Determinant Weight
Quantified that economic stability and coverage access are the primary statistical determinants of community diabetes rates, establishing quantitative support for prioritizing coverage enrollment initiatives.
Geographic Resource Precision
Provided clear geospatial cluster coordinates to deploy mobile screening and enrollment teams directly to high-vulnerability South and West Side community zones.