Healthcare access isn't evenly distributed, and the communities that fall through the gaps are often the ones least equipped to advocate for themselves. The problem is concrete: where are the access deserts in a city, who lives in them, and how unequal is the overall distribution? Answering that well matters to anyone deciding where to place a clinic, fund a shuttle route, or target outreach, and this talk shows how open geospatial tools can answer it rigorously, on free software and public data, reproducibly for any city. The talk makes four contributions: an enhanced accessibility method, findings from three real cities, a fully open-source geospatial pipeline, and reproducibility that lets any agency run the analysis on its own jurisdiction.

The method builds on the Two-Step Floating Catchment Area (2SFCA) model, a general spatial accessibility measure that balances facility supply against population demand within travel-distance catchments. Standard 2SFCA, though, treats every resident as identical in demand, so a low-income, uninsured community and an affluent, well-insured one of the same size register the same need, which means conventional ratios quietly under-measure deprivation in exactly the populations equity analysis cares most about. A key methodological contribution of this work is the weighting of sociodemographic and economic demand: effective demand is based on income, insurance coverage, and working-age fractions, so accessibility scores and inequality metrics reflect need, not just headcount. A second refinement, a truncated Gaussian distance decay that reaches exactly zero at the catchment boundary, gives cleaner, more interpretable results than the standard asymptotic kernel.

I apply this methodology to three facility types across three cities, each chosen deliberately. Dialysis centers (in New York City) deliver non-optional chronic care whose accessibility is acutely distance-sensitive; federally qualified health centers (in Los Angeles) are primary-care safety net sites serving underserved populations, a natural fit for equity analysis; and intermediate care facilities (in Washington, DC) provide long-term specialized care that has received comparatively little attention in accessibility research. Together they span different care models, geographies, and catchment patterns, which tests whether the method generalizes beyond any single facility type, rather than fitting one case.

The findings tell three different equity stories. DC shows the starkest inequality, with a population-weighted Gini of 0.80 and outer blocks systematically underserved. LA's primary care network is more uniform (Gini 0.68) but still strands pockets of population behind topographical barriers. NYC sits between them (Gini 0.73). Each result maps to a decision: the Gini coefficient says how unequal a city's access is, the Lorenz curve shows whether the shortfall is concentrated or widespread, and the zero-access count names how many residents are stranded.

Making this credible and reusable is as much an engineering story as an analytical one. Every component is free and open: GeoPandas, Shapely, and SciPy for the spatial work, DuckDB for analytics, and Streamlit for the dashboard, all running on public Census TIGER and CMS data. The pipeline follows a medallion architecture (raw → validated → analysis-ready), with the engineering that makes it robust at scale: spatial indexing for performance, careful coordinate-system handling, and graceful fallback when an API is down. To make results legible to non-technical stakeholders, I use choropleth maps to depict accessibility, bivariate maps that show population and access together (surfacing priority areas where high need meets low access), and a live public dashboard for interactive exploration.

The pipeline is fully reproducible as it is configuration-driven: every study area parameter (city, facility type, catchment radius, decay function, and coordinate system) lives in a YAML file, so a new jurisdiction means editing config, not rewriting code. Any agency or community group can point it at their city and compute accessibility scores from public data, with no proprietary software and no licensing cost.

I will also be candid about the limits: provider data is messy, Euclidean distance simplifies real street-network travel, and catchment-radius choices materially change the picture. This is a caveat that matters for anyone tempted to treat an accessibility map as ground truth.

Attendees will leave understanding what enhanced 2SFCA measures and why equity weighting matters, how to read and trust accessibility maps, and how to reproduce the whole workflow on open tools for their city. The method transfers directly beyond healthcare. It can be used for access analysis of evacuation resources, cooling centers, and other hazard-resilience services, showing how open tools turn a single research result into infrastructure any community can reuse. The MIT-licensed code and live dashboard are both publicly available.

Intended audience: public health and health equity analysts, urban planners, GIS practitioners, data scientists, and data engineers.