School Immunizations

The goal for this project is to format county-level data on school vaccinations obtained from the states.

Publish the granularity the state publishes. If a source is by school or by district, keep those rows and add school_name (with district/city where the source has them) rather than aggregating in the ingest — the combined file rolls them up to county itself, and it can only do that correctly if it can see the denominator on each row. Aggregating early is how WI came to report a 95% county religious-waiver rate: it averaged school percentages with no enrolment to weight them, so one small school set the county figure.

The output is wide: index columns (geography, time, and any stratum the source publishes, such as grade) plus one column per measure. The vaccine or exemption category goes in the COLUMN NAME — rate_mmr, rate_medical_exempt, N_full_exempt — not in a vax value column. Categories in use:

dtap · polio · mmr · hep_b · varicella
personal_exempt · medical_exempt · religious_exempt · full_exempt

Data source status by state

Not every state's data comes from an automated pull. data/DATA_SOURCES.md tracks, for every state, the public source, how it's accessed (API, static download, dashboard-only, or by-request), what the ingest produces, and the public sources that are known but not ingested yet. Check it before starting work on a state. Each state folder also carries a machine-readable sources.json with the URL, discovery pattern and publishing calendar of every source. A few things worth knowing up front:

• DC, GA and NE have no data: DC Health blocks non-residential clients and the other two have no public file; all three are stubs.
• DE and WY are statewide or registry-based: Delaware publishes only statewide charts, and Wyoming's county report cards come from the immunization registry rather than a school survey.
• AK, WY, ID, SD and WV raw files were supplied to the project, not downloaded by ingest.R; there is no automated source behind them.
• PDF sources are parsed with resources/pdf_table.R (IA, TN, AL, SC, DE); states whose PDFs are charts without value labels (MS) or whose pages carry no file (KY, KS, NJ, UT, VA) stay exemption-only.
• AR only has school-district-level data — Arkansas publishes no county file, so its standard/data.csv.gz has district rows only, not county rows.
• MA cannot be fetched from CI (the mass.gov WAF rejects datacenter IPs); new years are fetched from a workstation and committed.

Fetching and monitoring

Network ingests download through resources/fetch.R: the file goes to a temporary path, is validated (workbook signature, CSV header, PDF magic, bot-block page detection), and only then replaces the committed copy. A failed download keeps the committed file, warns, and records the failure in process.json. Parsing only runs when a raw file or the script changed.

Two reports are regenerated by CI:

• data/BUILD_STATUS.md, written by scripts/build_status.R at the end of the nightly build; the job fails if an ingest errored or a fetch failed with nothing to fall back on.
• data/SOURCE_STATUS.md, written by scripts/check_sources.R weekly (.github/workflows/check_sources.yaml): checks every declared URL, lists files posted upstream that are not in raw/, and flags states whose latest ingested year is behind the publishing calendar. Run it locally with Rscript scripts/check_sources.R --states=MN,TX.

Other notes

This is set up as a Data Collection Framework project, initialized with dcf::dcf_init.

You can use the dcf package to check the source projects:

dcf::dcf_check()

And process them:

dcf::dcf_process()