Historical SAMHSA Treatment Facility Directories
The website now displays preliminary national, regional, and state aggregates. These are current parser outputs, not a validated data release. Address-level downloads, county geography, service filters, and facility linkage remain blocked until their QA gates pass.
This project reconstructs address-level treatment facility listings from 22 SAMHSA directories: 1998, 2000, 2001, and 2003-2021. It preserves both the directory year and the survey year represented by the directory.
Open the preliminary dashboard
Planned Downloads
The first validated release will include:
- Facility-year files in compressed CSV, Parquet, and RDS formats.
- Facility-service records and year-specific service availability.
- Stable cross-year facility linkage with confidence and review status.
- Geocoding results, county FIPS, provenance, and parser QA flags.
- Source manifests, SHA-256 checksums, data dictionaries, and a QA report.
Historical phone numbers will remain in the research downloads with a stale-contact warning. They will not appear in the dashboard.
Methods
The parser uses year-specific positional layouts for early OCR-heavy, three-column, and 2018-2021 two-column directories. Every listing retains its source PDF, page, column, raw record text, and parser warnings. Service codes are interpreted against each directory’s introductory legend, so a service can be represented as offered, not offered, or not asked in that year.
Street addresses are submitted in batches to the Census Geocoder. ZIP/county crosswalk assignments are used only as a lower-confidence fallback. Stable facility IDs prioritize exact normalized address and name matches, followed by geographically constrained fuzzy matching; uncertain matches remain unlinked.
Coverage
Headline dashboard totals will cover the 50 states and District of Columbia. Territories remain in the downloads and will be summarized separately. Regional summaries include Census region, Census division, and a Pacific Northwest grouping of Washington, Oregon, and Idaho.
These are historical records and are unsuitable for locating current treatment. SAMHSA also cautions that pre-2021 N-SSATS and post-2020 N-SUMHSS data are not directly trend-comparable.
Code And Citation
The parser code is licensed under MIT. Source data retain SAMHSA provenance and public-domain terms. A source-rights audit found a public-domain and reproduction notice in all 22 local directories; this documentation is not legal advice.
The release citation and versioned download links will appear here only after the QA report marks the release ready.
