California Physician DUI Discipline: The Record, the Registry, and What the Register Cannot Explain

I. Take the names away

Take away the cities, the attorneys, and everything else a reader might unconsciously use to distinguish one physician from another. Now put a series of Medical Board of California disciplinary summaries on a table.

One physician has an alcohol-related driving conviction and receives a public letter of reprimand. Another receives a public reprimand. Another has the licence formally revoked — but the revocation is stayed — and continues to practise under three, four, five, seven years of probation. Another surrenders. Another is revoked outright.

These are not hypothetical dispositions. They are in the Board’s own published record, and after reconstructing a matter-level registry from that record the interesting question is no longer whether a DUI can produce medical-licence discipline. It plainly can. The harder question is the one an informed reader, or a regulator handed only these published summaries, would have to answer.

Frequently the public register does not contain enough to do it. That is the central finding of this project, and it reduces to a single sentence.

That claim is narrower than it may first appear, and the narrowness is deliberate. It is not a finding that the Board’s decisions are arbitrary, inconsistent, biased, or driven by the quality of a physician’s counsel. Those are claims this dataset does not establish and is not built to establish. It is a finding about the publication surface — about what the regulator’s own quarterly instrument carries out into the world — not about the adjudication behind it.

II. What this is, and what it is not

This is an observational research corpus. It is not a census of California physician DUI discipline, and no figure on this page should be read as one.

The registry holds 205 reconciled disciplinary matters drawn from the Board’s published record. Of those, 140 are physicians holding an active Medical Board licence on a driving disposition. That is the principal analytic universe, and it is the denominator behind every physician figure on this page. Applicants, podiatric and polysomnographic licensees, reinstatement petitions, alcohol matters with no driving element, and deliberately retained non-DUI comparators are each counted in their own universe and never folded into a physician number. Universe membership is published as a column on every row of the registry, so the denominator can be reproduced by filtering the table.

Completeness is asserted period by period, never for the span as a whole. One period is bounded-complete today: January–March 2025. Everything else is partial. The full status table is at §XIII.

Verified matters currently begin in 2011. The programme covers 2008–2026 for the reasons set out in §III, but the 2008–2010 window sits inside the programme’s stated span and not yet inside its data. Its candidates are disclosed, not counted.

The count by year is a map of the reading, not of physician behaviour. The recent Administrative Actions registers are the ones read most completely here. The pre-2020 layer is barely read. A rising line in this dataset means read more completely. Anyone reading it as happening more often has drawn the one conclusion this project’s method exists to prevent.

III. The archive decides what can be seen

A study of this record could begin in 2011, where a present-day search of the Board’s document system starts to return results reliably. That would be an artefact of where reading happens to begin, not of anything about California. The Board’s own publication history supplies a real boundary.

Until the summer of 2008 the Medical Board published a monthly Hot Sheet — a mailed summary of administrative actions covering the Board and, at the time, podiatric medicine, psychology, physical therapy and the physician assistant committee. The archive runs from December 2005 to July 2008. The July 2008 issue opens with a notice to subscribers that it is the final one: in place of the monthly mailing, notifications of disciplinary actions would be sent electronically, by email, throughout the month. The Board’s current publications page states the same thing from the other side — since August 2008 it has distributed notices of administrative and disciplinary actions electronically through a subscriber list service, as the actions become effective.

So the record has a seam in it, and the seam is dated. Before August 2008, the public surface is a monthly mailed bulletin. After it, a stream of individual emails, later summarised in quarterly registers. Those are different instruments with different completeness properties. The programme is therefore 2008–2026: Hot Sheets for January–July 2008, the electronic-alert era from August 2008 forward.

The seam has a practical consequence that matters more than the tidiness. A present-day search of the Board’s licence-verification interface is not a historical census mechanism. Public letters of reprimand age out of it after ten years, and the alert emails that carried the original accusation and final-action links live in subscriber inboxes rather than on any public index. Reconstructing 2008–2012 means reading preserved bulletins and preserved mail, not querying a database.

One example of what that turns up. A licensee absent from the 2021 register a source had cited for it appears instead in the Decisions section of the June 2008 Hot Sheet: same name, same city. The source had the right physician and the wrong decade. That record is a disclosed candidate awaiting a full read, not a matter, and it is exactly the kind of thing a 2011 start date is structurally unable to find.

IV. What the register means by its own words

The Board is precise about its vocabulary, and the precision is published. The Hot Sheets carried the definitions in the margin of every issue. They are worth restating plainly, because the shorthand travels badly.

An accusation, the Board writes, consists of charges and allegations which still must undergo rigorous tests of proof at later administrative hearings. A petition to revoke probation is filed when a licensee is charged with violating a prior disciplinary decision — a later stage of one history, not a fresh event. An accusation withdrawn means the charges were dropped before any hearing; an accusation dismissed means the matter went to litigation and the licensee prevailed. And a probationary licence, in the Board’s own framing, is what a licensing agency may issue where good cause exists to deny a licence outright: a conditional grant, not a disciplined practising licence.

Read those definitions back against the register and the argument of this page narrows usefully. The Board is not publishing a sanction table and has never claimed to be. Each of its categories encodes a procedural position, and the abbreviated entry that reaches the public frequently omits the facts that determined which position a case reached.

The defensible claim is therefore not that the Board has no framework. It is that the public record often does not carry enough of the framework’s inputs for an outside reader to reconstruct the output — a criticism of the publication surface, not of the adjudication behind it.

V. Four distinctions this registry keeps

VI. The outcome spectrum, with its denominator attached

Physician DUI universe only. n = 140 of 205 reconciled matters.

Arranged for navigation, not as a statutory scoring system.

Two further universes are stated separately so that nothing has to be reverse-engineered.

All reconciled matters — every conduct category. Denominator: 205

Every DUI universe combined — physicians, applicants, other Board licence types, probation enforcement. Denominator: 154

The combined figure is not the headline. It is published so that a reader who wants the whole driving-related caseload can have it without the physician denominator being quietly widened to produce it.

VII. “Revoked” is itself capable of misleading the public

When someone reads that a physician’s licence was “revoked,” the natural assumption is that the physician can no longer practise. That is not necessarily what the Board’s disciplinary vocabulary means.

A large category of cases is revoked, stayed; placed on probation. The revocation exists in the disciplinary order, but its operation is stayed while the physician complies with the probationary terms. The physician therefore continues to practise, subject to the order. That is fundamentally different from outright revocation.

Both appear as “revocation” in casual summary. Crude statistics about “revocations” that combine them are not slightly imprecise; they are wrong by an order of magnitude in the direction that matters most to a patient trying to work out whether the physician in front of them is practising under supervision or should not be practising at all. This registry keeps them separate everywhere, and that seemingly small coding decision changes how the entire disciplinary record reads.

A stayed revocation is also not a trivial outcome. It commonly carries years of biological-fluid testing, abstention requirements, employer notification, a worksite monitor, psychotherapy, evaluations, and restrictions on supervising advanced-practice clinicians. A public reprimand carries none of that. Compressing both into “the physician was disciplined” discards the distinction the whole disciplinary structure is built on.

VIII. One publication period, many outcomes

The contrasts are not artefacts of comparing the 2010s with the 2020s. They appear inside a single interval of Board reporting — Medical Board of California, Newsletter Volume 173, Administrative Actions.

Read merely as outcomes, that spectrum is striking: letter → reprimand → three years → four years → seven years → surrender.

It would nonetheless be a mistake to look at one line from each case and declare the Board inconsistent. These cases are not necessarily comparable, and that qualification is not a hedge — it is the point. The summaries tell us enough to see enormous variation. They often do not tell us enough to explain it.

IX. A BAC number does not function as a simple penalty table

One tempting theory is that the Board follows something resembling a severity scale: higher blood alcohol, harsher discipline. The public summaries do not support such a rule.

The Board’s current publication includes a matter describing a blood-alcohol enhancement above 0.20% that ended in a public reprimand, alongside another blood-alcohol matter at the per-se threshold that resulted in seven years of probation. That does not prove inconsistent enforcement. It proves that the measured value alone cannot explain the published outcome.

There may be crucial differences in the underlying records — prior history, multiple incidents, evidence about rehabilitation or current fitness, admissions versus disputed allegations, differences in expert evaluation, facts bearing directly on patient safety, settlement considerations, procedural posture, or facts in a decision that never reach the register paragraph. All are plausible. None should simply be invented, and this project does not invent them.

That gap — between what may explain a result and what the public summary actually discloses — is exactly where serious regulatory analysis should begin.

X. Procedural route: why identical labels are not identical cases

A quarterly register prints the last line of a proceeding. It does not print how the proceeding got there. Reading the underlying documents recovers that middle: accusations, amended accusations, petitions to revoke probation, interim suspension orders, cease-practice orders, stipulations, default decisions, correcting orders. Once route is visible, outcome labels stop being interchangeable.

Route is coded as a first-class field on controlled values: stipulated_settlement, stipulated_surrender, default_decision, contested_hearing, adopted_proposed_decision, probation_enforcement, applicant_probation, reinstatement, other, unresolved.

Route carries its evidence state. Every route value is published with the source that established it, on three states: decision_derived (established from the primary decision, order or packet), register_derived (read from the register summary only), and unresolved. A route read from a register summary is not a route established from a document, and is never presented as one. Of 205 matters, 73 carry a decision-derived route, 11 a register-derived route, and the remaining 121 are unresolved.

Applicants and reinstatements are separate universes. A probationary licence issued to an applicant, and a conditional grant on a petition for reinstatement, are never merged into a physician-discipline figure anywhere in this project. The distinction is structural in the data model, not a caveat in prose.

What route does not license. Inferring arbitrariness, bias, favouritism, a causal effect of blood-alcohol level, quality of counsel, the reason a matter settled, impairment, addiction, or future dangerousness. Route describes part of the observed spread. It does not convert a published distribution into a finding about any physician or any pending case, and no causal claim on this page rests on it.

XI. Inside the physician DUI universe

The denominator here is 140 — physicians holding an active Medical Board licence whose published basis is a driving offence. Not the 205-matter corpus, and not the 154-matter combined DUI caseload.

Published basis. Alcohol-impaired driving in 125. Driving under the influence of a controlled substance in 4 — a small number, and deliberately so: a matter enters that subgroup only where the register’s own words name a drug or controlled substance as the impairing agent, never on the Board’s boilerplate “drugs and/or alcohol” formulation, which carries no substance-specific content at all. A further 3 matters carry a drug tag taken from a contributed compilation’s coded label rather than from register language, with no decision document behind it; those are held out of the subtype and shown separately, for a tagged total of 7. A wet-reckless or alcohol-related reckless disposition in 9. Categories overlap where an entry cites more than one.

Measured blood alcohol. 86 matters cite a measured blood alcohol concentration. Of those, 78 are published at the per-se threshold, 14 at 0.15% or above, and 6 at 0.20% or above. The tiers overlap where one entry cites more than one threshold, which is why the three do not sum to the total. The statutory enhancement thresholds are 0.15 percent or more and 0.20 percent or more under Vehicle Code section 23578; tags and prose use ≥ throughout, because a strict inequality would exclude the boundary cases the statute includes.

Bodily injury. 7 matters cite it.

Conviction language. 57 state a conviction in terms. 3 are published on arrest language alone. The remaining 80 are entries whose source records the offence in coded form without carrying the register’s conviction wording. That residual is a property of how the source was read. It is not a finding that no conviction occurred, and it is published rather than left for a reader to discover by subtraction.

Arrest without a stated conviction. Three matters — MATTISON, MASE and APPERSON — are published on arrest language alone, and each resolved in a public letter of reprimand. Three arrest-only matters, all landing on the same rung, is a small pattern rather than an anecdote, and it is the clearest single demonstration that the public disciplinary spectrum cannot be described as DUI conviction → probation. The Board has published more varied outcomes than that.

Probation length. Where a stay was granted, probation runs from 12 to 96 months across the 80 stayed matters that state a length: mean 52.3, median 59.5. The count is even, so the median falls between the 40th and 41st ordered values, which are 59 and 60 months. Both statistics are published because they are not interchangeable and the distribution is not symmetric.

XII. On reading the decisions

107 of the reconciled matters carry a link to a primary Board decision, order or packet. In 73 of those, the procedural route is established from the document itself. In 34, the document was opened and the route remains undeterminable from it. A further 11 matters carry a route read from the register summary rather than from a decision, and are counted separately for that reason. Default decisions, stipulated surrenders, interim suspension orders and correcting orders are visible in this dataset precisely because the underlying documents were opened.

What has not happened is a full codebook second read: an independent re-coding of every applicable field from the document text by a second reader. That number is zero, and it is published as zero under that exact label. Document located and field independently re-coded are two different facts, and this project keeps them apart.

The 22 cross-validated matters are resolved matches, not open questions: each is one matter that two independent reads found separately. Unresolved historical candidates are a different quantity and are counted at §XIII.

XIII. Completeness, stated period by period

Completeness is a separate dimension from verification, and it is asserted period by period — never for 2008–2026 as a whole. A period is marked bounded-complete only after the entire applicable Board register or publication interval has been reviewed for driving-related language, candidate matters have been reconciled against Board alerts, and the necessary underlying documents have been checked. “Partial / strong” never becomes “complete” by accumulation of records alone.

Unresolved historical candidates are recorded in the evidence-gap register rather than silently dropped or silently counted — 7 at present, each named with the reason it failed verification.

XIV. Known limitations

Stated in full, because a corpus that hides these is not a research instrument.

This is a reconciled record set, not a census. Verified matters currently begin in 2011. The 2008–2010 window sits inside the programme’s span and not inside its data.

2013–2016 is the largest reconciliation gap in the period already read.

Public letters of reprimand drop off the Board’s website after ten years, so the earliest part of the span is the likeliest to be under-represented — and public letters are precisely the mild end of the distribution this page argues is the most analytically interesting.

The register is an abbreviated instrument by design. It does not contain findings, mitigation, aggravation, credibility determinations, or reasoning. Fields derived from a register summary are labelled as such at row level and are never presented as decision-level facts.

Three matters were found in a single register page opened to confirm an unrelated entry, and had never been catalogued by any source consulted. That is a floor on what remains unfound, not a ceiling.

Five matters have no established effective date. They are counted in the corpus and excluded from every time series rather than assigned to a plausible year.

Older public disciplinary documents may disappear from the Board’s current web presence. Where a primary document link has gone dead, the row is marked archived rather than silently downgraded.

Probation enforcement arising from a driving offence sits outside the physician denominator. Where a physician’s driving matter was resolved through enforcement of an existing probation, the published entry describes the enforcement step rather than the original conduct, and the matter is counted in its own universe. Two matters are affected. Both are visible in the registry under probation_enforcement, and both appear in the combined DUI denominator of 154. Stating this matters because the excluded class sits at the severe end of the range; a reader is entitled to know the principal distribution is drawn without it.

Reinstatement petitions are not sanctions. Petitions for reinstatement or modification of penalty under Business and Professions Code section 2307 are conditional grants on an application. Two matters are held in their own universe rather than counted as stayed revocations, which is where a naive read of the disposition text would put them.

XV. The registry

Every reconciled matter, in semantic HTML, present in the initial document and readable without JavaScript. Sorted most recent effective date first, undated matters last. Generated from the canonical matter file under the generation stamp in the footer — not hand-maintained.

Column specification

How to read a row

Each row carries its own verification badges. REGISTER means the displayed fields are supported by an official Board publication. DECISION means the underlying document was opened. CROSS-VALIDATED means two independent reads found the same matter. A second-coded badge exists in the vocabulary and no row carries it.

The line under the analytical tags is the procedural route’s evidence state, at field level, because a route established from a decision and a route read from a register summary are not the same fact and a single row-level badge cannot express the difference.

On a narrow screen, swipe the table sideways — the licensee column stays pinned

Both exports carry the generation stamp, the universe, the verification state, the source layer, the source URL and the ambiguity flag on every row.

XVI. Methodology and data dictionary

Provenance layers

Matters are reconciled from independent provenance layers, each named and each inspectable underneath the canonical layer. No source observation is counted twice because a second researcher found it independently.

Unit of analysis

One reconciled disciplinary proceeding = one matter. A licensee may hold several matters; the licensee count is published separately for exactly that reason. Two licensees in this corpus hold two matters each.

Evidence states are separate stages, not one label

SOURCE_VERIFIED (source exists, basic observation matched) · DECISION_REVIEWED (primary document reviewed by the primary coder) · RECONCILIATION_REVIEWED (checked for cross-layer and same-layer duplication) · INDEPENDENTLY_SECOND_CODED (a separate coder reproduced the designated fields). A single “second read” label would conflate these, so none is used.

Publication thresholds by field class

Descriptive route reporting is permitted from an explicit official source or a first-read primary decision, with the evidence state displayed on the row. Independent second coding is required before route or any decision-level field is used for causal explanation, multivariable interpretation, or a claim that a given factor explains differences between otherwise comparable matters. No such claim appears on this page.

Exports

One export policy applies sitewide. Each generation ships as a versioned CSV (one row per canonical matter) and JSON (full structured matter objects), with a schema giving each field, its allowed values and its null rule. Methodology-bearing fields travel with each row — matter_id, universe, verification_state, source_layer, source_url, ambiguity_flag, reconciliation_note and route evidence state — so that a downloaded CSV cannot be stripped of its context by being downloaded.

XVII. Automated validation, run before every deployment

Manual editorial review is not sufficient at this size. The build fails on any of the following.

source_observation_total        == sum(all named source layers)
observations_as_matters        == source_observation_total − matched_cross_layer_duplicates
canonical_matter_count         == observations_as_matters − confirmed_same_layer_merges
sum(outcome_counts per universe) == that universe’s denominator
no duplicated matter_id
no classification tag repeated within one matter
route_value ∉ outcome_vocabulary
every drug-impaired tag names a drug in its source
  AND that source does not match /drugs?\s+and\/or\s+alcohol/
every row satisfying the DUI-primary definition carries the DUI-primary tag
every blood-alcohol tier field carries a field-level source
procedural_route publication permitted by its evidence_state
published_median derivable from its own distribution
published_conduct_verbatim contains no editorial gloss
every live page references the same dataset generation
superseded dataset pages self-identify as superseded
every source URL passes a link check or is marked archived

Two of these deserve a note. The DUI-primary test runs in both directions — no matter may carry a tag its source does not support, and no matter satisfying the definition may be left untagged. Running it in one direction only is how a dataset can simultaneously over-tag one subtype and under-tag the parent category. And the median assertion exists because a median that cannot be derived from its own distribution is not a median, which is a class of error no amount of proofreading reliably catches.

XVIII. Where the investigation goes next

The first phase of this project answers what happened. The next asks why, and that requires moving past register summaries into the underlying accusations, stipulated settlements, proposed decisions and final orders, and coding the variables that abbreviated registers omit: number and timing of alcohol-related events; prior disciplinary history; blood-alcohol level and enhancements; collision or bodily injury; patient-care nexus and practising while impaired; admissions versus contested allegations; rehabilitation evidence, evaluations and treatment; prior probation and probation violations; settlement versus adjudication; findings actually sustained, aggravating and mitigating; Board modifications to proposed decisions; and the terms the parties actually negotiated.

There is a further question the corpus has already surfaced and that the second read is designed to answer. The route vocabulary contains contested_hearing and adopted_proposed_decision, and no matter in the current registry carries either. If that holds after the full read, it is a finding in its own right, and a substantial one: it would mean this corpus is composed almost entirely of negotiated and default resolutions, in which the Board’s reasoning about a driving offence was never adversarially tested at all. A publication surface that carries little explanation may, in part, be reporting proceedings in which little explanation was ever generated.

Until those variables are read from the documents, this project will keep resisting the easiest and least defensible conclusion. Variation is visible. The cause of the variation is not yet proved.

XIX. The question the data leave on the table

A professional licensing system is not merely supposed to impose discipline. It also performs an explanatory function. Physicians should be able to understand what conduct creates what regulatory risk. Lawyers should be able to advise clients from something more informative than scattered settlements. Patients should be able to distinguish a monitored physician from one whose licence was actually revoked. And the Board itself benefits when its published record makes its decision structure intelligible.

If two public summaries sound very similar but one produced a reprimand and the other seven years of probation, a reasonable reader should not have to guess what important fact is missing.

That is now a research question rather than a rhetorical conclusion, and it is the question the decision-level second read exists to test.

California does not have one publicly visible outcome for a physician driving offence. It has an extraordinary disciplinary range. What its abbreviated public record does not yet supply is an equally clear account of how a matter moves from one end of that range to the other.

Notes

Scope. Public-record research for educational and policy analysis. An accusation is not a finding. A published historical distribution is not a prediction about any physician or any pending case. Nothing here is legal or medical advice.

Method. The Observatory distinguishes a Medical Board publication summary from the underlying disciplinary decision. Register-supported fields describe what the Board published. Decision-level conclusions about aggravation, mitigation, adjudicative reasoning and causal explanation require examination of the underlying decision.

Read next: Observatory methodology · Osteopathic Board dataset · How Board panels review proposed decisions · Administrative hearings · Standards of proof · Why enforcement data need context · A complaint is not a finding · Medical Board newsletter archive