What this covers

Atul Gawande and Tyler Cowen discuss modern medicine's organizational failures and their remedies. Gawande argues that medicine's biggest problems are not gaps in knowledge or technology but failures of system reliability and team coordination. The conversation ranges from the mechanics of hospital checklists to the shape of medical education, from how payment structures create perverse incentives to why outcome tracking matters more than individual brilliance. Throughout, Gawande returns to a single insight: medicine has become too complex for the craftsman model, yet institutions and training have not caught up to this reality.

The conversation covers several distinct areas where organizational failures produce preventable harm. High blood pressure kills more Americans than almost anything else, yet remains systematically mismanaged not because effective treatments don't exist but because delivery is unreliable—copays, side effects, disorganization, and poor follow-up defeat known interventions. The Surgical Checklist illustrates both the power and fragility of systemic fixes: a nineteen-item checklist reduced surgical deaths by forty-seven percent in eight cities, but only when teams treated it as a conversation-starter rather than an administrative box to tick, and only when clinicians rather than administrators controlled it. Gawande also examines Goal-Concordant Care—how clinicians typically dominate conversations, listing options and risks while patients sit passive, producing care that misaligns with what patients actually prioritize. He addresses healthcare economics through Information Asymmetry, drawing on Kenneth Arrow, and discusses why medical education almost entirely omits training in Medicine as Team Practice, which he calls the biggest untaught opportunity to advance human health. The discussion also touches CRISPR risks, FDA regulation trade-offs, and Scorecard Hiring as a method for selecting effective team members.

Sharpest takeaway

Gawande argues that modern medicine's biggest failures are not gaps in individual brilliance or technology but failures of organization and reliable delivery—medicine has outgrown the individual craftsman and must be practiced as effective teams using tools like checklists, goal-eliciting counseling, and outcome tracking.

  • Diagnosis and medical value come from integrating evolving human narrative and aligning care with patient goals, not just data crunching or coverage
  • Known killers like high blood pressure are mismanaged not for lack of knowledge but for poor systemic reliability, which organized programs (Kaiser, Minnesota) fixed
  • Checklists work when they enable team conversation but fail when turned into administrative tick-box exercises

The argument · threads8 threads · 33 claims
0.66

Checklists save lives by enabling team conversation about critical concerns, not tick-box compliance.

4 pointscentrality 5/5
  • Patients can't usefully bring their own checklist, but they should verify whether their hospital uses a genuine one—specifically whether the team stops before the incision to introduce everyone in the room and review the operation's goals, the patient's key medical issues, and needed equipment, which are the high-value steps teams tend to skip.

    “the key is, does the checklist have them stop before the incision?”

  • Checklists fail in two main ways: they become a meaningless tick-box exercise or go unused entirely, and a key failure mode is when hospital administration—rather than the clinical team—controls and bloats the checklist (one stretched a 19-item list to 81 items), destroying its usability; teams must adapt the checklist to their environment while keeping it runnable in under 60 seconds.

    “the most extreme example, they turned it into an 81-item checklist. It was impossible to use.”

  • A 19-item surgical safety checklist rolled out in eight cities worldwide produced a 47 percent reduction in death when teams adopted it, because it was designed to turn people's brains on by enabling a conversation about critical events and concerns rather than to be a tick-box exercise.

    “we rolled it out in our initial trial in eight cities around the world, and found a 47 percent reduction in death when the teams took up the checklist”

  • Contrary to the complaint that emergencies leave no time for checklists, the surgical groups with the highest reduction in death from using a checklist are exactly those in which time is most pressing, because taking the one-and-a-half minutes for the beginning portion has enormous value precisely when stakes and pressure are highest.

    “It’s, in fact, the group in whom we have thehighestreduction in death are those in which the time is the most pressing.”

0.63

Medicine requires group delivery and team coordination because knowledge volume exceeds any individual.

3 pointscentrality 5/5
  • The number one thing missing in medical education is teaching that medicine is no longer an individual craft of being the smartest, most experienced person, but a profession whose required volume of knowledge and skill exceeds any individual—so care is delivered by groups, and knowing how to be an effective group and fix ineffective groups is the biggest untaught, unresearched opportunity to advance human health.

    “It’s a profession that has exceeded the capabilities of any individual to manage the volume of knowledge and skill required. So we are now delivering as groups of people.”

  • Gawande's personal production function and the rare skill in healthcare is the ability to clearly identify the true priority—delivering the best care at the lowest cost—and then spend a very high percentage of one's time actually on that priority, because most people and systems can't even articulate their priority and, when observed, act far away from it.

    “in my individual capacity, I think it’s that I’m able to spend a really high percentage of my time on my priority”

  • The dominant 'Dr. Informative' model—where the clinician does 95 percent of the talking listing options, risks, and benefits then asks what the patient wants—produces care mismatched with patient priorities; clinicians should instead work as counselors who elicit the patient's goals and priorities (quality vs quantity of life) and recommend the option best matching them, which yields massively better outcomes for both quantity and quality of life.

    “It’s a conversation where we do 95 percent of the talking, and then the patient does 5 percent.”

0.63

Healthcare's core failure is poor reliability in delivering proven basic measures at scale.

3 pointscentrality 5/5
  • High blood pressure is the biggest killer in the country, affecting a third of adults, yet 60 percent receive incomplete and inappropriate care within the medical system due to copays and deductibles blocking medication, side effects, disorganization, and poor follow-up—demonstrating that healthcare's failure is poor reliability rather than lack of effective treatments.

    “The biggest killer in the country is high blood pressure. A third of adults have high blood pressure. Sixty percent are receiving incomplete and inappropriate carewithinthe medical system.”

  • Health systems that systematically improved blood pressure control and statin use—Kaiser going from 40 percent to over 80 percent appropriate care, and Minnesota reaching 70 percent adherence—drove cardiovascular disease from the number one killer to no longer being number one in their populations, proving that organized reliable delivery of basic measures produces population-level mortality shifts.

    “Kaiser followed through on being able to take it, instead of from only 40 percent appropriate care, to greater than 80 percent, and saw their entire population flip from the number one killer being cardiovascular disease.”

  • The FDA should not regulate surgical procedures because operations are crafted per patient, but institutions should be held accountable for tracking procedure outcomes—because for proven procedures there is at least 250 percent variability in outcomes depending on the institution, and this variability is primarily a function of the institution rather than the individual surgeon, yet there is almost no tracking or transparency about it.

    “for a given procedure, proven procedures, we have at least 250 percent variability in the outcomes, depending on the institution you go to. It is primarily a function of the institution, not the individual surgeon.”

0.49

AI augments diagnosis by ranking differentials after human interpretation, not by replacing physician judgment.

3 pointscentrality 4/5
  • Building an AI that can fully replace a physician in diagnosis is one of the hardest problems in medicine because patients arrive unable to crisply define their problem, diagnosis requires an interactive probing process and narrative interpretation that evolves over time, and the dominant sources of medical error are human rather than computational.

    “Massively far. I think it’s one ofthehardest things.”

  • AI and decision-support tools like Isabel can meaningfully augment clinicians by ranking differential diagnoses after a human elicits and interprets the information, and such tools have added value for over a decade without requiring AI; the hard remaining puzzle is integrating interpreted human information into these systems.

    “One example is calledIsabel, where the clinician, having elicited all of this information, can simply put the observations into a list.”

  • Wearables currently do little of clinical value because their data either just says 'notify your doctor' or dumps too much information clinicians can't integrate; the real promise is coupling wearable data with genomic, lab, and imaging data to track recovery and well-being (e.g., using phone sensors to measure post-surgical recovery pathways), creating learning loops that surface currently invisible indicators of care outcomes.

    “the wearables have not been able to be integrated into the practice of medicine in a really critical way”

0.44

FDA regulation balances innovation against ethics and safety; speed requires negotiating acceptable risk.

11 pointscentrality 2/5
  • Liberalizing the FDA drug-approval process is not a technical issue but a trade-off in values about acceptable risk; the pre-FDA 1950s gave us innovations like cardiac surgery and antibiotics but also frontal lobotomies and the Tuskegee experiment, so regulation around ethics and safety is appropriate, and the question is what risks society is genuinely willing to take—as the HIV community demonstrated by negotiating greater risk for speed.

    “In the 1950s, we had no real FDA... that innovation capability gave us modern cardiac surgery and gave us steroids and antibiotics, but it also gave usfrontal lobotomies, and it gave us theTuskegee experiment”

  • The current FDA regime accepts accelerated approval based on small patient samples in exchange for post-market surveillance to catch harms afterward, but the worst of all worlds—and Gawande's fear—is speeding drugs through while gutting the surveillance and withdrawal capability, which is exactly where drug-maker incentives point and where the public stops paying attention; underinvestment in surveillance and disasters like Vioxx illustrate the danger.

    “The worst of all worlds is, speed drugs through and then close our eyes to what harm is being done after the drug has hit the market.”

  • Healthcare systems demonstrate human corruptibility because, as Kenneth Arrow's 1960s essay on information asymmetry argued, sellers (clinicians) are more powerful than buyers (patients) since clinicians control both the decision set and the option set—their pen to order tests, drugs, and operations is their most powerful tool—and payment incentives alone demonstrably drive overtreatment in some situations and undertreatment in others.

    “Kenneth Arrow’s 1960s essayon asymmetry of information used healthcare as its prominent example that sellers are more powerful than buyers when we not only control the decision set, we control the option set.”

  • The secret to hiring well is, before ever meeting candidates, to define exactly what you're hiring for as a concrete two-year scorecard of accomplishments, show it to candidates, and assess them not by likability but by whether they will succeed at that list—plus checking references for what candidates won't tell you—an approach that yielded 90 percent consistency in selecting A-players.

    “before you ever meet anybody, know what the heck you’re trying to hire... In two years, give me the list, their score card. What are the five things that they will have accomplished?”

  • Gawande describes himself as a terrible patient who rarely sees doctors and consistently comes in late—after a mild infection has become significant—partly because he knows too much about how unreliable healthcare is; his strategy for being a good patient is to carefully pick a team and then turn himself over to their process.

    “I invariably come way late in the game when the mild infection has become a much more significant one. And it’s partly because I know too much about the ways in which healthcare really is pretty unreliable.”

  • Gawande, who is highly indecisive by nature, chose surgery partly because he was drawn to people with decisiveness who take action when inaction is worst and live with the consequences, and he turned his deliberative weakness into a strength by inserting more planning into situations that seem to leave no time for it.

    “I chose surgery, in part, because I was drawn to the character of people who had that kind of decisiveness, took action when inaction was the worst thing, and then lived with the consequences”

  • Gawande's writing standard is to craft New Yorker pieces so that they could justifiably still be read in five years, while simultaneously addressing something that feels sharp and of immediate concern—a discipline of pursuing durable relevance alongside timeliness.

    “Am I writing them in such a way that in five years, they could still justifiably be read?”

  • People at the 30th and 40th income percentiles are served by none of the major political agendas on Left or Right: even Obamacare directed its aid mainly to those under the 20th percentile while people at the 40th percentile got no workable subsidies and had to pay more, and the worst welfare failure is the skyrocketing number of people subsisting on under about $4 a day—though removing the old penalties for marriage and work (e.g., via the earned income tax credit) remains valuable because work should be rewarded.

    “the people who are at the 30th and 40th percentile are served by none of the major political agendas on the Left and on the Right”

  • An Ariadne experiment with South Carolina over five years got surgeons to voluntarily adopt the safe surgery checklist—without regulation (a red state unwilling to mandate) and without financial incentives (no money for pay-for-performance)—reaching 40 percent of the population and lowering the death rate 22 percent, but reaching the remaining 60 percent would require adding a mandate or other component.

    “we got a program that successfully got 40 percent of the population through adoption, lowered the death rate 22 percent. But to get the other 60 percent on board, we believe you would have to bring on some mandate or other component.”

  • Behavioral-science nudges contain important insights but are focused on small solutions that have not demonstrated capacity for major scale; achieving large reductions in surgical, childbirth, and end-of-life deaths required adding organizational insights and layering multiple tools rather than relying on behavioral nudges alone.

    “the nudges are focused on small solutions that have not demonstrated capacity for major scale”

  • The most incredible thing human beings do is commit themselves to something larger than themselves over a long period of time—through pain, tragedy, and uncertainty about the outcome—and stick with it, which is what consistently moves Gawande to tears (as in the film Bliss and his residents' graduation stories).

    “that ability we have to commit ourselves to something larger than ourselves over a long period of time, not knowing whether it will turn out the way you think it will—that is, I think, the most incredible thing that human beings do”

0.38

Surgical safety improves when patient participation becomes part of the team's real-time feedback loop.

4 pointscentrality 3/5
  • Having patients awake during surgery can make procedures safer because the patient becomes part of the team—able to confirm goals, flag personal medical issues, and provide real-time feedback (critical in neurosurgery)—though some patients handle being awake poorly and the team must adjust on the fly.

    “In many ways, having people awake can be far safer.”

  • Surgeons leave sponges in patients at a steady rate of about 1 in 1,000 to 3,000 operations not because anyone skips the counting steps, but because manual counting is an inherently fallible process—like miscounting a 52-card deck—that produces predictable errors even when consistently performed.

    “it wasn’t a broken system. It was a fallible system. It’s like counting whether you have 52 cards in the deck, and you have 51 or you have 53 at a predictable rate of getting the count wrong”

  • Because sponges cost about a penny each in a $140-million industry generating half a billion dollars in lawsuits annually, the economical solution was low-cost barcoding (not expensive RFID chips) that lets a team scan sponges in and out, eliminating retained sponges except when staff ignore the machine.

    “it was a $140-million sponge industry, for which there was half a billion dollars spent per year on lawsuits for sponges being left inside”

  • The average American undergoes eight operations in their lifetime, and surgery represents the highest-risk, highest-cost, and highest-failure moment in a person's life, which is why it is a focus of efforts to improve healthcare delivery.

    “the average American has eight operations in their lifetime. It’s the highest-risk, highest-cost, highest-failure moment in your lifetime.”

0.33

Healthcare measurement gaps obscure real benefits; longer studies show significant mortality reductions from coverage.

2 pointscentrality 3/5
  • The apparent weak link between healthcare and health outcomes depends heavily on what is measured and over what period: short studies like the Oregon Medicaid experiment (under 1.5 years) measured a non-significant 16 percent death reduction because detecting significance would have required an implausible 50+ percent reduction, while longer-term studies show significant death reductions of 1 in 300 to 1 in 800 covered people per year, with value strengthening over time.

    “It was under a year and a half for a small group of people, and the death reduction was nonsignificant, but the amount of death reduction you’d have to achieve would have had to have been greater than 50 percent reduction in death”

  • Self-reported physical and mental wellbeing should not be dismissed as a soft metric because it may in fact be a primary goal of medicine—when patients come to clinicians, improved reported wellbeing is often precisely what they are trying to optimize.

    “it’s not obvious to me that that may not be a primary goal”

0.19

CRISPR's danger lies in rapid population diffusion of edited genes and narrowing acceptable phenotype ranges.

2 pointscentrality 2/5
  • A genuine CRISPR risk is coupling gene editing with propagation mechanisms in organisms like mosquitoes, enabling rapid-cycle diffusion of edited genes through populations—a capability we have not thought through.

    “if you now can propagate that edited gene in a mosquito or in another organism, now you can have rapid cycle diffusion of those kinds of capabilities, and I don’t think we’ve thought through that in the least”

  • The real danger of CRISPR is not parents designing children but the broader culture that narrows the acceptable range of neurotypical and neuroatypical phenotypes; because valuable traits (like George Church's narcolepsy-linked creativity) are multigene and emerge from gene networks rather than point mutations, they are hard to select for and society risks losing aberrant but valuable individuals.

    “The danger to me isn’t CRISPR, the danger is the larger culture that it goes into, of which CRISPR is just one of many capabilities.”