CITY SCHOLAR @ UNIVERSITY OF ILLINOIS URBANA-CHAMPAIGNPUBLIC HEALTH ANALYTICS

Chicago Diabetes Modeling:
Predictive Policy Dashboard

Quantified socio-economic determinants of adult diabetes across 77 Chicago zones using predictive modeling, deploying an AI dashboard to guide municipal healthcare funding.

01. The Public Health Challenge

Chicago experiences massive life expectancy disparities, with gaps spanning up to 30 years depending solely on a resident's zip code. Historically, public health funding often targets individual behavioral choices (e.g., lifestyle and anti-smoking campaigns).

The Goal: Shifting the Narrative

Chronic conditions like diabetes are deeply tied to socio-economic structural factors. The objective of this initiative was to quantify exactly which structural factors drive these disparities and build an accessible tool to help policymakers shift from qualitative assumptions to quantitative facts.

77 Zones

Chicago community areas

Predictive Model

Isolating predictive weight

Policy Pivot

Optimizing municipal ROI

02. Solution & Architecture

I developed a full-stack analytical pipeline, from ingesting and normalizing raw health data to deploying an interactive LLM dashboard for non-technical policymakers.

Predictive Modeling

Linear Regression

Accounted for ~46% of variance (error margin <5%) to isolate weights of Income, Uninsured Rates, and Care Access.

Geospatial Analysis

K-Means Clustering

Identified hyper-local "hot zones" (e.g., median income <$40k intersecting with diabetes rates >18%).

Dashboard Deployment

Interactive AI Dashboard

Integrated an LLM chatbot to democratize the data. The AI acts as a data translator, capable of explaining complex model coefficients to non-technical municipal users.

Security

Ethical Guardrails

Programmed strict constraints preventing the AI from generating or requesting PII.

03. Key Strategic Insights

Structural Supremacy

Standardized regression coefficients proved that wealth and insurance status are the primary "Invisible Architects" of Chicago's health disparities. These structural inequities mathematically outperformed individual behavioral choices (like smoking rates) in predictive modeling.

The "Doctor Paradox"

The model revealed a counter-intuitive positive correlation between primary care access and higher diabetes rates. Rather than representing a failure of healthcare, I interpreted this as a diagnosis bias, more doctors lead to better detection, whereas lack of insurance in other zones prevents diagnosis entirely, artificially deflating their reported rates.

04. Strategic Recommendations

Resource Reallocation

Recommended a strategic pivot for municipal funding, shifting capital away from generic, city-wide behavioral lifestyle campaigns toward hyper-targeted insurance enrollment drives.

Geographic Precision

Provided a data-backed roadmap to deploy mobile enrollment units directly to the high-vulnerability socio-economic clusters on the South and West Sides, maximizing the ROI of public health expenditures.