Record of Care · Episode 3 of 7 · Medical fiction

The Setting

The test encounter is complete. The patient’s next step still has no owner.

  • Medical fiction · Episode 3 of 7
  • With notes: The People Who Configure the Practice of Medicine

Published · revised as medical fiction

7 min story · 15 min notes · sources in the notes checked through (Pacific)

Conceptual illustration: a settings panel with switches, beside a completed step whose dotted line leads to an empty circle.
Conceptual illustration. No real patient, clinician, screen or record is depicted.

The test environment had a patient named Zzztest Marigold, who was ninety-four years old in the system and had been referred to every specialty in the building.

“Nobody use a real chart,” Inez said. “I mean it. If I see a real name on that wall I’m unplugging the projector.”

It was seven-thirty on Wednesday morning and the conference room smelled of the coffee Celia Ward had carried in herself, two cardboard carriers, because she had called the meeting and it was early. Celia ran operations for both Calder Bay sites. She had the waiting list on her tablet, as she generally did, and she had not come to be told that the referral workflow was broken.

“I want to say why it’s built the way it’s built,” she said, before Inez could start. “Because I was in the room.”

“Please,” said Julian.

“A year ago the referral desk had six hundred open tasks. Half were finished and nobody’d closed them. Tomas was spending the first two hours of every day working out which ones were real. Clinicians had notes they couldn’t close because of an open item three screens away. So when the spring update gave us the choice to close the task automatically once the referral transmits, we took it. Open tasks dropped by two-thirds. Tomas got his mornings back.” She looked down the table at him. “True?”

“True,” Tomas said. “I asked for it.”

“So I’d like us to remember it solved something.”

“Nobody’s saying it didn’t,” Lena said, and tried to mean it.

Inez dimmed the lights.


They ran the visit the way Lena had run the real one. Inez played the clinician. She opened a note for Zzztest Marigold, let the drafting tool produce its paragraph from a script she read aloud in a flat voice, chose rheumatology from the referral list and signed.

On the wall, the destination had filled itself in.

“Seaview,” said Julian.

“It’s the first on the preference list,” Inez said. “The list wasn’t changed when the contract was. That’s a five-minute fix and it’s mine. I’ll do it this morning in production, with a ticket.” She wrote it on the whiteboard under her own name. “That’s not the interesting part.”

She switched to Tomas’s view. The referral desk’s queue appeared, long and gray. Near the top a line for Zzztest Marigold showed a transmission time and a small green mark.

“That’s how it gets to me,” Tomas said. “Already green. I don’t open the green ones. That’s the point of green.”

“And when Seaview faxes it back?”

“Fax goes to the shared folder. If the task’s open, I attach the fax and work it. If the task’s closed, there’s nothing to attach it to.” He shrugged, not carelessly. “Last Thursday I had a rejection in my hand and no place to put it. That’s why I wrote the report. I wasn’t reporting Dr. Rowan. I was reporting that I was holding a piece of paper.”

Lena had not known until then how the report had read from his side. She found she was grateful, and a little ashamed of having wondered.

“Show the patient’s view,” Julian said.

Inez opened the test portal. Zzztest Marigold’s after-visit summary said that a referral had been made and that the specialist’s office would be in touch.

“Would they?” Celia asked.

“Seaview? No.”

“Then who tells her?”

Nobody answered, and after a moment Celia saw that the silence was the answer. She set her tablet down on the table.

“Before the update,” she said. “What happened before?”

Inez had the old configuration notes open on her laptop. She had gone looking for them the night before, and it had taken her longer than she liked.

“Before, the task had a second state. Sent, awaiting acceptance. It stayed in Tomas’s queue until the receiving office confirmed, or ten days went by, and then it turned amber and somebody called. When we turned on auto-close, that state went away. It wasn’t a separate decision. It came with the setting.”

“Did we know that?”

“It’s in the release notes.” Inez turned the laptop so Celia could see the line. “Page nine.”

“Who read page nine?”

“I did,” Inez said. “I’m not sure I understood what it meant for a person. I understood what it meant for the queue.”

It was a harder thing to say than fixing a preference list, and the room let her say it.

Julian had been writing. “Who approved the change?”

“I did,” Celia said. “With IT. Tomas asked for it, the vendor demonstrated it, I signed the form.”

“Was a clinician in the demonstration?”

Celia opened her mouth and closed it.

“Dr. Amari was invited,” she said. “The medical director. He had clinic. The demo ended when the referral transmitted. It turned green and everyone was pleased.” She rubbed the bridge of her nose. “It was a good demo.”

“I’d have been pleased too,” Lena said. This time she did mean it. “I didn’t know there’d been an amber. I only knew that at some point I’d stopped getting calls from Tomas, and I thought that meant things were going well.”

“They were going quietly,” Tomas said.


They argued for twenty minutes about what to do, and it was a real argument.

Lena wanted auto-close turned off that day. Celia would not have it. Six hundred open tasks was not a number she intended to see again. It was not the vendor’s number either. It was Tomas’s mornings, and the forty-one clinicians who would start leaving notes unsigned by the following week. Julian wanted to know whether the choice was truly between the two, and Inez said it was not. The tool allowed a task to close on transmission and a separate, lighter item to open, a confirm-acceptance check with a due date that would belong to the referral desk and nobody else.

“Why didn’t we do that in the spring?”

“Nobody asked for it. You have to know to ask.”

“Can you build it?”

“In test, today. In production, after it goes through change review, which is Tuesdays.” She saw Lena’s face. “I’m not skipping change review. Skipping it is how we got page nine.”

“And between now and Tuesday?”

That was Celia. She had picked the tablet back up, but she was not looking at the waiting list.

“Between now and Tuesday,” Tomas said, “I can pull every referral that auto-closed since the spring and went to a practice on the changed-contract list. It’s a report. I know how to run it.” He paused. “I don’t know how many there’ll be.”

Julian said what the others were thinking. “It won’t only be Ms. Ellison.”

“No.”

“Then that’s the fallback,” Celia said. “Written down, with a date on it. Tomas runs the report today. Every one of those patients gets a call from the desk, not a letter. Inez builds the check in test and brings it to change review Tuesday with a clinician’s name on the form. I’d like it to be yours, Lena, since you seem to have opinions.”

“I do.”

“And I’ll sign for the limit we’re accepting in the meantime, which is that for six days the system will keep turning these green and a human being is checking behind it by hand.” She wrote her own name on the whiteboard, under Inez’s. She had been handed plenty of other people’s problems. This one she intended to return in better condition. “I approved the first one without knowing what I was approving. I’d like to know this time.”

Inez photographed the board.

“One page,” she said. “What changed, which version, who looked at it, what we tested, what we’re living with, who to call. I’ll write it up. I should have had one in the spring.”


The room emptied the way such rooms do, quickly and in the wrong order. Julian left first, writing as he walked. Celia took the coffee carriers. Inez stayed to shut down the projector, and Zzztest Marigold went dark on the wall with all her referrals intact.

Lena found Tomas at the door. He had his list on a legal pad: the report, the calls, the county clinic for Ms. Ellison, where he had left a message the afternoon before and where nobody had yet called him back.

“Thank you for filing it,” she said. “The report. I read it as being about me.”

“It had your name on it. I’d have read it that way too.” He folded the pad under his arm. He was a courteous man, and he said the next thing gently. “Can I ask you something, doctor?”

“Of course.”

“This check they’re giving me. Every referral, every day, did somebody on the other end say yes. I think it’s right. I’d have wanted it before.” He nodded toward the whiteboard, where his name now appeared three times. “I had my mornings back for about six months. I’m not complaining. But I’m one person, and the queue’s the same queue it was.”

He looked at her as if she might know.

“When am I supposed to do it?”

The questions behind this episode

Clinical and governance notesThe People Who Configure the Practice of MedicineA physician’s judgment depends partly on decisions made before the patient enters the room. Clinical AI makes those decisions harder to keep invisible.

Everything below is commentary on real clinical, technical and legal questions. It refers to the story only as an illustration. Characters named here are fictional.

Evidence at a glance

Established by the cited sources
HRSA assigns Health Center Program boards substantial authority over services, budgets, personnel policy and quality; Medical Board of California guidance describes corporate-practice concerns, including control over treatment, record contents and patient volume; California’s AB 1979, AB 1637 and SB 503 take effect January 1, 2027; FDA’s January 2026 clinical decision support guidance addresses function-specific nondevice criteria.
Not established, and not claimed
That any real clinic configured software unlawfully, excluded physicians from implementation or engaged in the corporate practice of medicine. An administrative reporting chain or an EHR purchase is not, by itself, a violation.
The analysis in these notes
A configuration that changes what a clinician sees, accepts or sends is a clinical decision. Meaningful participation means input that can change the decision, a working route to disagree and a versioned record of what was approved.

In the episode a test encounter is followed past the point where demonstrations usually stop: into the referral desk’s queue and the instructions a patient would receive. The software does what it was configured to do. One step is marked complete, and no person has accepted responsibility for the next one. The original decision solved a real problem and was made in good faith. It still needed a second look once its consequences were visible.

The notes below ask who should take part in decisions of that kind, what legitimate management authority includes, what the law does and does not say, and how an organization can record a configuration choice so that it can be found again.

The consultation starts before the consultation

When we imagine medical judgment, we usually picture a clinician and a patient. One describes a problem. The other listens, examines, weighs possibilities and recommends a course of action.

Much of the encounter has already been arranged by then.

Someone decided how much time appears on the schedule. Someone determined which information will be visible, which reminders will interrupt, which fields must be completed and which choices require additional explanation. Someone chose who can edit a note, what a signature releases and where an unfinished task goes.

These decisions can be necessary and sensible. They are also capable of changing what the clinician can practically accomplish.

In Rowan’s clinic, the AI assistant was introduced to reduce the burden of writing. Its later configuration also influences how a plan is presented and how work is handed to other staff. That is a different operational proposition, even if the product keeps its name.

The most consequential question is therefore not whether a physician was invited to a demonstration. It is whether clinicians had a meaningful role in deciding the conditions under which the system would shape care, and whether they retain a route to change those conditions when experience exposes a problem.

An invitation to training answers how to use the chosen system. Participation in governance asks whether the chosen system should work that way.

The manager has a patient, too

The clinic’s operations director sees a different part of the same problem.

People are waiting for appointments. Staff repeat work across systems. Clinicians finish documentation after the building closes. The organization cannot purchase every requested tool or support a different electronic workflow for every preference.

A functioning clinic needs coordination. It needs reliable budgets, secure access, consistent handoffs and enough capacity to serve the next patient. Standardization can support those aims. A physician’s objection does not prove that a proposed change is unsafe, and the word autonomy cannot settle every disagreement over resources.

The director’s strongest question is fair: if the organization cannot establish a common process, how does it deliver dependable care across multiple locations?

Rowan’s strongest response should be equally practical: the common process must make the necessary clinical distinctions possible.

For a health center participating in the federal Health Center Program, governance is not simply a contest between management and individual physicians. HRSA assigns the governing board substantial responsibilities for services, budgets, personnel policy and quality oversight. Those powers must coexist with compliance with applicable law. They cannot be erased by a proposal that gives one clinician control over every organizational decision.[1]

The aim is an arrangement in which each decision reaches the people qualified and authorized to make it. The budget belongs in an accountable governance process. A patient’s clinical assessment belongs with the appropriately licensed professional. A change that connects the two requires both kinds of expertise.

A setting can carry a clinical decision

Consider what appears to be a simple configuration choice: should a generated recommendation be displayed as an option, inserted into a draft, or sent automatically to the patient?

The content may be identical. Its status is not.

An option invites evaluation. A draft invites correction. A released instruction can tell a patient what to do. Moving between those states changes the responsibilities of the people who encounter the information.

Another setting controls whether a referral can remain visibly unresolved after a note is signed. Someone might reasonably want to reduce the number of open items. Yet closing a documentation task and completing the clinical follow-up are different achievements.

In the fictional test, the green indicator records the first. The patient needs the second.

This is why an information-technology decision can deserve clinical review without every person who works on the software thereby practicing medicine. Engineers have essential authority over technical design. Administrators have legitimate responsibilities for operations. The governance question is whether a particular choice determines clinical content or constrains professional judgment, and what qualified review it receives.

California’s Medical Board guidance on corporate practice identifies several forms of control that can intrude on physician judgment, including decisions about treatment, medical-record contents and patient volume. It also recognizes consultation with people who are not physicians while preserving physician responsibility or approval for the decisions it describes.[2]

That guidance should sharpen the inquiry. It does not establish that buying an EHR, employing an implementation team or disagreeing with a physician about a configuration is automatically unlawful.

Entity structure matters. California recognizes exceptions permitting physician employment in settings including qualifying nonprofit community clinics. A federally qualified designation does not, by itself, tell the reader everything about an organization’s state-law structure. A responsible analysis identifies the applicable structure and the actual conduct before announcing a corporate-practice violation.[3]

The broader governance lesson remains available without making that allegation: when a configuration affects care, clinical participation should be built into the decision that authorizes it.

Ask while the answer can still change

In a better version of Rowan’s clinic, the first substantive meeting occurs before the implementation plan becomes a completed fact.

The operations director brings the problem to be solved: unfinished documentation, duplicated work and unreliable follow-up. The technical team explains what the proposed configuration can do. Practicing clinicians describe what a safe encounter requires. Staff who receive referrals explain what happens after the physician presses the button.

Each participant can now discover something useful.

The physician may learn that a preferred shortcut leaves another team without essential information. The administrator may discover that a supposedly redundant field carries a distinction the receiving clinician needs. The engineer may find that the requested behavior can be achieved through a different setting with fewer interruptions.

That is meaningful input because it can alter the decision.

It also requires representative experience. A clinician who enjoys the demonstration may not work with the same patient population, language needs, staffing or connectivity as colleagues at another location. A useful review includes people who encounter the exceptions as part of ordinary work.

The federal SAFER guidance recommends practicing-clinician involvement in EHR safety decisions, analysis of administrative and clinical workflows, and continuing collaboration between organizations and vendors. These are safety recommendations, not a declaration that every suggested process is independently required by law.[4]

The story’s test is modest. Follow an invented encounter all the way through the clinic, including the places the demonstration usually ends before reaching: the nurse’s queue, the referral team’s inbox and the instructions a patient would receive.

The test reveals a missing handoff. It also shows that the proposed assistant makes several steps easier. The team can preserve those benefits while redesigning the troublesome transition.

This is a stronger result than either automatic approval or automatic rejection. The clinical objection has become a testable design question, and the operational problem still has an owner.

Who belongs in the decision

Meaningful participation is not measured by the number of people invited. It is measured by whether the people who understand each consequence of a change can affect it before it reaches patients. One possible composition for a consequential clinical-AI configuration decision:

ParticipantWhat they contributeThe question only they may think to ask
Practicing clinicians from more than one siteKnowledge of the encounter as it actually unfolds“Where would I correct this, and does the correction persist?”
Clinical leadership with decision authorityAccountability for the clinical standard the change must meet“Who answers for this if it behaves differently in practice?”
Nursing and medical assistant staffRooming, triage, messages and the work that follows the visit“Who receives the task this button creates?”
Referral and scheduling staffThe pathway after the clinician’s order“What happens when the receiving practice declines?”
Interpreters and language-access staffHow meaning survives translation“Does the summary reflect what the patient actually agreed to?”
Health information managementRecord integrity, amendments and release of information“Where else will this statement travel?”
Informatics and ITConfiguration, integration and version control“Which release and local setting are we actually testing?”
Compliance and privacyApplicable law, consent and retention“What is recorded, kept and disclosed, and for how long?”
Operations and financeCapacity, budget and staffing“What does this change cost, and what work does it move?”
A patient representativeThe experience of receiving the result“Will someone without portal access understand what to do next?”

Not every change requires every participant. A cosmetic adjustment does not need a patient representative; a new automatic patient message probably does. Composition should follow the potential consequence, and the record of the decision should show who actually took part.

The table is a proposal, not a description of any statute or accreditation standard. Its purpose is modest: to give the person who will discover a problem in the third week of use a chance to describe it at the first meeting.

The right to disagree needs a working route

An organization can write that the clinician retains final judgment while making disagreement unusually difficult.

A suggested plan may be easy to accept and cumbersome to revise. A rejected item may return after the draft is regenerated. A physician may be able to change the narrative while lacking permission to correct a linked field. These are possibilities to examine in the actual system, not assumptions about every product.

The practical test is what happens when the clinician reaches a different, defensible conclusion.

In Rowan’s proposed workflow, the physician can keep a software suggestion separate from the adopted plan. A correction remains recognizable. A patient-specific decision reaches the staff who must act on it. If a feature cannot support that work reliably, an approved alternative remains available while the problem is investigated.

That alternative needs boundaries. It should protect continuity, preserve necessary documentation and tell colleagues what to do next. It should not become an undocumented private system that leaves the rest of the clinic guessing.

The organization should also distinguish clinical disagreement from refusal to perform essential work. A physician remains answerable for appropriate assessment, accurate records, communication and follow-up. A safety concern deserves an evidence-based response; it does not confer immunity from legitimate review.

For the same reason, adoption statistics are an incomplete measure of success. A high acceptance rate might reflect useful recommendations. It might also reflect interface design or pressure to finish. A rejection rate cannot tell a reviewer, by itself, whether a clinician is thoughtful, mistaken or struggling with irrelevant output.

The useful unit of inquiry is the decision in context: what information was available, what the software proposed, what the clinician concluded and what happened to the patient’s care.

California’s new laws reach different parts of the problem

California enacted several relevant measures in 2026. Their ordinary effective date is January 1, 2027. Their enactment should not be used to declare earlier conduct unlawful under provisions that had not yet taken effect.[5]

AB 1979 requires covered health facilities, clinics and physician offices to take reasonable steps to preserve a licensed provider’s independent judgment when clinical decision support informs care. It restricts specified uses of AI for clinical functions requiring licensure, while preserving an exception for documentation and communication that do not involve professional judgment.[6]

AB 1637 addresses physician patient-note responsibility and who may alter those notes, with specified exceptions for delegated authority, transferred care and supervision of trainees. It does not convert the physician’s signature into proof that the physician personally originated every sentence.[7]

SB 503 assigns developers and deployers responsibilities concerning foreseeable biased impacts. Among its provisions are developer documentation associated with initial sale, requests and applicable material updates, and continuing deployer monitoring and proportionate mitigation.[8]

These are distinct obligations. None should be reduced to a slogan that the physician owns every risk. Nor does their existence establish a violation by a particular organization or product. The operational question is how an organization will demonstrate the relevant duties in the way it actually works.

Federal device classification asks another question. FDA’s current clinical decision support guidance addresses statutory criteria for excluding certain software functions from the device definition. One criterion concerns the professional’s ability to independently review the basis of a recommendation. The analysis depends on the function and intended use; it is not settled by calling a system a scribe.[9]

Failing one exclusion criterion does not automatically make a function subject to full device regulation. FDA expressly recognizes other nondevice categories and policies of enforcement discretion.[10]

A clinic should obtain a sound regulatory explanation for the functions it uses. That explanation still does not answer whether its local implementation gives Rowan the information, time and practical authority needed for patient care.

The update is another decision

Approval at purchase cannot settle every later change.

An assistant may gain a new function. An integration may begin placing output in another part of the record. A local administrator may change a default. The vendor may repair one defect while altering a behavior on which the clinic’s staff had come to rely.

The relevant question is what changed in the work.

Rowan’s clinic develops a short, versioned decision record for consequential changes. It identifies the behavior being changed, the reason, the clinical and operational reviewers, the test performed, the limits accepted and the person authorized to respond if the result differs from expectations. The record refers to the release and local configuration actually evaluated.

This is a proposed governance practice. Its purpose is to make responsibility recoverable without turning every minor interface adjustment into a committee proceeding.

A spelling correction and a new automatic patient-message function do not need the same review. Neither do a cosmetic rearrangement and a change that removes an unresolved follow-up item from view. The review should follow the potential clinical consequence.

When a concern arises, the record also prevents a familiar confusion. The physician remembers a previously tested behavior. The technician describes the latest release. The manager cites the original approval. All may be accurately describing different versions.

A versioned record gives them a common object to inspect.

It can also protect a useful innovation. If the problem concerns one feature or one local setting, the organization may be able to limit that function while preserving the rest of the tool. A precise account of the change supports a proportionate response.

The clinic should not confuse a successful demonstration with proof of sustained performance. A useful next step is a defined period of observation in ordinary, authorized use: do the expected people receive the tasks, do corrections persist, and can staff identify work that still needs attention? The evaluation should examine whether the improvement creates additional work somewhere else.

Nor should a quiet incident inbox end the inquiry. Staff may have found a workaround, misunderstood the new behavior or concluded that reporting will take longer than repairing the problem themselves. The organization can ask the people doing the work to show how they now complete an encounter. A discrepancy between the approved process and routine practice is information to investigate. It is not, by itself, proof of either misconduct or software failure.

A change record on one page

The versioned decision record described above need not be elaborate. One page can preserve what later disputes most often require.

FieldWhat it records
The changeThe behavior being added, altered or removed, described in clinical terms
Version and configurationThe vendor release and the local settings actually evaluated
PurposeThe clinical or operational problem the change is meant to solve
ReviewersClinical and operational reviewers by role, and who held authority to decide
Test performedThe scenario used, including downstream queues, patient instructions and structured fields
Known limitsBehaviors accepted with awareness, and why
FallbackThe approved alternative if the feature fails in practice
Observation periodWhat will be measured in ordinary use, by whom and for how long
Response ownerWho acts when the result differs from expectation
Reconsideration triggerThe evidence that would reopen the decision

A record of this kind does three things at once. It tells the clinician what was decided and why. It tells the technical team what behavior it is responsible for maintaining. And it tells a later reviewer, examining a disputed chart, a patient complaint or a quality finding, which version of the system was in use and what the organization knew about it at the time.

That last function matters for fairness. A physician’s encounter should be evaluated against the system as it was configured on the day of care, not against a later release, an earlier demonstration or an administrator’s recollection of what the tool was supposed to do.

Consultation must leave a trace

Meaningful physician input needs a visible outcome.

That outcome may be acceptance of the recommendation, a modification, a request for further testing or a reasoned decision to proceed despite the objection. Agreement is not the only evidence that consultation occurred. A concern disappearing into an inbox is poor evidence of a functioning process.

The clinic should be able to explain who considered the clinical issue, what evidence was reviewed, who had authority to decide and how the decision will be reconsidered if the expected behavior does not hold.

Participation also needs time. An organization cannot sensibly treat careful testing as important while requiring the participating clinicians to absorb all of it between full clinical sessions. If the work is necessary to deployment, it belongs in the deployment’s resource plan.

The same applies to staff outside the physician group. A referral coordinator may spot the missing recipient. An interpreter may recognize that a generated instruction obscures what the patient actually agreed to. A nurse may know that a seemingly complete task still requires clinical triage.

Patient involvement can expose a different blind spot. In health centers subject to the ordinary federal governance requirements, patient-majority boards already express a principle of accountability to the people served, with defined exceptions and waivers.[11] A patient representative does not need to select a model architecture to explain why an instruction that assumes portal access can fail.

Clinical participation should make those observations actionable. It should not become a claim that one professional group understands every part of the system.

What the patient receives

Return to the fictional clinic after the workflow has been revised.

Rowan’s patient leaves with instructions that explain what has been decided and what remains to be arranged. The referral team receives an identifiable task. Someone is responsible for checking whether the next step occurs. If the usual route fails, the staff know where to escalate it.

The patient never sees the configuration meeting. The patient does not know which option was changed, which recommendation was rejected or which person had authority to approve the revised workflow.

But the patient receives the consequences.

The operations director benefits as well. The organization has a common process that can be explained, taught and examined. The technical team has a defined behavior to maintain. The physician’s review has a clear purpose. The clinic can ask whether the change improved the care process, rather than treating a green screen as the answer.

There will still be errors, disagreement and difficult tradeoffs. Governance cannot promise otherwise. It can make the organization capable of recognizing a problem, locating the relevant decision and correcting it without pretending that responsibility begins and ends with the last person to sign.

Clinical AI draws attention because it can produce language that resembles medical reasoning. The quieter influence belongs to the people who decide where that language appears, when it becomes actionable and how another person can challenge it.

Those people help configure the practice of medicine.

They should be able to explain their choices to the clinician who must use them—and to the patient who must live with the result.

Sources and scope

Sources checked through 19 November 2026 (Pacific). Each source establishes only what its own scope supports; vendor documentation describes product capabilities, not any organization’s configuration. Source numbering restarts in each essay of the series.

  1. HRSA, Health Center Program Compliance Manual, Chapter 19: Board Authority. Requirements and exceptions apply to the relevant Health Center Program governance structure.bphc.hrsa.gov
  2. Medical Board of California, Information Pertaining to the Practice of Medicine, Corporate Practice of Medicine section. Board guidance; application depends on actual structure and conduct.www.mbc.ca.gov
  3. California Research Bureau, The Corporate Practice of Medicine in a Changing Healthcare Environment, April 2016, pp. 4–7, including the nonprofit community-clinic category under HSC 1204(a). Historical government analysis used only for the existence of differentiated entity structures, not as an exhaustive statement of 2026 law. Current CDPH primary-care-clinic guidance also describes the nonprofit licensed-clinic framework.sbp.senate.ca.gov www.cdph.ca.gov
  4. Federal 2025 SAFER Guide, Organizational Responsibilities, especially practicing-clinician participation, workflow analysis and shared safety responsibilities. These are recommendations, not legal-compliance guarantees.www.healthit.gov
  5. California 2026 Legislative Calendar, compiled by the Assembly Chief Clerk and Secretary of the Senate, revised 27 March 2026, p. 2; regular-session effective-date rule under California Constitution, Article IV, section 8(c). The three chaptered texts cited below have no separate urgency or delayed-operation provision.www.assembly.ca.gov
  6. AB 1979, Chapter 854, Statutes 2026, BPC 22758.5; chaptered legislative text reproduced by LegiScan. Official MBC staff analysis corroborates the chapter and provisions; staff analysis is not a subsequent Board ruling.legiscan.com www.mbc.ca.gov
  7. AB 1637, Chapter 536, Statutes 2026, BPC 2266.1; chaptered legislative text reproduced by LegiScan.legiscan.com www.mbc.ca.gov
  8. SB 503, Chapter 857, Statutes 2026, BPC 22758; final chaptered legislative text reproduced by LegiScan. Earlier proposed audit and reporting requirements are not substituted for the final text.legiscan.com
  9. FDA, Clinical Decision Support Software, final guidance issued 29 January 2026, superseding 6 January 2026.www.fda.gov
  10. FDA, Clinical Decision Support Software Frequently Asked Questions, especially Q4.www.fda.gov
  11. HRSA, Health Center Program Compliance Manual, Chapter 20: Board Composition.bphc.hrsa.gov

Related reporting and resources on this site

About this page. This is Episode 3 of Record of Care, a fictional medical drama in seven episodes, followed by its clinical and governance notes. The story is fiction: its characters, clinic, documents and events are constructed, and nothing in it is evidence about a real person, organization, product or deployment. The notes cite real sources with their limits and label their proposals as proposals. First published 19 November 2026 as part of the essay series Clinical AI, Physician Autonomy, and the Record of Care; revised as a serialized work of medical fiction on 19 November 2026. The earlier edition is retained in the site’s editorial archive. Drafted with AI assistance and published on the author’s instruction. Sources were carried over from the first edition, checked through 19 November 2026 (Pacific). General information, not individualized medical or legal advice. Corrections: editorial standards · methodology.

Medical fiction with clinical and governance notes. First published 19 November 2026; revised as a serialized work of medical fiction on 19 November 2026, on the author’s instruction. Sources in the notes checked through 19 November 2026 (Pacific). Corrections: editorial standards.

Search KPSGill.com

Search the full publication — patient guides, policy, investigations, enforcement records and research.