MaternalCompass

MaternalCompass: mapping Georgia’s maternal care deserts and using risk-weighted algorithms to help NGOs and governments access vital maternity care data and pinpoint where new clinics are needed.

Comment

Inspiration

Georgia is currently facing a silent emergency: it is one of the most dangerous states in the U.S. to give birth. The crisis isn't just medical; it's geographic.

The Maternal Desert: Over 50% of Georgia’s 159 counties lack a single OB-GYN or labor and delivery provider.

The Distance Gap: For a mother in a rural "care desert," the nearest Level 3 facility - capable of handling high-risk complications can be over 80 miles away. In an obstetric emergency, those 80 miles represent the difference between life and death.

A Preventable Crisis: The CDC reports that over 80% of pregnancy-related deaths are preventable. The barrier isn't a lack of medical knowledge; it's a lack of access.

We built MaternalCompass because local governments and NGOs shouldn't have to "guess" where to deploy mobile clinics or build new centers. We wanted to move beyond static spreadsheets and create a high-fidelity, algorithmic map that identifies exactly where the "Golden Hour" of care is being lost. We aren't just mapping data points; we are mapping the path to survival for mothers across Georgia.

What it does

MaternalCompass is a spatial analysis tool that identifies and visualizes the geographic gaps in Georgia’s maternal healthcare. By synthesizing complex public health datasets into an interactive dashboard, we empower NGOs and governments to make data-backed decisions about where to expand infrastructure.

Key Features

  • Interactive Mapping: A county-by-county visualization of Georgia using React-Leaflet, allowing users to see the maternal health landscape at a glance.
  • Granular County Insights: Clicking on any county triggers a deep-dive data panel that reveals: Risk Assessment, Average Distance from Hospital, Number of OB Beds, Percent of Births With No or Late Prenatal Care
  • Bed Allocation Simulator: Simulate adding OB beds at a new clinic location in a county to instantly show impact on risk level assessment
  • AI Chatbot: Powered by the Gemini API and trained on our county risk scores, hospital and OB bed data, and prenatal care statistics, the AI Assistant helps users understand the maternal health crisis in Georgia.

How we built it

We built MaternalCompass using a modern, lightweight tech stack optimized for speed and data visualization. Our process focused on three main pillars: data synthesis, algorithmic scoring, and geospatial rendering.

The Tech Stack

  • Frontend: Leaflet.js 1.9.4, Vanilla JavaScript, HTML5, CSS3 for an interactive and responsive user interface.

  • Backend: Python 3.14 with Flask and Flask-CORS to handle API requests and server logic.

  • APIs & Libraries: Google Gemini API (google-genai) and pandas for AI integration and data processing.

  • Data: CSV files for structured input and storage.

The Data Pipeline

We integrated three distinct data types to create a unified view:

  • GeoJSON Boundaries: To define the 159 counties of Georgia and create the interactive "clickable" regions.

  • Health CSVs: We ingested datasets containing prenatal care percentages, OB bed counts, and hospital proximity metrics.

  • Facility Mapping: We geocoded every maternal care center in Georgia, tagging them by their ACOG Level of Care (1, 2, or 3).

Challenges we ran into

Data Fragmentation: Our biggest hurdle was synthesizing data from multiple sources. We had to merge geographic boundaries (GeoJSON) with medical facility locations and medical statistics from various CSVs. Ensuring that a specific coordinate correctly mapped to the right county and its corresponding risk data required meticulous data cleaning.

The Risk Algorithm: Defining a "Risk Score" isn't just about distance. We struggled with how to weight different variables, like whether a lack of OB beds is more critical than a high percentage of late prenatal care. We had to iterate on our logic to ensure the algorithm reflected the lived reality of Georgia's healthcare landscape.

Mapping Performance: Rendering a high-detail map of all 159 Georgia counties with interactive layers and custom icons in React-Leaflet initially caused performance lag. We had to optimize how our frontend handled the GeoJSON data to keep the user experience smooth and responsive.

Scope Management: As a hackathon team, we had a dozen "must-have" features. Deciding to prioritize the core mapping and Risk Score over the search bar or chatbot was a tough but necessary call to ensure we delivered a polished, working MVP.

Accomplishments that we're proud of

Translating Data into Impact: We successfully took "cold" data from CSVs and JSON files and turned it into a visual narrative. Seeing a "Risk Score" actually reflect the known maternal care deserts in Georgia was a huge win for the team.

The Technical Integration: Seamlessly connecting React, Vite, and Leaflet to handle real-time data filtering. Getting the county boundaries to talk to our risk algorithm was a major technical milestone.

User-Centric Design: We’re proud of building a dashboard that doesn't require a PhD to understand. We focused on making the UI clean and intuitive so that an NGO worker or a local politician could use it immediately to identify where resources are lacking.

Mathematical Modeling: Developing a multi-variable risk formula that accounts for both physical distance and socioeconomic barriers like late prenatal care. It's not just a map; it's a diagnostic tool.

What we learned

The Reality of Health Inequity: Beyond the code, we learned the staggering reality of Georgia's maternal healthcare landscape. Seeing the data firsthand - where some counties have zero resources for thousands of residents, deepened our understanding of why technology is so vital in public health.

Geospatial Data is Messy: We learned that mapping is about more than just coordinates. Handling GeoJSON layers and ensuring they sync perfectly with custom data points taught us a lot about spatial indexing and data normalization.

The Power of the "Risk Score": We realized that raw data can be overwhelming. We learned how to synthesize multiple complex variables (like prenatal care percentages and number of OB beds) into a single, digestible metric that tells a story at a glance.

Full-Stack Agility: Building with Vite, React, and Supabase taught us how to move fast without breaking things. We learned how to prioritize a functional MVP, making the tough calls on which features would provide the most value to our target users (NGOs and local governments) under a tight deadline.

What's next for MaternalCompass

Hyper-Local Search: We plan to implement a high-precision search bar using Google Maps or Mapbox API, allowing users to enter a specific zip code or address to find the nearest facility and the exact travel time in minutes.

NGO & Government Portal: Developing a secure dashboard where organizations can "drop pins" on the map to simulate how a new mobile clinic would lower the surrounding area's Risk Score.

Expanded Geographic Scope: While we started with Georgia, the maternal health crisis is national. We intend to scale our data ingestion to cover the entire "Black Belt" region and other rural areas across the Southern United States.

Real-Time Resource Tracking: Moving beyond static CSV data, we aim to integrate live API feeds from hospitals to show real-time OB bed availability and current staffing levels.

Share this project:

Updates