Inspiration
Pharmacy deserts are a major problem in the United States. According to a 2025 JAMA article an estimated 17.7% of the people in the US (57.1 million people) live in areas with inadequate access to prescription drugs. Unsurprisingly, this burden disproportionately falls on the most vulnerable populations. Pharmaceutical companies like McKesson are working to address this through their Oasis program – simultaneous increasing access to drugs as well as their revenue. Given the scope of need, one of the challenges for such initiatives is ascertaining which areas stand to benefit the most. Our team sought to create a tool to assist in this process.
What it does
The tool pulls in data related to pharmacy deserts and outputs locations based on the relative level of need.
How we built it
The team assembled data organized by zipcode from the US Census, CDC, National Plan and Provider Enumeration system (NPPES), and EPA which highlighted 1) areas with highest disease burden 2) areas with least financial means 3) areas with least access to pharmacies 4) areas with greatest vulnerability to extreme heat and air pollution 5) areas where the greatest number of people stand to benefit. We convert each ZIP’s access, health, income, and density into comparable percentiles, use Isolation Forest and an Autoencoder/PCA to score how unusually severe each ZIP is, and blend that anomaly score with a simple mathematical baseline to rank the best locations for a new pharmacy.
Challenges we ran into
There was a significant challenge finding as current datasets as possible as well as figuring out how to integrate all of the parameters of interest
Accomplishments that we're proud of
Our team managed to create a working prototype in 24 hours
What we learned
How challenging a project of this scale can be
What's next for AI pharmacy desert finder
Making the project available with industry partners who can use this information to expand pharmacy access
Built With
- geopandas
- python
- scikit-learn
- streamlit
- tensorflow
Log in or sign up for Devpost to join the conversation.