SEIR Modeling of UTI Spread in Water Networks

ISEF Category: Computational Biology and Bioinformatics

Subcategory: Computational Epidemiology  ·  Difficulty: Advanced  ·  Setup: Home Setup  ·  Time: Full Year

The Hook

A city’s water system can connect thousands of people in ways you never see. That makes it a perfect place for a computer model. You can test how antibiotic resistance and water reuse might change infection risk across neighborhoods. This project turns public health data into a map of spread.

What Is It?

This project asks you to build a spatial SEIR model. SEIR stands for Susceptible, Exposed, Infectious, and Recovered. Think of it like a relay race. People move through stages as an infection spreads, and your model tracks how that movement changes across a map.

The twist here is the water-reuse network. Water reuse means treated water gets used again for things like irrigation or industrial use. You would combine public AMR surveillance data, such as NARMS, with EPA water data to estimate where resistant bacteria might appear or spread more easily. You are not proving that water causes UTIs. You are testing whether certain network patterns, plus resistance data, line up with higher modeled risk.

Because this is a computer project, your main work happens in data cleaning, parameter choices, and simulation design. That gives you room to ask sharp questions without needing a wet lab.

Why This Is a Good Topic

This is a strong science fair topic because you can test clear variables, like network density, reuse patterns, and resistance rates, then measure their effect on modeled spread. It connects to a real public health problem, antibiotic resistance, and a real infrastructure issue, water reuse. You can learn modeling, data cleaning, GIS-style thinking, and basic epidemiology from public data alone.

Research Questions

• How does changing the density of a municipal water-reuse network affect the simulated spread of antibiotic-resistant UTI cases?
• What is the effect of different resistance prevalence estimates from NARMS on the final number of modeled infections?
• Does adding geographic clustering of reuse sites change outbreak size compared with a random network?
• To what extent do seasonal water-quality patterns from EPA data alter transmission estimates in the SEIR model?
• Which network nodes contribute most to spread when you remove or strengthen them in the simulation?
• How does the model response change when you vary the contact rate between neighborhoods linked by reused water?

Basic Materials

• Laptop or desktop computer with enough memory to run simulations.
• Python installed with NumPy, pandas, SciPy, and matplotlib.
• Public NARMS data tables or downloads.
• EPA water-quality or water-system datasets.
• Spreadsheet software for organizing source data.
• External hard drive or cloud storage for backups.
• Notebook for tracking assumptions, parameters, and model versions.

Advanced Materials

• Access to geospatial software or Python mapping libraries such as GeoPandas.
• Access to census or municipal boundary data for spatial units.
• Command-line Python environment or Jupyter Notebook.
• Version control with Git.
• Statistical package for sensitivity analysis.
• Public health datasets with finer geographic resolution, if available.
• Network analysis library such as NetworkX.

Software & Tools

• Python: Builds the SEIR simulation, cleans public datasets, and runs sensitivity tests.
• Jupyter Notebook: Keeps code, notes, and plots together in one place.
• pandas: Organizes surveillance and water data into analysis-ready tables.
• GeoPandas: Handles spatial boundaries and mapping for the reuse network.
• NetworkX: Models the reuse system as a connected network and measures node influence.

Experiment Steps

1. Define the exact geographic unit you will model, such as a city, county, or utility service area.
2. Choose the infection outcome and decide how public resistance and water data will feed model parameters.
3. Build the network structure first, then decide how movement or contact will happen across connected nodes.
4. Set up a baseline SEIR model, then add resistance and spatial variation one layer at a time.
5. Plan sensitivity tests so you can see which assumptions change the results the most.
6. Decide how you will compare scenarios, such as higher reuse, lower resistance, or different cluster patterns.

Common Pitfalls

• Treating water reuse as a direct cause of UTIs, which weakens the public health logic of the project.
• Mixing data from different years without checking whether the surveillance and water datasets overlap in time.
• Using county or city labels that do not match across datasets, which creates false pairings in the model.
• Building a complex spatial model before testing a simpler baseline, which makes it hard to find errors.
• Reporting one simulation run instead of repeating the model with varied parameters and uncertainty ranges.

What Makes This Competitive

A strong version of this project goes beyond a basic simulation. You can compare multiple network structures, test uncertainty in public datasets, and show which assumptions really drive the output. Better entries often include sensitivity analysis, clean visual maps, and a clear explanation of why the model matters for public health planning. If you can compare your results against real surveillance patterns, your project gets much stronger.

Project Variations

• Use hospital discharge data or local health department reports as the infection signal instead of only NARMS resistance rates.
• Swap the municipal reuse network for a river basin or watershed network and test how spatial flow changes the model.
• Compare two modeling approaches, such as SEIR versus a simpler compartment model, to see which explains the public data better.

Learn More

• NARMS by the CDC and FDA: Search for annual reports and data tables on antibiotic resistance surveillance in foodborne and related bacteria.
• EPA Water Data: Search EPA datasets for drinking water, wastewater, and reuse-related public records.
• NOAA Climate Data Online: Use weather and climate records to test whether seasonal conditions track with your model parameters.
• NIH PubMed: Search for review articles on antibiotic resistance, urinary tract infections, and spatial epidemiology.
• MIT OpenCourseWare: Look for free courses in epidemiology, network science, and modeling to build your methods background.
• NetworkX Documentation: Read the free docs to learn how to build and analyze contact networks in Python.