One Health AMR Data Dashboard

ISEF Category: Animal Sciences

Subcategory: Other  ·  Difficulty: Advanced  ·  Setup: Home Setup  ·  Time: 1 to 2 Months

The Hook

Antibiotic resistance does not stay in one place. A county with more livestock, more antibiotic sales, and more human infections may tell a story you can map. Your job is to see whether those signals move together, or whether the pattern breaks once you control for geography and population.

What Is It?

One Health means people, animals, and the environment affect each other. Think of it like three panels in one comic strip. If one panel changes, the others may change too. In this project, you would build a dashboard that lines up county livestock density, antibiotic sales, and human AMR infection rates.

The hard part is not just plotting dots. You need to decide whether the same county should be compared by raw counts, per-capita rates, or grouped trends. You also need to remember that correlation is not cause and effect. Your goal is to find patterns that are strong enough to survive better controls.

Why This Is a Good Topic

This is a strong science fair topic because it starts with public data, yet still asks a real research question. You can test links across place, animal density, and human health, then check whether the pattern holds after you add controls. You will learn data cleaning, mapping, correlation, and how to explain limits without overclaiming.

Research Questions

• How does county-level livestock density relate to human AMR infection rates after you control for population density?
• What is the effect of using per-capita antibiotic sales instead of raw sales on the strength of the correlation with human AMR infection rates?
• Does the relationship between livestock density and AMR infection rates change when you compare cattle-heavy, swine-heavy, and poultry-heavy counties?
• To what extent do rural counties and urban counties show different AMR patterns at the same livestock density?
• Which county controls, such as income, population density, or hospital access, explain more variation in AMR infection rates than livestock density alone?
• How does the pattern change when you compare the same county across multiple years instead of using a single-year snapshot?

Basic Materials

• Laptop with internet access.
• Spreadsheet software such as Google Sheets or Excel.
• Public USDA livestock data downloaded at the county level.
• Public CDC AMR or infection-rate data.
• County boundary file from the U.S. Census Bureau.
• Notebook for research notes.

Advanced Materials

• Laptop or desktop with enough memory for large county datasets.
• QGIS or ArcGIS Pro for spatial joins and choropleth maps.
• Python or R with data-cleaning and statistics packages.
• Secure access to restricted CDC or hospital-level data, if your mentor provides it.
• County-level shapefiles and demographic control datasets.
• Version control system such as Git for tracking your analysis.

Software & Tools

• QGIS: Maps county data and helps you spot spatial clusters.
• Python: Cleans merged datasets and runs correlation or regression models.
• R: Builds statistical tests and publication-style graphs.
• Google Sheets: Checks joins, flags missing values, and keeps a small working copy.
• Tableau Public: Builds an interactive dashboard from cleaned county data.

Experiment Steps

1. Define the county unit, the year range, and the exact AMR outcome you will measure.
2. Match the USDA, CDC, and census datasets on county and year, then clean missing or suppressed values.
3. Build a first-pass map and scatterplot set so you can see the pattern before modeling.
4. Add control variables like population density, rurality, and income so you can separate confounding from signal.
5. Test the relationship with a correlation or regression model, then repeat the analysis by livestock type or region.

Common Pitfalls

• Joining files with slightly different county names or FIPS codes, which creates bad matches and missing rows.
• Using raw counts instead of rates, which makes large counties look more affected than small counties.
• Ignoring suppressed or blank CDC values, which can flatten the pattern or bias your map.
• Treating antibiotic sales as the same thing as antibiotic exposure in livestock, which weakens your claim.
• Calling a correlation a cause, which goes beyond what county data can prove.

What Makes This Competitive

A stronger version of this project goes past a simple map. You compare multiple livestock types, add real control variables, and test whether the pattern survives when you change scale or remove outliers. You can also show that you understand the limits of public surveillance data, which makes the analysis more credible.

Project Variations

• Compare cattle-heavy, swine-heavy, and poultry-heavy counties to see whether one livestock type drives the pattern.
• Swap in hospital infection rates or antibiotic prescribing data to test a different human-health signal.
• Rebuild the dashboard at the state level to see whether geographic scale changes the strength of the relationship.

Learn More

• USDA NASS Quick Stats: Search county livestock counts and agricultural density data on the USDA NASS website.
• CDC Antibiotic Resistance Patient Safety Atlas: Find U.S. resistance maps and background data on the CDC website.
• CDC WONDER: Pull county and state health data for trend checks on the CDC website.
• U.S. Census Bureau American Community Survey: Get county controls like population density, income, and rural context on the Census website.
• PubMed: Search review articles on One Health, antimicrobial resistance, and livestock exposure.