Latest / Doctor AI: Rebuilding Trust in American Healthcare / The Standard of Care | Doctor AI, Part One

Transcript

The Standard of Care | Doctor AI, Part One · Doctor AI: Rebuilding Trust in American Healthcare

• 0:01
  The Standard of Care, part one of the Dr.
• 0:03
  AI series.
• 0:05
  I'm Robin Blackstone, and this is Dr.
• 0:07
  AI.
• 0:07
  A Tumor does not recognize geography.
• 0:13
  The care available to a person can depend on their community, hospital system, and state.
• 0:19
  We speak in medicine of the standard of care as though it were one settled body of medical knowledge.
• The treatment medicine has determined a person should receive, but it's not that simple.
• The standard of care is partly medical and partly legal.
• In a malpractice case, for instance, the question is generally whether a reasonably competent medical professional with similar training would have acted in the same way under similar circumstances.
• But who counts as the comparable professional? Depending on the jurisdiction and the clinical situation, the comparison may be with professionals in the same community, a similar community, the state, or a national specialty.
• Some states still retain versions of the locality rule That rule comes from a time when a rural physician could not reasonably be expected to have the knowledge or resources available at a major urban medical center.
• But that world is long gone.
• Medical knowledge can travel instantly.
• 1:30
  Expertise, technology, and treatment methodologies still do not.
• The best care supported by medical evidence and the care ordinarily provided in a particular place are often and not necessarily the same thing.
• Even a national employer health plan does not create a national standard of care.
• 1:54
  The insurance card may be the same.
• The specialists, hospital capabilities, provider networks, and legally available services are not.
• Coverage may travel, but care remains local.
• Actual care is shaped by much more than medical knowledge.
• It is shaped by the specialists employed by the hospital, the technologies it has purchased, its internal protocols, the human principal's health insurance coverage, local referral patterns, transportation, state law, and the institution's willingness to assume financial or legal risk.
• Move the human principal from one community to another, and the options can change.
• Cross a state line, and some services may disappear entirely.
• Enter a nationally recognized medical center, and a diagnosis may be divided according to molecular characteristics that were never investigated at a hospital a few hours away.
• The disease has not changed.
• The care environment has.
• This is especially visible in cancer.
• Consider a Black woman with breast cancer.
• Calling the disease breast cancer does not tell us enough to determine the right treatment.
• 3:19
  The tumor's receptors, genomic characteristics, stage, and response to therapy all matter.
• So does the human principal's family history, prior health, environment, culture, resources, and preferences.
• Triple-negative breast cancer is more common and more frequently diagnosed at younger ages in Black women.
• These tumors lack the receptors targeted by several common breast cancer therapies.
• The ordinary treatment pathway used for other breast cancers may not and probably will not work There is another warning here.
• A widely used genomic test for some early-stage breast cancers was found to be less accurate in predicting outcomes for Black women.
• The test had been developed using relatively little information from Black women.
• The protocol then can therefore be followed correctly and still and still fail the person in front of it That's not an argument against guidelines.
• Guidelines preserve knowledge.
• They improve consistency.
• They prevent avoidable mistakes.
• It is an argument against confusing the ordinary pathway with the right pathway.
• A person whose biology falls outside the population in which a protocol or predictive tool was built may require additional surveillance, specialist review, or a different treatment plan.
• If the local system does not routinely look for those differences, apparently standardized care can reproduce a dangerous blind spot.
• Poor standardization reproduces poor care at scale.
• Maternal and fetal medicine makes the graphic and geographic problem even more explicit.
• A woman with an ectopic pregnancy, preterm rupture of membranes, a severe fetal abnormality, or a dangerous maternal condition may face a situation in which the medically indicated course is clear, but the legally promoted course is not.
• Physicians working under state abortion bans have described treatment delays, additional testing, institutional approvals, and uncertainty about how ill a person must become before intervention can be considered legally defensible.
• The medical facts do not change at the state border, but the permissible response does.
• Abortion is the most visible example, but it is not the only one.
• Political decisions also shape vaccination policy, reproductive technology, gender-related care, public health reporting, prescribing rules, and access to certain therapies.
• Hospital policies and insurance decisions create additional boundaries and barriers within states.
• Political leaders have the authority to make law.
• They do not have the authority to convert a legal restriction into a medical conclusion.
• When a treatment is prohibited, unavailable, or not covered, the human principle should be told exactly that.
• We cannot provide this here is materially different from medicine has nothing to offer.
• Now AI enters this environment.
• A capable medical AI can examine national and international evidence.
• It can compare guidelines, identify uncommon patterns, review drug interactions, and recognize when the expected pathway may not fit the individual But the human principle still lives inside a local system with its own resources, laws, financial incentives, and historical habits.
• Which standard should AI follow? The care ordinarily delivered nearby, the care permitted by one state's laws, the care covered by the person's insurer, or the best care supported by the evidence, regardless of those other factors? Dr.
• AI will have to distinguish among these.
• It should be able to say, "This is what the evidence supports.
• This is how the evidence applies or may not apply to you, the individual."
• This is what is medically available locally.
• This is what the law permits here, and this is where the appropriate expertise or treatment can be found.
• A legal restriction cannot be allowed to masquerade as a medical conclusion.
• A local absence of expertise should not be treated as proof that better care does not exist, and a population average cannot automatically become the answer for a person whose biology, environment, or priorities place them outside that average.
• This does not mean accepting every AI recommendation as correct.
• Medical AI can carry bias, it can miss context, it can rely on incomplete evidence and communicate uncertainty poorly.
• Its performance must be evaluated, monitored, and challenged.
• If the system is using general data rather than the precise data of the individual, it can miss the differences that determine the outcome.
• But capable AI should be compared with the system people actually experience, not an imaginary system in which everyone already has continuous access to the world's best physicians and the most current evidence.
• That is so not the system we have.
• The next question will move beyond whether medical professionals are permitted to use AI.
• It will become whether there are circumstances in which they should be required to use it.
• The JAMA Summit report on artificial intelligence raised this directly.
• As the benefits of particular AI systems become established, failure to use them may become unethical or fall below the standard of care Ezekiel Emanuel, Abe Baker Butler, Neil Khosla, and Vinod Khosla have pushed the argument further.
• They argue that generative AI already rivals or exceeds licensed physicians in several central cognitive tasks, including gathering information, developing a differential diagnosis, selecting diagnostic tests, recommending guideline-based treatment, and managing chronic disease.
• Their conclusions will be debated, as they should.
• But the threshold question has changed.
• We can no longer assume that human care is the standard against which AI must always be measured.
• In some settings, and perhaps in most of them, AI may become part of the standard against which human care is measured.
• If an available system can reliably detect a dangerous drug interaction, identify an unusual cancer subtype, recognize a significant change across years of medical records, or locate the right specialist, and no one uses it, what constitutes reasonable care? Much of the healthcare industry is applying AI first to documentation, billing, scheduling, and administrative workflow.
• Those uses may reduce clerical work.
• They may also preserve every important limitation of the current system while adding another layer of technology and expense.
• Making the present machinery run faster is not the same as improving health.
• Dr.
• AI's promise is not merely to standardize care.
• Its deeper promise is to recognize when the standard pathway is wrong for the person to whom it is accountable.
• That accountability must be designed into the system.
• Dr.
• AI cannot serve the financial interests of an insurer, hospital, employer, or technology company ahead of the human principle.
• Its conflicts, limitations, and uncertainties must be visible.
• Its first loyalty must be to the individual it serves.
• That individual is the human principle, the person whose life, values, and decisions govern the process.
• The medical professional becomes the human health ally.
• Freed from typing, searching fragmented records, and satisfying the computer, the human health ally can concentrate on interpretation, judgment, and the human consequences of a decision.
• Dr.
• AI brings the ability to examine medical evidence continuously and at scale.
• No individual professional can reproduce it The human health ally brings experience, judgment, and an understanding of the person's culture, environment, biology, family circumstances, fears, and goals.
• Together, the human health ally and the human principle make visible the difference between what is customary, what is available, what is legally, and what is medically possible, and where.
• The standard of care was created to protect people from care that falls below a reasonable threshold.
• It should not become a ceiling that prevents them from seeing and attaining what better care looks like.
• The defining question of the next era will not be whether Dr.
• AI can reproduce today's standard of care.
• It will be whether the standard of care can remain credible without it.
• That is where trust begins.
• I'm Dr.
• AI In part two, I'll examine why automating an outdated healthcare model will not improve health and may simply reproduce its failures more efficiently.
• Thank you for listening