Skip to main content

Scientific question and hypotheses

Why the study is needed​

Public-health datasets and health-system EHR data observe different populations through different measurement systems. Population surveys and modeled small-area estimates are designed for population inference. Chicago Health Map summarizes diagnoses recorded during care among adults observed in participating health systems.

Many public products report at ZIP-code or community-area scales. A single label for a larger area can hide clinically recorded variation among the tracts inside that area. The study therefore asks an information question: does the smaller geography add a materially different description?

Primary question​

Among eligible Chicago census tracts during 2022–2024, do direct tract-level Chicago Health Map measures contain geographic information beyond the direct Chicago Health Map community-area label?

Primary hypothesis​

If tract resolution adds information, we expected to observe all or some of the following:

  • nonzero variation among tracts within the same community area;
  • meaningful absolute differences between tract and community-area percentile ranks;
  • disagreement between tract and community-area quartile classifications; and
  • movement into or out of the highest tract quartile after applying the larger-area label.

Evidence against added tract resolution would be near-zero within-area variation, small rank gaps, near-complete quartile agreement, and little movement across the highest-quartile boundary.

Secondary questions​

  • Does the conclusion change when Census ZCTA is the coarser geography?
  • How closely do Chicago Health Map and CDC PLACES rank the same tracts?
  • Do Chicago Health Map and the Healthy Chicago Survey similarly order community areas for hypertension?
  • Is remaining spatial structure detectable after the supplementary ecological model?

What was not hypothesized​

The study did not hypothesize that Chicago Health Map estimates population prevalence, identifies individual risk, establishes causality, measures unmet need, locates services, or proves that one data source is correct.