data/README.md

# RHTP Evaluation — Data Folder

This folder is the **single replacement point** for the dashboard. Every
file here is synthetic, but its schema mirrors the real source it would
come from in production. To switch the dashboard from synthetic to live
data, drop a real file with the same name and column schema in place of
the synthetic one.

## File-by-file source mapping

### `geography/`
| File | Real-data source |
| --- | --- |
| `montana_counties.csv` | US Census TIGER county FIPS + MT DPHHS region map |
| `county_population_panel.csv` | American Community Survey 5-yr / Census Population Estimates Program |
| `montana_hospitals.csv` | MT DPHHS facility roster + CMS Provider of Services file |

### `facilities/`
| File | Real-data source |
| --- | --- |
| `facility_quarter_ops.csv` | CMS Hospital Cost Reports (HCRIS) + state quarterly facility metrics |
| `facility_quarter_treatment_status.csv` | RHTP CoE program tracker (state-internal) |
| `coe_implementation.csv` | RHTP CoE program tracker |
| `telehealth_activation.csv` | Facility self-report + CMS telehealth claims |
| `shared_services.csv` | RHTP shared-services participation roster |

### `workforce/`
| File | Real-data source |
| --- | --- |
| `workforce_county_year.csv` | HRSA Area Health Resource File + Montana DPHHS licensure |
| `service_commitment_awards.csv` | RHTP service-commitment program tracker |
| `residency_slots.csv` | ACGME accreditation reports + RHTP rural track tracker |
| `training_participants.csv` | RHTP training program rosters |
| `workforce_supports.csv` | RHTP supportive services program tracker |

### `claims/`
| File | Real-data source |
| --- | --- |
| `member_month_pmpm.csv` (+ `.parquet`) | Montana Medicaid + Medicare duals enrollment / claims |
| `county_quarter_treatment_status.csv` | Aggregated rollout indicator panel |
| `ed_high_utilizers.csv` | Medicaid claims (≥4 ED visits / 12mo) |
| `tnt_cpt_usage.csv` | EMS CPT 99X claims + state EMS billing extract |
| `pharmacy_prescribing.csv` | Medicaid pharmacy claims (NPI-of-prescriber level) |
| `outpatient_share.csv` | Hospital outpatient vs inpatient revenue (HCRIS / claims) |
| `ems_modernization.csv` | RHTP EMS modernization program tracker |
| `vbc_attribution.csv` | Montana Medicaid VBC attribution roster |

### `prevention/`
| File | Real-data source |
| --- | --- |
| `prevention_county_year.csv` | BRFSS, MT Vital Statistics, MT YRBS, HEDIS-aligned Medicaid measures |
| `school_sites.csv` | RHTP school-based health expansion tracker |
| `mobile_care_runs.csv` | Mobile clinic logbooks / EMR extract |
| `chap_rollout.csv` | Community Health Aide Program (CHAP) certification tracker |
| `crisis_infrastructure.csv` | DPHHS Crisis Continuum dashboard |

### `technology/`
| File | Real-data source |
| --- | --- |
| `hie_participation.csv` | Big Sky Care Connect participation roster |
| `ehr_modernization.csv` | ONC certified EHR registry |
| `bed_registry.csv` | DPHHS Bed Registry connection log |
| `bh_bed_wait_time.csv` | Behavioral health bed placement queue logs |
| `dashboard_usage.csv` | Internal RHTP analytics-hub usage log |

## Regenerating synthetic data
```bash
python3 scripts/generate_data.py
```
The seed in `RNG = np.random.default_rng(20260425)` makes every run reproducible.

## Synthetic-data notes
- All hospital names are clearly fictional / representative — the schema is
  real but no row maps to a specific real facility.
- Treatment effects are baked in deterministically (with realistic noise),
  so the models in the dashboard *should* recover positive effects on the
  outcomes the RHTP plan is designed to move.
- Years run 2019 → 2031 (5 baseline + 7 RHTP years).
- Member-level claims are limited to 2022 → 2031 (10y panel, 5,000 synthetic
  duals members → 600k member-months).