# California Medical Board DUI Discipline Dataset — V2.1

**Version** V2.2 · **Released** 3 September 2026 · **Supersedes** V2 (221 source rows presented as dispositions) and V1 (103 records)

## Recommended file

**`mbc-dui-matters.csv` / `.json` — CANONICAL DATASET. Use this for analysis.**
199 rows, one per disciplinary matter.

`mbc-dui-source-observations.csv` / `.json` — SOURCE OBSERVATIONS. Use this to audit provenance.
221 rows. **Never sum these for totals.**

## The V2.1 correction

V2 reported 221 source observations as 221 dispositions. That double-counted 22 matters found independently in both provenance layers. V2.1 introduces a canonical matter layer; every public aggregate is now computed from unique matters, and the source observations remain inspectable underneath.

## Definitions

| Term | Meaning |
| --- | --- |
| **SOURCE OBSERVATION** | A row extracted from one source by one read. 221 exist. |
| **MATTER** | A unique disciplinary proceeding after reconciliation. 199 exist. All public totals use this. |
| **LICENSEE** | A person. One licensee may hold several matters. 191 exist. |
| **REGISTER VERIFIED** | The displayed fields are supported by an official Medical Board Administrative Actions publication. |
| **DECISION VERIFIED** | The underlying final decision or order has also been read and reconciled against the register. |
| **CROSS VALIDATED** | The same matter was found independently in both provenance layers. This strengthens the evidence; it does not create a second matter. |

## Audit fields

| Field | Value |
| --- | --- |
| `source_observation_count` | 232 |
| `unique_matter_count` | 210 |
| `unique_licensee_count` | 202 |
| `matched_source_pairs` | 22 |
| `same_license_distinct_matters` | 13 |
| `unresolved_merge_candidates` | 3 |
| `open_date_discrepancies` | 2 |
| `decision_verified_count` | 0 |
| `register_verified_count` | 210 |
| `cross_validated_matters` | 22 |
| `dui_primary_matters` | 151 |
| `dui_primary_physician_matters` | 139 |

Invariant: source observations (232) − merges (22) = matters (210).

## Reconciliation

**Rule.** Two source observations merge into one matter only when they share a license number, share a disposition class, sit within four months, AND come from different provenance layers. Same-layer near-duplicates are flagged, never merged.

**Evidentiary order.** 1. Board case / accusation / decision number where available 2. medical license number 3. direct Board document identifier 4. effective date proximity within four months 5. disposition class 6. publication volume 7. conduct description

**Gupta.** GUPTA, ABHAY (A 88550) was NOT merged. The KPSGILL read records a 2025 probation-violation disposition citing prior case 800-2019-057074; the contributed read records a 2022 disposition. Three years apart with a prior case referenced: two distinct matters, two matter_ids.

## The DUI universe

dui_primary is true only where the Board publication expressly identifies driving under the influence, driving with a blood alcohol concentration at or above the statutory threshold, wet reckless or alcohol-related reckless driving, drug-impaired driving, an equivalent out-of-state driving offense, or a directly identified DUI arrest. Generic "drugs and/or alcohol" language is NOT classified as DUI.

- `dui_primary` matters: **140**
- `dui_primary` AND `physician_primary`: **128** ← principal analytic universe

## Two verification states

**REGISTER OBSERVATION** — The displayed fields are supported by an official Medical Board Administrative Actions publication. (199 of 199)

**DECISION-VERIFIED MATTER** — The underlying final decision or order has also been read and reconciled against the register. (0 of 199)

The Observatory codebook originally required a concluding document to exist *and have been read*
before a record was created. That conflicted with publishing register-derived observations, so the
codebook now permits a register-derived observational layer while reserving findings about
mitigation, aggravation, findings of fact, credibility, procedural route, ALJ reasoning, Board
reasoning, causal explanation, and charged-versus-sustained counts for decision-verified matters.

## Outcome distributions, each with its denominator

### All matters (n = 199)
- Public letter of reprimand: 16
- Public reprimand: 17
- Stayed revocation with probation: 100
- Stayed revocation, probation extended: 3
- Stayed revocation with actual suspension: 4
- Surrender of license: 34
- Outright revocation: 8
- Probationary license (applicant): 17

### DUI-primary (n = 107)
- Public letter of reprimand: 14
- Public reprimand: 7
- Stayed revocation with probation: 51
- Stayed revocation, probation extended: 1
- Stayed revocation with actual suspension: 2
- Surrender of license: 19
- Outright revocation: 5
- Probationary license (applicant): 8

### DUI-primary physician (n = 96) — principal universe
- Public letter of reprimand: 14
- Public reprimand: 7
- Stayed revocation with probation: 50
- Stayed revocation, probation extended: 1
- Stayed revocation with actual suspension: 2
- Surrender of license: 18
- Outright revocation: 4

### Revocation-type matters
Of the revocation-type matters, only the outright group ends the license. The stayed group contains a revocation whose operative effect is stayed while the licensee practices under the stated probationary terms. Stayed: **107**. Outright: **8**.

## Data dictionary — canonical matter layer

| Field | Type | Meaning |
| --- | --- | --- |
| `matter_id` | string | Stable identifier. One per disciplinary matter. |
| `licensee_name`, `license`, `city` | string | As printed in the source publication |
| `effective_date` | date | Effective date of the disposition |
| `effective_date_status` | enum | `single source read` / `DISPUTED` / `NOT ESTABLISHED` |
| `effective_date_alternatives` | date list | Competing dates from other layers, retained not overwritten |
| `outcome` | string | Published disposition, source wording |
| `disposition_class` | enum | letter / reprimand / stayed / stayed-ext / stayed-susp / surrender / revoked / applicant |
| `stay_present` | boolean | Order contains a stayed revocation |
| `published_conduct` | string list | Source publication's own charge language. **Not a finding.** |
| `route` | string | Resolution route where printed |
| `probation_months`, `actual_suspension_days`, `obligations_count`, `costs`, `probation_end` | number/date | As stated; null where not stated. **Null is never treated as zero.** |
| `dui_primary` | boolean | Meets the explicit DUI definition above |
| `driving_alcohol_explicit`, `driving_drug_explicit`, `wet_reckless_explicit`, `dui_bodily_injury`, `bac_008_or_more`, `bac_015_or_more`, `bac_020_or_more`, `arrest_only`, `conviction_explicit`, `plea_or_no_contest_explicit`, `alcohol_related_non_driving`, `condition_or_impairment_only`, `practice_under_influence`, `probation_violation`, `non_dui_comparator`, `applicant_case`, `physician_primary` | boolean | Controlled classification flags |
| `source_observation_ids`, `source_observation_count`, `provenance_layers` | list/number | Provenance |
| `cross_validated` | boolean | Found independently in both layers |
| `source_publications`, `source_urls`, `decision_urls` | list | Sources and recovered primary documents. A blank `decision_urls` does not mean no document exists. |
| `register_verified`, `decision_verified`, `matter_status` | boolean/enum | Verification state |
| `contributor_confidence` | enum | Retained as supplied by the contributing layer; not re-graded |
| `ambiguities` | string list | Fields the source does not establish |

## Known limitations

1. 0 of 199 matters are decision-verified. No finding about mitigation, aggravation, findings of fact, credibility, procedural route or causal explanation is published.
2. The register is a summary. Similar abbreviated summaries do not establish that the underlying cases were comparable.
3. No denominator for the enforcement funnel. Counts and distributions only, never rates.
4. Public letters of reprimand are removed from the Board website after the ten-year posting period, so the earliest years are systematically under-represented in any present-day search. Historical documents remain obtainable from the Board by request.
5. The contributed layer states it is not an exhaustive fifteen-year census and identifies 2013–2016 as its largest reconciliation gap.
6. Contributor confidence is retained as supplied and was not re-graded upward: 93 High, 16 Medium-High, 9 Medium of 118 contributed rows.

## Change log

| Version | Date | Change |
| --- | --- | --- |
| V2.6 | 2026-09-04 | **Data-integrity release.** Headline denominator corrected: a registry named for DUI had been totalling the whole corpus. Seven universes now exist; the headline is MBC_MD_DUI_ACTIVE_LICENSE = **135** physicians holding an active licence on a driving disposition. **Five duplicate matters merged** (M0005, M0008, M0010, M0047, M0049) after the "same-provenance near-duplicates are never merged" rule was deleted — three pairs shared one decision document ID, one date and one disposition. Matters 210 → **205**. **"0 decisions reviewed" corrected**: 107 primary decision documents reviewed, 105 procedural routes established from them; the zero belongs to `full_codebook_second_read` and is published under that label alone. Span moved to **2008–2026** on the Board's own publication seam. Fifteen historical candidates disclosed, zero promoted. Dataset 02 (Osteopathic) opened, never merged. New gate: `scripts/validate-dui-counts.mjs`. |
| V2.4 | 2026-09-04 | **6 more matters added (M0205–M0210)** in a second same-day supplemental pass: MAZZARELLA (G82089) and CIRESI (G60858) confirmed against the same audit supplement's cited registers; KIM, JOON (A77889) added with the register's own outcome after the supplement's 'surrender' claim was disproved; BUNCKE (G72447), DOMINGUEZ (A114123) and MARTINEZ (A78740) found on direct primary-source read of a register page opened for another matter — none were in either provenance layer or the supplement. CISEK/APAYDIN re-checked and still absent from the cited register; MOFU/NEILAN/SOHAL/LI/SALINAS remain unresolved. This dataset is not a full census: ~60 quarterly MBC newsletters span 2011–2026 and most have not been read in full — the discovery of 3 uncatalogued matters in one already-open page is a strong signal more remain. See `changelog_v24` in the JSON. |
| V2.3 | 2026-09-04 | **5 matters added (M0200–M0204)** from a contributor-supplied audit supplement, each independently re-verified against a freshly-fetched official MBC register before inclusion (see changelog_v23 in the data file for the full reconciliation note, including a contributor error caught and a mismatched citation excluded). Matters 199 → 204; DUI-primary 140 → 145; physician 128 → 133. |
| V2.2 | 2026-09-03 | **Classification fix.** Per-se BAC offenses printed as "BAC per se" / "BAC 0.08% or greater" without the word "driving" were not flagged `dui_primary`. 33 matters reclassified; DUI-primary 107 → 140, physician 96 → 128. No other field changed. |
| V2.1 | 2026-09-03 | **Canonical matter layer introduced.** All aggregates recomputed from 199 unique matters instead of 221 source rows. Explicit `dui_primary` definition replaces implicit DUI selection, resolving an unexplained 49-vs-42 denominator. Codebook reconciled with the two verification states. Page rebuilt with no V1 remnants; every figure carries a visible denominator. Standardized on "license". |
| V2 | 2026-09-03 | Contributed 2011–2026 registry integrated as a second provenance layer. **Defect: 221 source rows were reported as 221 dispositions, double-counting 22 cross-validated matters.** |
| V1 | 2026-09-03 | 103 records, KPSGILL layer only, Volumes 163 and 167–172, May 2022 – September 2025. |

## Citation

> KPSGILL California Medical Discipline Observatory, California Physician DUI Discipline
> Registry, 2008–2026, matter-level, last reconciled 4 September 2026, compiled from
> Medical Board of California publications.
> https://kpsgill.com/research/medical-discipline-observatory/california-dui-discipline/dataset/