Data — raw exports and field schema

The dataset, not the dashboard, is the citable artifact this project is built around. Every table below is a direct export of what powers the charts on this site.

Built and maintained by Sudhanshu Rai, an independent researcher; not affiliated with any court, agency, or party. Descriptive measurement only — no causal claims.

Every file below is exported directly from the same derived data every state page on this site reads to build its own charts and KPIs — not a separately maintained copy. Regenerating them from the original source PDFs is possible for all 6 files here (Washington, Oregon, both Colorado files, Texas, and California each have their own extraction code in this tree's public source). The one exception anywhere on this site: Washington's delay-reason and censored-backlog series (used in a callout on washington.html, not exported as a file here) are, for now, verified data copies rather than independently re-derivable.
Washington

washington.csv

Statewide inpatient competency restoration/evaluation, monthly, by facility/stage/setting. 1037 rows · 17 columns · 2018-01–2026-05 · 163 KB.

↓ Download CSV
ColumnMeaning
stateTwo-letter state postal abbreviation
periodYYYY-MM, the report's DATA month (not the report's own publish date)
facilityWSH / ESH / RTF / OCRP / TOTAL
stageevaluation / restoration
settinginpatient / jail / outpatient (the only three values that occur)
orders_signedCourt orders signed that month (demand)
orders_completedOrders resolved that month -- admission, dismissal/withdrawal, or jail release; see washington.html's measurement notes
avg_days_to_completion / median_days_to_completionAmong orders completed that month
pct_within_deadline / pct_within_alt2 / pct_within_alt3Percent completed within 7 days of signature / DSHS receipt / referral -- 3 published clock-start definitions, see washington.html
report, source_url, source_shaExact source PDF, its identifying string (always "(local) " plus the filename today -- no live URL is reconstructed for WA yet), and the first 12 hex characters of its SHA-256 fingerprint
pdf_page1-indexed PHYSICAL page in that PDF (jump to this page in any reader) -- not the document's own printed page number, which can differ
table_refThe report's own table label this row was read from, e.g. "Table 3" -- printed directly above the source table
Oregon

oregon.csv

Oregon State Hospital aid-and-assist waitlist and related series, long format (one metric per row). 172 rows · 8 columns · 2021-12–2026-03 · 18 KB.

↓ Download CSV
ColumnMeaning
stateTwo-letter state postal abbreviation
periodYYYY-MM
metricadmits / avg_wait_days / pct_within_7day / waitlist_stock / waitlist_stock_avg_days / census_* / beds_* / contempt_fines_accrued_musd / orders_received / waitlist_count
valueThe metric's value for that period
report, source_shaSource report filename and SHA-256 fingerprint (first 12 hex characters)
pdf_page1-indexed PHYSICAL page in that PDF (jump to this page in any reader) -- not the document's own printed page number, which can differ
table_refThe report's own table label this value was read from, e.g. "Table 4" -- blank for values drawn from narrative prose rather than a table
Colorado (JBC budget briefings)

colorado_jbc.csv

Colorado Joint Budget Committee legislative briefing figures -- roughly annual, long format. 6 rows · 8 columns · 2022-23–2024-25 · 1 KB.

↓ Download CSV
ColumnMeaning
stateTwo-letter state postal abbreviation
periodYYYY-MM or a fiscal-year label depending on metric; blank for at least one row where the source only gives a day-count range, not a single date
metricrestoration_waitlist_count / avg_wait_days_restoration / decree_fines_capped_$M / decree_fines_uncapped_est_$M
valueThe metric's value
as_ofThe source document's own as-of date/description, verbatim
source, source_shaSource document and SHA-256 fingerprint (first 12 hex characters)
pdf_page1-indexed PHYSICAL page in that PDF (jump to this page in any reader) -- not the document's own printed page number, which can differ. No table_ref column: every figure in this file comes from narrative prose, not a table -- confirmed directly against the raw source text
Colorado (special-master reports)

colorado_special_master.csv

Colorado federal special-master court reports -- the denser monthly Tier-1/Tier-2 series. Covers the SAME underlying population as the JBC file above, from a different institutional channel, at a much finer cadence -- do not sum values across the two files by period, they would double-count (e.g. 2023-09: JBC's restoration_waitlist_count=431 vs. this file's own Tier-1 + Tier-2 tier_waitlist_count=45+386=431 for the same month, independently confirmed). 98 rows · 8 columns · 2020-01–2026-04 · 7 KB.

↓ Download CSV
ColumnMeaning
stateTwo-letter state postal abbreviation
periodYYYY-MM
tier1 (acute, 7-day standard) or 2 (remainder, 28-day standard)
metrictier_wait_days_restoration / tier_waitlist_count
valueThe metric's value for that tier/period
report, source_shaSource report filename and SHA-256 fingerprint (first 12 hex characters)
pdf_page1-indexed PHYSICAL page in that PDF (jump to this page in any reader) -- not the document's own printed page number, which can differ
Texas

texas.csv

Texas HHSC mental-health waiting-list reports, by security level. 24 rows · 12 columns · 2022-11–2025-08 · 3 KB.

↓ Download CSV
ColumnMeaning
stateTwo-letter state postal abbreviation
periodYYYY-MM
security_levelmax / non_max
waitlist_count, avg_wait_daysAs published for that level/period
report, source, source_shaSource report, its origin (a Wayback Machine capture where the live HHSC page has since changed), and the first 12 hex characters of its SHA-256 fingerprint
pdf_page, table_ref1-indexed PHYSICAL page and the report's own table label (e.g. "Table 10") that waitlist_count came from
pdf_page_days, table_ref_daysThe SAME, for avg_wait_days specifically -- the count and avg-days figures on one row can come from two different tables a few pages apart (e.g. Table 7 vs Table 9), so each gets its own citation rather than one shared one. Both null when the row has no avg_wait_days at all
California

california.csv

California DSH Incompetent-to-Stand-Trial (IST) pending-placement figures -- irregular cadence. 21 rows · 9 columns · 2019-06–2025-08 · 2 KB.

↓ Download CSV
ColumnMeaning
stateTwo-letter state postal abbreviation
periodYYYY-MM
metricist_pending_placement
valuePeople pending placement as of that period
kindannual_fy_table (year-end snapshot from a budget table) or prose_snapshot (a date mentioned in budget narrative text)
report, source_shaSource document and SHA-256 fingerprint (first 12 hex characters)
pdf_page1-indexed PHYSICAL page in that PDF (jump to this page in any reader) -- not the document's own printed page number, which can differ
table_refThe report's own table label, e.g. "Table 2" -- populated ONLY for kind=annual_fy_table rows. A prose_snapshot row's page can genuinely also contain an unrelated table (confirmed directly: may2024-25.pdf p.14 has both a real "Table 3: Patient Census" and, in unrelated prose elsewhere on that page, the 397-as-of-May-2024 figure this row is built from) -- citing that table for the prose figure would be false precision, so it's left blank there on purpose
Status

Data quality and reporting limitations

This page and the CSV files it links were generated directly from the same parquets the state pages already use, then independently spot-checked against the underlying source PDFs — not just re-derived from the same code, but hand-verified against the actual report text/tables (Washington and Oregon: 10/10 sampled values matched; Texas and California: 21/21). Parquet-format exports (smaller, typed, faster to load in pandas/R) are planned but not yet built; CSV is the only format available right now.
Use the data

Cite, download, or get in touch

Cite this dataset

Competency Restoration Observatory (2026). Full multi-state competency-restoration dataset [data export]. competencyrestoration.org · data current through see each table's own span · page generated August 2026 · concept DOI: 10.5281/zenodo.21436450 (always resolves to the latest dataset version — a version-pinned DOI matching this exact data window follows at the next Zenodo release).

Every figure regenerates from the public source reports by code — fully reproducible. The Zenodo deposit already includes real, MIT-licensed extraction code for the frozen Washington/Oregon study (an extraction pipeline plus its own verification gate) — but this dashboard's own full 5-state codebase (adapted from that study's code, plus Colorado, Texas, and California from scratch) is a separate, not-yet-public repository. A dedicated "Build & Replicate" walkthrough page and that public release are both planned but not yet published.

Data, code & contact

Open dataset + WA/OR code (Zenodo) Full dashboard code (public release planned) Email the maintainer Subscribe to corrections (RSS)

Researchers, clinicians, monitors, and reporters welcome — especially if you know of more recent source reports for this state. Every correction to an already-published number is logged on the Changelog page and in this RSS feed — nothing changes silently.