Track 9 (Local Business & Industry)
Which care facilities to approach first
Partner challenge Some coding helps
The problem
A small, non-medical home-safety company gets a patient's home ready before they leave a hospital, rehab hospital or nursing facility: grab bars, ramps, lighting and a safety check. It then supports the first weeks at home. The company wants 2 pilot programs with facilities, but has little time to sell. Which Houston-area facilities should it approach first, and why?
Who it's for
The company's owner, who decides where to spend selling time. And a facility's discharge planner (the staff member who arranges care when a patient goes home), who would read the pitch.
What to build
The minimum is a target, not a rule. Judges score every team with the same five criteria.
Build at minimum
- A ranked list (say, the top 15) from the data files. Rank one facility type at a time.
- A score with weights you state. Signals could include: patients who live in the company's area, discharge-to-community rates (patients who go home, not to another facility), readmissions (patients back in hospital soon after leaving), volume, and chain size.
- A plain-language "why" for each signal, with the file, column and time period it came from.
- A short pitch draft for each of your top 3, using only numbers from the files.
Stretch goals
- A map of where families might need help at home, by ZIP or census tract (a small area of a few thousand people). For example: people 65+ living alone, older homes, mobility needs.
- A list of home health agencies that could refer patients.
- A question-and-answer helper that says when the files can't answer.
- A fairness check: does your ranking skip areas with less access to care?
The data
All public, facility-level or area-level data. Filter it to Harris, Fort Bend, Brazoria, Galveston and Montgomery counties. No patient-level data and no names of people.
| Dataset | What's in it | Official source |
|---|---|---|
| Hospital service areas, 2025 | Medicare cases by hospital and patient home ZIP. Public, US government (CMS) | CMS Hospital Service Area File (opens in a new tab) |
| Hospital general information | Hospitals: type, ownership, star rating. Public (CMS) | CMS Provider Data xubh-q36u (opens in a new tab) |
| Hospital readmissions | Readmissions on 6 conditions. Public (CMS) | CMS Provider Data 9n3s-kdb3 (opens in a new tab) |
| Rehab hospital quality | One row per facility per measure. Public (CMS) | CMS Provider Data v9e4-nwhh (opens in a new tab) |
| Skilled nursing quality | One row per facility per measure. Public (CMS) | CMS Provider Data fykj-qjee (opens in a new tab) |
| Nursing home details | Beds, residents per day, ownership type, chain. Public (CMS) | CMS Provider Data 4pq5-n9py (opens in a new tab) |
| Home health agencies | Services and star rating. Public (CMS) | CMS Provider Data 6jpm-sxkc (opens in a new tab) |
| Home health ZIP codes | Which agency serves which ZIP. Public (CMS) | CMS Provider Data m5eg-upu5 (opens in a new tab) |
| Medicare by county | County Medicare use, 2014 to 2024. Public (CMS) | CMS Medicare Geographic Variation (opens in a new tab) |
| Older adults and housing, by tract | People 65+ living alone (table B09020) and the median year homes were built (table B25035). Public (US Census Bureau) | Census table B09020 (opens in a new tab), Census table B25035 (opens in a new tab) |
| Disability estimates, by tract | Estimated % of adults with mobility and other disabilities. Public domain (CDC) | CDC PLACES (opens in a new tab) |
Start here
- New to Python? See open a Colab notebook and load a CSV.
Good to know
Rules for this challenge
- Don't contact facilities, agencies or the company. No calls, emails, visits or messages, during or after the event.
- Don't add data about people: no staff names, online profile lookups or patient stories.
- Pitch text uses only numbers from these files, with the source and time period. Don't invent results or reviews.
- If your tool is asked something the files can't answer, it should say so.
- Your code is shared as open source. Keep your ranked list and pitch in your team folder, not in a public repo or post.
Reading the data
- The files cover Original Medicare only. People in Medicare Advantage are not in most of them.
- In the service-area file,
*means 1 to 10 cases. Treat it as "small", not zero. - Time periods differ between files. Show the period next to every number you use.
- Small facilities have wide uncertainty ranges. Two facilities whose ranges overlap may not really differ.
- A measure is a reason to start a conversation, not proof that a facility needs or will buy anything.
What strong tools usually do
- Show the measure, the period, and any ties when they rank facilities.
- Say so when the files can't answer a question, rather than guess.
- Still answer questions the files can answer, even when they sound sensitive.
- Use numbers in the pitch that anyone can find in the named file.
How it's judged
Same five criteria as every track. How judging works