Time Stamps

  • 02:51 Broad workup for reversible causes and other etiologies of AFib that may occur post-op
  • 05:10 Considerations for management of post-op atrial fibrillation
  • 13:00 Stroke risk in atrial fibrillation
  • 20:49 Outpatient management of atrial fibrillation
  • 25:54 The role of AI in medical decision-making

Sponsor:Ā Amazon Pharmacy

Search ā€œAmazon Pharmacy Nationwide Home Delivery”Ā in your EHR to get home delivery (often same-day). Learn more HERE!

Show Notes

Post-op afib, like all afib, depends on the clinical contextĀ 

  • Broad workup for reversible causes and other etiologies of afib that may occur post-op
    • Pulmonary embolism (check the EKG for signs!)
      • EKG changes:
        • Step 1: Lead I – Is there an S wave? S wave in Lead I = Suspicious for PE
        • Step 2: Leads V1/V2 – Is there RSR’? RSR’ in V1/V2 = Suspicious for PE
    • Sepsis
    • Post-op hemorrhageĀ 
    • Electrolyte imbalances
    • Fluid shiftsĀ 
    • Thyroid function testsĀ 
  • Considerations forĀ  management of post-op atrial fibrillation
    • Symptomatic vs. asymptomatic
      • Symptomatic afib should prompt more aggressive treatment measures as this has a direct effect on our patientsā€™ quality of life
      • Symptoms sometimes are unmasked during physical therapy or other exertionĀ 
    • Cardioversion
      • Patients generally feel better when they are in sinus rhythmĀ 
        • allows them to participate more actively in rehab
          • Can affect the quality of their post-op recoveryĀ 
    • The duration of atrial fibrillationĀ 
      • Very short periods of AFibafib generally put patients in a lower risk category than patients that remain in afibĀ 
      • Historically studies have defined episodes greater than 5 minutes to exclude artifact oversensing from cardiac devices, however this is a prespecified time point and has not been studied in randomized trialsĀ 
    • EchocardiogramĀ 
    • AnticoagulationĀ 
      • For patients who tolerate surgery without major bleeding complications and elevated CHADsVASc scores, the benefit of anticoagulation, at least in the short term after atrial fibrillation develops, may outweigh the risk
      • Each patient is different and deserves their own nuanced risk vs benefit discussion
  • Stroke risk in atrial fibrillationĀ 
    • Afib-related stroke is traditionally thought of as related to stasis in the left atrial appendage, which can create a clot that can embolize to the brainĀ 
    • afib is likely a reflection of broader atherosclerotic riskĀ 
      • The CHAā‚‚DSā‚‚-VASc score serves as a proxy for atherosclerotic risk, with elements that are markers of systemic vascular health
      • When we are caring for patients, it is important not to get trapped into only managing anticoagulationĀ 
        • Consider general reduction of stroke/atherosclerotic risk factors including lipid lower therapies, better control of diabetes, aggressive management of hypertension, treatment for sleep apnea, etc.Ā 
  • Outpatient management of atrial fibrillationĀ 
    • Long term management of anticoagulationĀ 
      • Risk vs benefitĀ 
      • Patient preferenceĀ 
      • Insurance and affordability of medications
    • Role of holter monitors and loop recorders
      • NOAH-AFNET
        • Clinical question: Do patients with subclinical atrial high rates (afib, atach, etc) have decreased rates of stroke when treated with edoxaban vs no anticoagulation?
        • Trial type: Multicenter, double blinded, randomized control trial
        • Patient population
          • >= 65 years of age
          • Atrial high rates (atrial rate of at least 170bpm for > 6 minutes) deteceted on implanted deviceĀ 
          • Presence of one or more of: heart failure, hypertension, diabetes mellitus, previous stroke or transient ischemic attack, vascular disease (previous myocardial infarction, aortic plaque, or peripheral, carotid, or cerebral arterial disease), or an age of 75 years or olderĀ 
        • Intervention: patients were randomized to receive either edoxaban or a placebo pill. The placebo pill was either aspirin if patients had indication for aspirin or one with no active compound if there was no such indication
        • Results: anticoagulation with edoxaban did not result in a lower incidence of a composite outcome of cardiovascular death, stroke, or systemic embolism but did result in a higher incidence of a composite of death from any cause or major bleeding.The trial was terminated early given concerns of safety and futility
      • ARTESIA
        • Clinical question: do patients with short episodes of subclinical atrial fibrillation have decreased rates of stroke when treated with apixaban vs aspirin?Ā 
        • Trial type: Multicenter, double blinded, randomized control trial
        • patient populationĀ 
          • Age >= 55 years
          • subclinical atrial fibrillation episodes 6 minutes ā€“ 24 hours in duration that was detected by an implanted pacemaker, defibrillator, or cardiac monitor,Ā 
          • CHAā‚‚DSā‚‚-VASc score of 3 or higherĀ 
        • Intervention: patients were randomized to receive either apixaban 5mg twice daily or aspirin 81mg daily
        • ResultsĀ 
          • Patients that received apixaban for short episodes of subclinical atrial fibrillation had less strokes and also had higher bleeding risk compared to those that received aspirin

        • Of note, a subsequent metaanalysis that pooled the data of both trials did show a significant benefit (relative risk 0.68 [95% CI, 0.50-0.92]) in reduction of ischemic stroke in patients with device detected atrial fibrillation, though this benefit was not seen in NOAH-AFNET alone
        • All of these studies are not directly translatable for each patient
  • The role of AI in medical decision making
    • Uncertainty is a part of medicine
      • Evidence is incomplete
      • Experts may reasonably disagree and there may be more than one correct answer as long as it is justifiableĀ 
    • Clinical Judgment vs Algorithm
      • Medicine requires contextual, individualized decisions
      • The role of doctors is to help patient live better and to live longerĀ 
      • Clinical guidelines frequently establish the ā€œminimum acceptable careā€, but physicians can and should provide much more than this which requires:
        • Customization of treatment based on patientā€™s individual situations
        • A mental model of real-world barriers and the development of feasible and realistic treatment plansĀ 
        • Ability to apply scientific literature to each individual patient

Transcript

Dr. Greg Katz: That’s honestly how everything in medicine works, which is when we start out, and we have seen things a couple of times, we’re like, “Oh, this is simple, and it’s straightforward.” And then there’s something about you, do this a hundred times and then 200 times and then 500 times. And the questions that seem like they’re so basic and straightforward and like, “Oh, duh, that’s obvious.” Well, you often realize that, actually, it’s just convention. It’s not actually based on the level of evidence that any of us would want our family to be taken care of. But it’s passed down, and it’s codified through the unwritten language of medicine, and I’m looking forward to talking about the case.

Dr. Shreya Trivedi: Welcome to the Bread and Butter Series, a new series where we compare and contrast cases to help you better appreciate the nuances that we all tend to oversimplify. Today, Dr. Angela Wang, a resident from NYU, will present scenarios that come up often in post-op AFib to one of our favorite cardiologists, Dr. Greg Katz, and we also have another special guest, OpenEvidence.

Dr. Angela Wang: Hi, Greg. So tonight I wanted to talk with you about a case I had on the wards at the end of intern year. It was a case my thought was relatively straightforward and then later realized there was a lot of nuances that I didn’t really think about. So I put it into OpenEvidence who gave me super confident answers on how to manage this patient, but I’m now really curious to hear your thoughts on the case and also on how you think OpenEvidence answered my questions.

Dr. Shreya Trivedi: And throughout the episode, you will hear the MedEd King, Dr. Gurpreet Dhaliwal from UCSF, on his reflections on human versus AI when it comes to management reasoning. I just love this episode and really have sent it to so many people because I just appreciate having that headspace to see, okay, what does human versus AI bring to the convo? It’s not a fair comparison by any means, but at least we can appreciate it a little bit more in this specific context of post-op AFib.

Dr. Angela Wang: Amazing. Okay, let’s get started. So this is a 63-year-old man, past medical history of hypertension, hyperlipidemia, type two diabetes. His last A1C is 8.4, not on insulin. He was initially admitted for a total knee replacement. He underwent this knee replacement with no complications, and then in post-op, he develops AFib with RVR. His heart rates are 140s, otherwise hemodynamically stable. And because of this, he’s transferred to medicine for further observation and management of his AFib. We started him on Metoprolol Tartrate. His heart rate decreases to the 100s, but he remains in AFib. So, what further information do you want to know to figure out how to manage this patient?

Dr. Greg Katz: Did he have any kind of preoperative evaluation? Do you have a baseline EKG? Do you have a baseline set of the other labs?

Dr. Angela Wang: He has a baseline EKG from a couple years ago, which is normal sinus, no abnormalities on it, and otherwise his labs are all within normal limits.

Dr. Greg Katz: Yeah. So post-op AFib is insanely common, and how you approach it or how I approach it is completely based on who is this person, what was their surgery, and how are they doing? And so when I hear a patient goes for a knee replacement and then they develop post-op AFib, the first five things on my mind are make sure this patient does not have a PE. And so I’m going to say, show me the 12-lead EKG. I’m going to stare at that EKG and I’m going to look in two places specifically. First, I’m going to look at lead one and I’m going to examine it for the hint of an S wave because an S wave in lead one is telling me about late rightward forces that make me suspicious for PE. He told me in his 60s, he has an A1C of over eight. And so it is very plausible that even an S wave in lead one or even an S1Q3, T3 is actually just somebody who has untreated sleep apnea, but it’s going to go through my mind as something that I’m paying attention to. And I’m going to look to see is there any RSR prime in V1 V2 because that’s telling me the same story. I’m going to say my job in a post-op orthopedic surgery patient is I need to rule out a PE, and I need to be pretty sure that there is no PE before I decide I’m not going to scan this guy. And then I want to know how much blood he lost during surgery. I want to know the nature of what his labs look like now. I want to make sure that he doesn’t have an undetectable TSH and surgical site infection, make sure he’s not septic, because I think of AFib with RVR in a hospitalized patient as the same way I think of sinus tachycardia. And so I’m going to go through all of the things that make somebody tachycardic.

Dr. Angela Wang: Okay. So we largely followed your thought process to rule out reversible underlying etiologies of this new AFib. We sent a basic workup. His CVC showed a white blood cell count of 7. His hemoglobin was 13.4, which is stable from pre-op. His platelets are 256. His BMP had normal electrolytes. His liver function tests were normal. His TSH was 2.1. On infectious workup, his urine was bland with no reflex to culture, and his blood cultures were so far no growth. His chest X-ray had a largely normal cardiac silhouette, no pleural effusion, and no pulmonary congestion.

Dr. Greg Katz: Does he have symptoms? How does he feel?

Dr. Angela Wang: On exam, he appears really comfortable. No chest pain, no palpitations. He doesn’t feel lightheaded sitting in bed. He has no lower extremity edema, and he has no pain in his legs. He isn’t requiring oxygen, and he’s sat in fly on room air.

Dr. Greg Katz: Okay. And if he was not hooked up to a telemetry monitor, he’d have no clue that he was in AFib.

Dr. Angela Wang: That’s correct. So in summary, he is an asymptomatic AFib hemodynamically stable with some slowed rates with metoprolol, but no cardioversion back to sinus. Now what’s your approach to managing him?

Dr. Greg Katz: He stayed in AFib despite an evaluation and some beta blocker. And so for me, I’m treating this as though my default is going to be anticoagulate him at least for the time being, especially because I’m going to cardiovert him. And the reason I’m going to cardiovert him is I don’t actually know whether he has exertional symptoms. He is sitting in bed postoperatively after a knee replacement. That rehab stinks, and anything that is going to limit his ability to participate fully in that rehab is a true problem for this guy moving forward because the way that somebody recovers from surgery is very much dictated by how much mobility do they do and how hard do they work on their rehabilitation. And so if there’s even a small chance that he’s going to be limited rehab-wise because of his AFib, I want to put him in sinus rhythm, and I want to see whether or not he feels better. And there’s a fair number of people who are in AFib and don’t think that they have symptoms, and then you put them into sinus rhythm, and they’re like, “Oh, actually I did have symptoms, and I just didn’t realize it.” And a TEE with cardioversion is a pretty low-risk procedure. He literally just had this surgery, and so if he hasn’t bled in the hospital on the medications he’s on, his bleeding risk over the short-term horizon in that 30-day period after I put him in sinus, which is when thromboembolic risk is higher, it’s just like reasonably low. And so that’s like my immediate sort of approach to a case like this.

Dr. Angela Wang: Okay, this is great. Let me give the AI a chance to catch up, and let me tell you what it said to do up to this point.Ā 

Dr. Greg Katz: Sure.Ā 

Dr. Angela Wang: Interestingly, open evidence told me that he could be discharged once we get adequate rate control and there’s no evidence of any post-op complication, but we ended up doing what you suggested while he was still in the hospital.

Dr. Greg Katz: So, open evidence is not wrong that the patient can be discharged from the hospital if feeling well. The question of like ā€œcan beā€ versus ā€œshould beā€ and like how would I want to be taken care of versus what is the minimum medical legally acceptable nature of medical care? Those are very different things. And so if that were me and I had post-op AFib and my A1C was 8.3 and I was 63 years old and I just had a knee replacement, honestly, I would want a little bit more done for myself, and I would want the same thing for my family. It’s totally okay, but the difference between okay and the level of care that you want to provide to somebody that you really are trying to take wonderful care of is different.

Dr. Shreya Trivedi: Okay. With that, let’s turn to Dr. Gurpreet Dhaliwal on his reflection so far.

Dr. Gurpreet Dhaliwal: One of the things that you saw that Greg does, which AI systems can’t do, is he exhibited moral judgment. So if you listen to how he said ā€œcanā€ versus ā€œshouldā€, or what’s the minimal versus the optimal, or really like wrong versus right. There’s just a lot of moral judgments that go into it, and you’re a human to sign off on it, but you actually really need someone who cares that much to make the judgment. I think that’s a mistake almost all of us make when we start medical education is you’re like, I just want to ascend to all the scientific knowledge. And if I can master it, then my decision-making will be highly, not perfect, but it’ll keep getting better and better. Understanding the scientific literature is a modest part of what we do as doctors. People don’t hire us or seek our care because we master this whole canon of knowledge and apply it to every person the same way. They seek us because you’re really exceptional at customizing it per patient. It’s the exact opposite of science. I think what Greg did was a really great job of explaining why he would do things the way he did for this patient. He said rehab is very difficult. In order to go through rehab, you have to be at your very best. It’s very possible if I can get him into normal sinus rhythm, he will be at his very best. But he has a mental model of what needs to happen post-op. I mean, that is an incredibly sophisticated number of steps that we may find routine until you try to program that into a computer.

Dr. Angela Wang: What is your threshold of keeping this patient in the hospital to get an echo before discharge?

Dr. Greg Katz: 100% I want the echo before. If this patient is going to have quick follow-up, that’s totally fine, but he shouldn’t wait four weeks for an echo. And honestly, I probably would not have him leave the hospital without restoring sinus rhythm. And so I’m going to look at his heart on a TEE before I cardiovert him. He’s a patient who has somewhat poorly controlled diabetes. And so my suspicion that he has some kind of underlying coronary disease is pretty high. A 60-year-old diabetic patient who just went for a knee replacement probably has some plaque in his arteries. And so I don’t know that his ejection fraction is not 35%. This decision of, do they need to stay in the hospital or do they not need to stay in the hospital? I would actually sort of put it back on you, and I’d be like, “How confident are you have any idea what is going on with this guy’s heart?” And the answer is you shouldn’t be that competent because you don’t have very much information. And when you don’t have very much information, and somebody has an unexpected complication, and I would describe post-op AFib as an unexpected complication, your job is not to satisfy the O:E ratio. Your job is to take care of the patient, and it’s to figure out what’s going on. And if you discharge this guy with no echo in AFib and he gets an echo 48 hours later and his EF is 35%, you’re going to feel like you missed the boat taking care of him in the hospital. And so my sense is that when somebody has an unexpected complication happen, the right thing to do is to figure out as much as we can about what went on.

Dr. Angela Wang: Okay. I like that. And we’ve kind of hit on this already, but now let’s talk a little bit more about anticoagulation and stroke prevention in AFib.

Dr. Greg Katz: He’s at elevated thromboembolic risk, and he’s at elevated thromboembolic risk because his CHADS VASc score is quite high. He’s in his 60s. I think he’s 63, but is 63 truly that different than 65? Probably not. He has diabetes with an A1C of eight. He’s hypertensive. And so I have enough points in CHADS VASc score for me to say that I’m probably going to anticoagulate him. And he didn’t have a self-limited episode of AFib. And so the persistence of the AFib for even just the duration of the hospital stay says to me that I am treating him now as a paroxysmal AFib case. And the other part of this that’s really important is he had no clue that he was in AFib. I have no idea whether he’s had AFib on and off for the last five years of his life. And so he’s 100% somebody I’m anticoagulating regardless of whether or not we cardiovert him based on the data that we have.

Dr. Angela Wang: We talk about stroke risk and AFib a lot, and we use the CHADS VASc score to risk-stratify, but how exactly does AFib cause stroke? Is it really as simple as stasis in the left atrial appendage?

Dr. Greg Katz: Stroke risk in AFib is something that doctors are overly simplistic and concrete about too often. And so what I mean by that, it’s like, why do people have strokes when they have AFib? Well, one possibility is you have a thrombus form in your left atrial appendage from the disorganized atrial contraction, and then a piece of that breaks off, and it travels to the brain. But that’s only one mechanism of AFib. And so what’s in the CHADS VASc score? Nothing to do with the left atrial appendage, nothing to do with the left atrial appendage emptying velocity, the size of the left atrium. Nothing about the mechanics of the heart. The CHADS VASc score is basically asking how likely is this person to have atherosclerotic cardiovascular disease? And the CHADS VASc score, it’s not magical. It was derived like every other risk calculator in medicine, which is a retrospective cohort of patients. They figured out through some complex statistical analysis what seemed to independently cause an increased risk of stroke, and then they prospectively validate it, and then it just sort of gets put into the medical canon as the way that you are supposed to do it. And if you have a CHADS VASc score of 0 or 1, well, you can’t have a stroke with AFib. And if you have a CHADS VASc score of 3 or more, 2 or more, then God help the doctor who doesn’t prescribe that person anticoagulation. And you look at it like the Eliquis is all that matters, but I think it’s super, super important that when you think about a patient who has AFib, you don’t fixate on the CHADS VASc score and the left atrial appendage as the be all and end all of stroke because even if you take patients who have AFib and strokes and continuous monitoring, not all of the people with AFib who have strokes are in AFib at the time of their stroke. And not all people who have AFib and have strokes have a blood clot leave their left atrial appendage and go up to the brain. And so if you don’t think about all of the other competing risks of stroke, that’s an incomplete mental model of how to think about stroke in AFib. And I would just recommend that everybody who thinks about stroke doesn’t get fixated on just the left atrial appendage.

Dr. Angela Wang: Okay. Makes sense. Next question, what if he self-converted to sinus? Like, what if the AFib only lasted five minutes while he was post-op? Would that change your decision-making?

Dr. Greg Katz: If the AFib only lasted five minutes, it’s a wildly different ballgame, and then my default is not going to be to anticoagulate him. It’s going to be to monitor him more. And so the duration of AFib to me is really influential, and whether or not somebody like converts because we did something to them versus just converts on their own is also an influential thing for me. As far as I know, there’s no really good data to assess whether self-conversion versus medical conversion actually makes that much of a difference. But my perception is that if somebody’s converting on their own without intervention means that their phenotype is somewhat lower risk. But if somebody has a really short self-limited run of AFib, then I think that you need to look at them like they’re probably not the same level of risk as somebody who’s in AFib for 36 hours and needs to be cardioverted or leaves the hospital whilst they’re still in AFib. There’s a bunch of different clinical trials. There’s Artesia, which looked at subclinical AFib. There’s NOAH AFNET, which looked at subclinical atrial high rates that were really, really short and they all kind of like tell a different story. NOAH AFNET was less impressive than Artesia was. Artesia basically looked at pretty long episodes, but those are not post-op patients. Both of those studies are not post-op AFib. Both of those studies are people who have pacemakers who just incidentally find that the pacemaker has detection of atrial fibrillation or atrial high rates. And in medicine, you kind of need to be careful. So the question of what is reasonable extrapolation and what is unfair, it’s a gray line. I’m super curious what open evidence had to say.

Dr. Angela Wang: Right. So for the question of if it only lasted five minutes, it said a five-minute episode of AFib does not warrant anticoagulation based on current guidelines, even with the CHADS VASc score of 3. And then it goes on to quote the studies that you just talked about. So it quoted Artesia, and then it quoted NOAH AFNET to back up its decisions.

Dr. Greg Katz: NOAH AFNET was not AFib. It was atrial high rates. And so is atrial tachycardia the same? One of those things, as you dig into, there’s no way you come away with a higher level of confidence that anybody knows anything. It’s nice to think that we really understand it, but I think that really it’s way more complex, and we need to have a fair amount of intellectual humility when we think about how we counsel patients, and just the level of uncertainty is truly profound in cases like this.

Dr. Gurpreet Dhaliwal: Greg had one interesting thing. It was sort of philosophical. He was talking about the trials, and then he was talking about their shortcomings and how you can’t really extrapolate them to our case, so we do so in faith, even though we have a reference class problem. And at some point, he says, I think he said, nobody knows anything philosophical, but it gets back to the point, like there is no answer. I think sometimes you’re like, well, there’s just some answer in the literature that knows it, but it’s all imperfect information. The patient data isn’t perfect, our brain’s imperfect, the literature isn’t perfect, and what a professional does is makes a judgment, but there’s always uncertainty. And that’s a key point. That’s why we’re training everyone. If there’s no uncertainty, then you don’t need a doctor. That’s our job.

Dr. Angela Wang: Is there a specific time marker that you kind of change your mind about? Is five minutes where you draw the line, or is it still a super gray area?

Dr. Greg Katz: It’s a super gray area, and it’s also like, who is this patient? And so for this patient, if he has five minutes of AFib, he is so much better to be just put on a GLP-1 agonist than he is to be put on an anticoagulant when it comes to stroke reduction. And does he have sleep apnea? How are the rest of his atherosclerotic risk factors controlled? And then he’s had CTs in the past. How much calcium does he have in his arteries? And so I really try to be holistic when I think about stroke prevention in patients who have uncertain amounts of AFib, and I think that you need to look at the entirety of what his stroke risk profile is.

Dr. Gurpreet Dhaliwal: Greg was talking about CHADS VASc and he raised a great point. He’s like, “All the things in CHADS VASc are not exactly directly mechanistically tied to how AFib causes a stroke. They’re sort of correlations.” And so he has a deep understanding of mechanisms in a way that I wouldn’t expect the LLM to know. And then the ability of humans to think outside the box, outside of this whole conversation, he’s like, “Actually, if I wanted to reduce stroke, I might say maybe the GLP-1 is a more promising avenue than an anticoagulant.” That is a purely human instinct. LLMs are great at remixing things that humans have come up with in the past before in the literature, but humans are still, right now at least we have the advantage in creative solutions. And that was a really great example of it. I myself thought about that, maybe when I learned from that, that maybe when I’m focusing on stroke reduction in my AFib patient, there’s other processes I might want to think of. And if we don’t go down the anticoagulant route, that doesn’t mean that there aren’t some really promising and who knows, maybe even better approaches.

Dr. Angela Wang: All right. I’m curious now about your thoughts and how do you use Zio patches or loop recorders for more long-term monitoring of AFib in these patients?

Dr. Greg Katz: We make a decision based on a snapshot in time, but to really take care of a patient like this, you need longitudinal follow-up. You need to have somebody wear a heart monitor. I don’t know whether OpenEvidence recommended a loop recorder, but a loop recorder, it’s something you would consider, but loop recorders have been studied really interestingly because they’ve been studied for detection of AFib in patients with cryptogenic stroke. Very importantly, like Crystal AF, which is the study looking at detection of AFib in cryptogenic stroke with loop recorder versus standard monitoring, which is like 28 or 30 days, which found over the course of a year, I think around 12 and a half percent of patients with a loop recorder were found to have AFib. It’s really, really important that we remember Crystal AF was not a stroke prevention trial. Crystal AF was an AFib detection trial. And so we don’t actually know whether using a loop recorder to detect AFib and then choosing to prescribe anticoagulation because of the presence of AFib is a stroke reduction strategy, because that’s not how it was studied. Crystal AF tells us when we look for AFib for a long time, we find more AFib when we look for a short time, but it doesn’t give me any confidence that I have truly identified the mechanism of stroke.

Dr. Angela Wang: Okay, makes sense. So we did end up discharging him with a Zio patch. So, say now that you see him in clinic and the Zio patch is completely negative and he is still super asymptomatic, now where do we go from there?

Dr. Greg Katz: He got the Zio when he was in sinus, and then he leaves the hospital, and he wears the Zio for two weeks in sinus, and he’s on Eliquis, and he follows up in clinic. I would probably keep him on the Eliquis for at least a full month because he was in AFib for a pretty long period of time. And then I’m at least repeating a monitor after another couple of months because my level of confidence that we really understand things over two weeks is pretty low. But I also look at this case as a case where, at least in the short term, the risk of staying on anticoagulation is pretty asymmetric in a way that favors anticoagulation. And so what I mean by that is the likelihood of a 63-year-old non-chronically ill, non-frail man, what’s his risk of having a serious bleeding complication? It’s like pretty low, and you can plug in whatever risk calculator you want it, but to me, that misses some of the nuance of like how I actually think about this decision. And so we put him on anticoagulation for a short period of time. And what’s the downside of that? Well, he might have a self-limited bleed. The chance of him having a serious, life-threatening, life-changing bleeding complication is pretty low. And if he has a stroke, it has the potential to just totally ruin his life. To me, the burden of proof is on why we should stop the anticoagulation super early. In real life, you run into practical implications like he wants to go skiing, or he hates being on a blood thinner, or his copay is a million dollars a month. I think that ultimately the real-world implications are going to dictate what we actually do, but that point of risk is asymmetric is a really vital component of how I think about what that decision-making is going to be.

Dr. Angela Wang: That makes sense. He seems like he could tolerate AC just fine.

Dr. Greg Katz: Yeah. What did Open Evidence say?

Dr. Angela Wang: OpenEvidence said that his chance of bleeding is much less than his chance of having a stroke, and so it would advocate keeping this patient on Eliquis indefinitely. And then I asked, what if he follows up in one year and he’s not had another episode of AFib? What are its thoughts on continuing AC at that time? And it said you should still just continue it because the benefits of continuing AC far outweigh the risks for this patient.

Dr. Greg Katz: Yeah. Maybe you should, but maybe you shouldn’t. That’s just like a gray area. Nobody knows what the answer is.

Dr. Gurpreet Dhaliwal: When I was earlier in my career, I would be frustrated that three neurologists would handle it this way or that two GI doctors would approach this GI bleed differently. But as you go on further and further, you just realize there’s no one right way because there’s too many variables and there’s too much complexity. It’s really sort of, is it a defensible plan that’s put in place rather than sort of being fixated on there as an optimal one, and if I get the right consultant or if I get the right AI, I will get there. We got Greg consulting on this case, but if Angela picked up the phone and got a different cardiologist, there’s no doubt she would’ve heard some different advice and guidance and it would’ve been defensible and logical in its own way, but non-identical. In AI, I think many of your listeners know this, if you typed in the same case a second time or third time, there’ll be different variations on the advice. It’s not fixed advice. It’s stochastic in what it puts out.

Dr. Angela Wang: I thought it was really interesting because AI never asked me if the patient was symptomatic from AFib.

Dr. Greg Katz: That’s pretty wild.

Dr. Angela Wang: Right? You asked me, but the initial stem I gave didn’t comment on this on purpose and then halfway through the prompt, the AI assumed this patient was symptomatic and based all of its answers on a symptomatic patient.

Dr. Greg Katz: What’s our job as a doctor? I think we have two. One is to help patients live better, and one is to help them live longer. And so if you don’t figure out if somebody has symptoms, you completely lose the ability to figure out the first one of how do you make somebody live better. And so I hope the chatbot learns its lesson and is going to ask, do a little bit more thorough history taking next time.

Dr. Gurpreet Dhaliwal: I think what it really showed is that humans, no other humans, and perhaps more intellectually, they have a mental model of how humans operate. But Greg, just like all doctors, over time learned that patients may minimize patients may not recognize symptoms. He had a great point. He’s like, “Oftentimes patients feel better after our treatment and then they recognize what they thought was an asymptomatic state was.” But I just want to point out that humans have mental models of a lot of things that computers don’t have. So they have mental models of how humans behave and may minimize symptoms. They have mental models of how the health system works. I mean, so many of the things in the conversation were about how can this be done in healthcare, right? Can the TEE happen? Can the echo happen before discharge? What things might be a barrier to taking anticoagulation That is an extremely rich mental model. And again, it goes way beyond what would be captured in PubMed. There’s nothing in PubMed that would help someone make those judgements. And if you accept the premise that our job is to make professional judgments, you have to know humans, and you have to know the health system. And if you don’t know either one of those two things, you will be ineffective in judgment. I really enjoyed the conversation. I learned a lot of perspectives on post-op AFib and maybe AFib in general. I came away thinking that OpenEvidence was a useful tool, but more than anything, I think it was a really great example of the power of the human brain and the role of what we do in medical education, which is that we train for professional judgment. That is what we do, and that was on full display there.

Dr. Shreya Trivedi: Today, we talked a lot about what we should do for our patients versus what must be done. So I do want to acknowledge in real life, there’s so many other factors that go into what may be the best thing that we can do for our patients versus what we actually do. And I will say, I think the thing that hurts me the most is whenever I hear like, “Oh, that’s not something we do inpatient with to do that outpatient.” That always trips me up. And I just want to acknowledge that aspect of the application, right? Whether it comes from a thoughtful cardiologist friend or from AI, there are just very practical barriers to implementing it for our patients, and that is so real. But thank you so much for joining us today. We hope you appreciated this episode. I know for me it was so eye-opening, and I can’t wait to do more of these. If you’d like to join the team of people who help think about bread and butter cases, love medicine, love to compare management nuances, plus minus, where does AI compare and help in the fold of things, please reach out via coreimpodcast.com. Send us an email, coreimpodcast@gmail.com. We’d love to hear from you and be a part of the team. And if you liked this episode, as always, our one ask is to please send it to one other colleague. It really helps people find us and hopefully get some thought-provoking ideas too from this episode.Ā Thank you to everyone who made this episode possible, both on and off air. As always, opinions expressed are our own and do not represent the opinions of any affiliated institutions. Thank you.

References