OSHA Severe Injury & Fatality Dashboard

Interactive dashboard for exploring 101,996 severe injury reports (Jan 2015 – Jun 2025) and 22,865 fatality/catastrophe records reported to federal OSHA — built with Streamlit and Plotly.

Live demo: osha-dashboard-cenuscpehyevclrnznyhtv.streamlit.app

What it shows

• KPIs — total severe injuries, hospitalizations, amputations, hospitalization rate, and fatalities for any filtered slice
• Trends — monthly injury counts and severity (hospitalization vs. amputation) over a 10-year span
• Geography — US choropleth of severe injuries by state, with a sortable state breakdown
• Analysis — top industries (NAICS mapped to plain-English names), injury type distribution, body parts affected, and top injury mechanisms
• Raw data access — filterable record view with CSV export

Filters: date range, state, injury type (OIICS nature), and industry (NAICS).

Data sources

Dataset	Source	Records
Severe Injury Reports	OSHA Severe Injury Reports — employer-reported hospitalizations and amputations under 29 CFR 1904.39	101,996
Fatality/Catastrophe Data	OSHA fatality reports	22,865

Important caveat for interpreting this data: severe injury reporting under 29 CFR 1904.39 covers federal OSHA jurisdiction only. The 20+ State Plan states that operate their own programs (California, Washington, Michigan, and others) are absent or undercounted here — which is why Texas, Florida, and Ohio dominate the map. Counts reflect reported events, not incidence rates; they are not normalized by employment. This is a known limitation of the dataset, not a finding.

Run locally

git clone https://github.com/dgrinnell/osha-dashboard.git
cd osha-dashboard
python3 -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
streamlit run app.py

Project structure

osha-dashboard/
├── app.py                 # Streamlit entry point: layout, filters, KPIs, tabs
├── src/
│   ├── data_loader.py     # Cached data loading with optimized dtypes
│   ├── charts.py          # Plotly chart builders
│   └── utils.py           # NAICS and state lookup tables
├── data/
│   ├── severe_injuries.csv
│   └── fatalities.xlsx
└── .streamlit/config.toml # Theme

Performance notes: columns load with categorical/32-bit dtypes to keep the 100k-row dataset memory-light, and st.cache_data (5-min TTL) avoids re-parsing on every interaction.

Deploy your own

1. Fork or clone this repo to your GitHub account
2. Sign in at share.streamlit.io with GitHub
3. New app → select the repo, branch main, main file app.py → Deploy

About

Built by Dan Grinnell, MS, CIH — Certified Industrial Hygienist working at the intersection of AI, data, and EHS practice. More at ai4ehs.com and the AI 4 EHS newsletter.

License

MIT