Overview

This project checks the Pennsylvania Department of Health's (PDOH) website every day for new measles cases in each county, and keeps a running daily and weekly count of the outbreak.

How it works

A scheduled job runs every day at 5:05pm ET:

1. The scraper checks PDOH's measles page and compares the numbers it finds there to what's already saved in measles_daily.tsv
2. If a county's case count went up, it adds a new row to measles_daily.tsv for that increase
3. It then updates measles_weekly.tsv: for each week, it adds up that week's new cases from measles_daily.tsv, skipping any week we've corrected by hand, and recomputes cumulative_cases as a running total across all weeks
4. It also updates measles_daily_age_group.tsv: PDOH's dashboard only exposes a statewide year-to-date cumulative case count broken out by age group (not by day), so each run compares each age group's cumulative total (cumulative_cases) to what's already recorded and, for any group that went up, adds a row for the increase (new_cases)
5. It also updates measles_daily_hospitalization.tsv: PDOH's Hospitalization tab exposes a statewide running total broken out into three cards — total, under 18 ("children"), and 18+ ("adult") — but not by day or county, so each run adds one row per category, recording that category's total as cumulative_hospitalizations and working out new_hospitalizations by diffing it against that category's most recent prior day's cumulative total on record

You can also run it manually: Actions → Scrape → Run workflow

Repo structure

measles-tracker/
├── scraper/
│   └── measles_scraper.R     
├── data/
│   ├── measles_daily.tsv      
│   ├── measles_weekly.tsv     
│   ├── measles_daily_age_group.tsv
│   └── measles_daily_hospitalization.tsv
└── .github/
    └── workflows/
        └── scrape.yml

Known limitations

PDOH changed how it displayed new cases on its website in early July 2026, which broke the scraper from July 8–24, 2026. That gap is filled in measles_daily.tsv as one row per county, all dated July 24, 2026 (the day the scraper caught up) rather than broken out day by day. In all, 38 cases were confirmed during this 16-day gap.

Left alone, all cases would land in the week of July 20, 2026 in measles_weekly.tsv, creating an artificial spike in cases. PDOH separately reported that 29 new cases were confirmed in the 7 days before July 24, so we used that number to split the 38 by hand: 29 cases to the week of July 20, and the remaining 9 (38 minus 29) to the week of July 13. Those two weeks' new_cases values are manually adjusted and excluded from the scraper's usual auto-sync (see ADJUSTED_WEEKS in measles_scraper.R) so they don't get overwritten on the next run.

Hospitalizations were originally tracked weekly (as a hospitalizations column on measles_weekly.tsv) before measles_daily_hospitalization.tsv started tracking them daily, by category, on August 28, 2026. That earlier weekly history wasn't carried over; measles_daily_hospitalization.tsv instead starts from a baseline of three rows (August 28, 2026: total 83, children 24, adult 59) and tracks each category day by day from there.

Age-group tracking started August 26, 2026. Each age group's first row in measles_daily_age_group.tsv (dated August 26, 2026) leaves new_cases blank rather than recording PDOH's full cumulative total as if all of it were new that day — it was just where each group's count already stood when tracking began. cumulative_cases on that row still carries the correct starting total.

Local development

Rscript scraper/measles_scraper.R

Requires the R packages listed at the top of measles_scraper.R (install once via install.packages(c("httr2", "rvest", "dplyr", "readr", "stringr", "jsonlite"))).