Every Cut Is a Prediction
1. The Spleen That Migrated
Beverly Bryan, a retired nurse, says that after an operation in Miramar Beach, Florida, on August 21, 2024, a surgeon told her that her husband's spleen had been four times its normal size and had migrated to the other side of his body.
Her husband, William Bryan, was seventy. According to the lawsuit she later filed, he came to the hospital with left upper quadrant abdominal pain. The spleen sits on the left side of the abdomen. The liver sits on the right. The planned operation was a laparoscopic splenectomy. He died on the table. The lawsuit says the specimen weighed 2,106 grams. At autopsy his spleen was still in his abdomen. The suspension order alleges that his inferior vena cava had been cut.
The surgeon had removed his liver.
Florida's emergency suspension order alleges that the surgeon, Thomas Shaknovsky, fired a stapling device blindly into an abdomen he could not see for blood, then placed the liver on the table and told the staff it was a spleen. According to the order, one of them felt sick, and the staff member who labeled the specimen knew the label was wrong and wrote it as instructed. The order says the operative note blamed a ruptured splenic artery aneurysm and the autopsy found none. In April 2026 Shaknovsky was indicted for second-degree manslaughter. He has pleaded not guilty.
How did this happen? The simplest explanation is skill. According to the order, staff questioned whether he had the skill for the operation, and the hospital did not regularly perform splenectomy. Regulators allege two earlier cases in 2023; in one, part of a pancreas was removed instead of an adrenal gland. All of that is allegation.
None of it explains the liver. Skill explains why an operation gets into trouble. It does not explain why the operation never stops. A surgeon with limited skill and a calibrated estimate converts, packs and holds pressure, calls for help, or stops before the irreversible step. Knowing that you are lost is part of skill, and certainty is what switches it off.
2. The First Prediction
To operate is to harm someone on purpose. We cut because we predict that the short-term harm will become long-term benefit. That is the first prediction, and every operation rests on it.
After the incision, every step is another prediction: that this move gets you closer to the goal. A dissection is a small harm. Moving an instrument risks injury. The work is to progress through a series of predictions, each paid for with some harm, that together vindicate the first one: that you will help this person.
Most of us learned surgery as a list: step one, step two, step three. A list says nothing about how you know a step is finished, or what to do when it is not working. Each move, down to the smallest turn of an instrument, is a question you are asking the tissue.
3. Moves That Change the Odds
You can spot a surgeon who has not learned this by watching how they treat the tissue. They touch it, nothing happens, and they touch it again in the same place, the same way, hoping something will change. Two or three times.
In Bayesian terms, the second touch is not a new test. Its result is already predicted by the first, so it adds almost no information and still adds harm. Repetition that advances the plane is progress: keep going while the plane is giving. Repetition that returns the same answer is not. If I make a move and the plane does not come, I reposition the instrument. I change the traction. I turn the camera. I come from another side. Each change is designed to produce evidence I do not have yet.
In a rectourethral fistula repair after radiation, the question is the plane between the prostate and the rectum. What is the evidence I am in the right one? Is that perirectal fat? Is the plane white? Lifting the prostate is not a dissection, but it still advances the case. It raises the probability that the next cut is in the right plane. No single arm position is correct. Some make the next move more likely to be right.
Every move should either advance the operation or change your estimate. A move that does neither only adds harm.
Five rules follow.
- Know what a move is supposed to show before you make it.
- If it did not change what you believe, do not repeat it. Change the angle, the traction, or the instrument.
- Before anything irreversible, such as dividing a vessel, firing a stapler, or removing an organ, the evidence has to be in already. Name the structure from its attachments, not from the plan. Does the renal artery actually go to the kidney? Surgeons have divided the superior mesenteric artery believing it was the left renal artery.
- When the field surprises you, widen the list of what could be true. Do not narrow it.
- When you hear yourself explaining away an observation, stop and ask yourself: am I trying to make myself feel better, or is this a reasonable explanation of what is happening? What else could be true?
4. A Prior of One
Surgeons say it about themselves with a half smile: sometimes wrong, never in doubt.
Never in doubt is a prior of one. Under Bayes' theorem, a hypothesis held with a probability of one cannot be lowered by any evidence. Statisticians call the rule against this Cromwell's rule: reserve probabilities of exactly one and zero for statements that are true or false by logic alone.
In The Wicked Problem of Surgical Failure, I argued that good surgical judgment is partly well-calibrated uncertainty. Never in doubt is uncertainty set to zero.
No surgeon holds a probability of exactly one. The working version is simpler: the alternative never makes the list. With one hypothesis, every finding gets explained by it. Bleeding does not lower the estimate; it becomes a reason to finish. An organ on the wrong side does not change the certainty. The organ migrated. An organ four times too large does not lower it; it is diseased. Bleeding from an unexpected vessel does not lower it; it was an aneurysm. The last three are the explanations given afterward in this case.
By Shaknovsky's own account in a deposition for the civil suit, as reported by NBC News in May 2026, Bryan's heart stopped, and Shaknovsky searched for the source of the bleeding while the team did chest compressions. Shaknovsky said he had learned the difference between a liver and a spleen in junior high. He also said he could not tell the difference in the room because he was so upset. If that is true, knowing the difference was not the problem. Seeing it was. That is panic, not calm conviction. It ends in the same place. Calm certainty and panic both take the alternatives off the list, and with one name left, every finding gets fitted to it.
The order reads the label and the operative note as deception. The family's attorney says depositions in the civil case show that Shaknovsky examined the specimen after the operation and still called it a spleen. The record does not settle which, and no one outside that room knows what he believed. What the record does show is the frame: a named object, a field that disagreed, an operation that did not stop, and explanations afterward that kept the name.
Under that kind of certainty, staff who can see a liver stop being evidence. They become noise. Aviation has a name for this: plan continuation bias, the tendency to keep flying the original plan after conditions have changed. Once the object is named and the obligation to finish is set, each new cue gets fitted to the plan.
Certainty takes away the body's veto. Once the organ is certainly the spleen, and the spleen certainly has to come out, nothing the field shows can stop the operation.
Doubt and hesitation are different things. Packing a bleeding field is a fast, decisive move made by a surgeon who does not yet know what is bleeding. You can act at full speed on an estimate below one.
5. Highest Reliability
Brian Christian opens the chapter of The Alignment Problem titled "Uncertainty" with a night in 1983. At Serpukhov-15, the Soviet Oko early-warning system reported one American ICBM launch, then four more, at its highest reliability rating. The duty officer, Stanislav Petrov, was expected to pass the warning up the chain. He reported a false alarm instead. A real first strike, as he understood it, would not come as five missiles. Ground radar showed nothing, and visual checks could not confirm it. The satellites had mistaken sunlight on high clouds for rocket launches.
The story is usually told as a rescue, which overstates it. Petrov did not hold launch authority, and accounts of the night differ in their details. What the record supports is smaller and more useful. The machine could name a category and attach a confidence. It could not say it did not know. The Oko system was sometimes wrong, never in doubt. Petrov kept his own estimate below one and asked the question the machine could not: would an attack look like this?
6. A Prior of Zero
The rule cuts both ways. A prior of zero is as closed as a prior of one. If you are certain something cannot be true, no evidence can raise it either.
In revision reconstruction the prior of zero has a name: inoperable. Patients arrive with it, along with "hostile abdomen" and "nothing more to do." Strictly, "inoperable" is a decision, not a probability, but it behaves like a prior of zero on whether a workable plane exists. It is a conclusion that reads like a finding, sometimes drawn from a note and a scan of a body that has since scarred, leaked, and moved. A ureter left on the psoas may now sit in the wall of the bowel. The plane on the CT may not exist in the pelvis.
"Inoperable" is often right. It is also sometimes a verdict nobody tested, because the only test is an incision. To move it, someone has to be willing to start: open, look, and find out whether this is doable. That first look is a move like any other. Its job is information. Write the rules down before the incision: what I need to see to keep going, and what will make me stop.
Some of those looks end in an injury that would not have happened if no one had looked. That injury belongs to the surgeon who chose to look. A verdict of inoperability that was never checked has a cost too, and the patient pays it. The two costs are not the same kind, and Bayes does not say which is larger. That is why the rules go on paper before the incision.
7. The Play That Sets Up the Play
I like basketball, and I like the film breakdowns more: players explaining the thinking behind a play. As in surgery, setting up the play has as much to do with success as the play itself.
The clearest example I know is Game 6 of the 2013 Finals. Miami trailed San Antonio by three in the final seconds. LeBron missed a three. Chris Bosh got the offensive rebound and passed out to Ray Allen, who was backpedaling to the right corner. Allen tied the game with 5.2 seconds left, and Miami won in overtime.
The shot is the highlight. The setup is the lesson. Early in his career Allen built a drill for exactly this: lie down in the middle of the lane, spring up, backpedal to the corner, catch a pass, shoot. After the game he said he worked on it often. He had broken one decisive moment into components and rehearsed each until the position that made the shot most likely was automatic. Rehearsal belongs in practice, not on the tissue.
A good operation looks like this from the inside. The goal is to raise the probability of one outcome for the patient. Every position, every retraction, every repositioning of an arm is chosen because it makes the next move more likely to work.
What is the best move to make an operation possible? Unglamorous retraction. What was the best move to win Game 6? The backpedal to the corner.
8. Teach the Moves
Surgical teaching mostly gives trainees two things: a list of steps and an instruction to get experience. Experience of what? Measured how? Logged hours are not the same as the right practice. A list of how Dr. So-and-so does an operation does not help unless you know why each move was made. The tactics of surgery, the individual moves and the reasoning behind them, are the least discussed part of it.
Uncut video helps, up to a point. If you could do exactly what I do, you would get close. But your patients would have to be exactly the same as mine, and they are not. Watching a thousand full cases with outcomes attached teaches less than it seems, because you do not know what you are watching for.
Video editing is now easy enough to do this differently. Take a full case and break it apart. Watch the whole case. Watch a summary of what I did. Watch each part. Watch each move, how it led to the next, and the decision that came before it. Link them, so a trainee stuck at a point can ask the question that matters: which move gets me out of here? That is closer to what goes on in my head while I operate than any list of steps, and it is how I want to teach.
Here is one stuck point, broken down.
I am often called into another operating room to find a ureter in the middle of a pelvic dissection, and I see the same stuck point in trainees. The usual move is to look at a tube, look again, and say, "That's the ureter. That's got to be the ureter." Each look is the same test, so the estimate never moves. What rises is the conviction. I have seen the internal iliac artery and the gonadal vein taken for the ureter this way. In a scarred pelvis the obturator nerve is another candidate.
The ureter has competitors, and each move is built to rule one out.
- Start at the iliac vessels, where the ureter crosses them. Lower in the pelvis, the internal iliac artery can look the same.
- Check for a pulse. A tube that pulses is usually an artery, but an artery can transmit its pulse to a ureter lying against it. A pulse moves the estimate. It does not settle it.
- Watch for peristalsis. Seeing it points to the ureter. Not seeing it means less, because a scarred ureter may not peristalse.
- Trace it backward, up and over the iliacs. The ureter crosses over them. The gonadal vein does not, but it runs close. Dissecting above the iliacs and looking for a tube, you can take it for the ureter, and fibrosis can pull it to where you expect the ureter to be. The internal iliac artery does not cross them; it branches from them. The obturator nerve runs deep to them, not over them.
- Trace it forward. The ureter heads toward the bladder. The gonadal vein does not: in a man it turns laterally toward the inguinal canal, and in a woman it runs in the suspensory ligament to the ovary. The obturator nerve runs along the sidewall to the obturator foramen. If the tube does not reach the bladder, it may already have been cut.
In a scarred field, that may not be enough. Then I stop looking and test. I pass a wire, a stent, or a ureteroscope up from the bladder, or, when the system is dilated, I put in a needle and aspirate. The needle is a puncture, and its answer is not clean: urine can be bloody, and clear fluid can be lymph. Clear fluid from a dilated tube raises the estimate. It does not settle it.
Each of those moves can change what I believe. Repeating the same look cannot.
That is what a trainee should copy: the components, and the decision before each one.
9. What Else Could Be True
Every cut is a prediction. Each is paid for in harm, and each should earn its cost by moving the patient toward the outcome or moving your estimate toward the truth. Sure it is the spleen, you keep cutting after the field has said no. Sure it is inoperable, you never make the incision that would find out.
The question that guards against both is short:
What else could be true?
Sources
- Brian Christian. The Alignment Problem: Machine Learning and Human Values. New York: W.W. Norton, 2020. Chapter 9, "Uncertainty."
- The 26 September 1983 Serpukhov-15 / Oko incident: History.com; Chatham House, nuclear decision-making case studies; Wikipedia, 1983 Soviet nuclear false alarm incident.
- Florida Department of Health, emergency suspension order, Thomas Shaknovsky, 24 September 2024, as reported by Becker's ASC Review, Get the Coast, Law&Crime, and WSB-TV. Allegations in a disciplinary record.
- Beverly Bryan, civil complaint against Shaknovsky, his former practice, and Ascension Sacred Heart Emerald Coast: complaint (PDF); coverage by HealthExec and The Washington Post. Allegations, not findings.
- Alabama Board of Medical Examiners filing (the May 2023 pancreas / adrenal case and the July 2023 bowel case), as reported by NBC News. Allegations in a disciplinary record.
- Shaknovsky deposition in the civil suit and his not-guilty plea: NBC News, 7 May 2026.
- The family attorney's account of the depositions and the post-operative specimen examination: 1819 News, 13 March 2026.
- Grand jury indictment for second-degree manslaughter, April 2026: Walton County Sheriff's Office. Allegations, not findings of this essay.
- Cromwell's rule, named by the statistician Dennis Lindley.
- SKYbrary, "Plan Continuation Bias".
- Sayegh AS, et al. Superior Mesenteric Artery Injury during Robotic Radical Nephrectomy: Scenarios and Management Strategies. J Clin Med. 2023.
- Ray Allen's backpedal drill: NBA.com. Allen's postgame comments: Sports Illustrated, 19 June 2013.
- Lee C. Zhao. The Wicked Problem of Surgical Failure. 10 February 2026.