---
title: RHTP Evaluation Dashboard
emoji: 🏥
colorFrom: red
colorTo: blue
sdk: docker
app_port: 8501
tags:
- streamlit
- rural-health
- causal-inference
- montana
pinned: false
short_description: Preregistered causal evaluation of all five RHTP initiatives
---

# RHTP Evaluation Dashboard

Interactive dashboard implementing the **RHTP_models_full** modeling playbook for
the Montana Rural Health Transformation Program. Six pages:

1. **Overview** — interactive Montana county map with all 65 hospitals, top-line stats
2. **Initiative 1 — Workforce** — DiD + lagged-intervention TWFE + event study (county-year)
3. **Initiative 2 — Facility Sustainability** — staggered event-study + multi-intervention intensity (facility-quarter)
4. **Initiative 3 — Innovative Care & Payment** — county-quarter + member-month PMPM with risk adjustment
5. **Initiative 4 — Community Prevention** — TWFE with 3-yr smoothing for rare outcomes
6. **Initiative 5 — Technology & Data** — TWFE for direct effects + logit with marginal effects

Every initiative page follows the same three-section pattern: **Data Explorer →
Models → Was the initiative successful?** (six-criteria verdict).

## Run locally
```bash
pip install -r requirements.txt
python scripts/generate_data.py     # one-off, creates /data
streamlit run src/streamlit_app.py
```

## Run in Docker
```bash
docker build -t rhtp-eval .
docker run -p 8501:8501 rhtp-eval
```
The Dockerfile generates synthetic data at build time. To run with real data,
mount a host directory over `/app/data` containing files of the schema described
in [`/data/README.md`](data/README.md) — no code changes needed.

## Switching from synthetic to real data
Every file in `/data/` has a real-world source mapping documented in
`/data/README.md`. Drop in a real file with the same column schema and the
dashboard works without code changes.