Complex spine surgery has become extraordinarily precise about anatomy.
Surgeons can measure alignment, model sagittal balance and plan where a reconstruction should begin and end. Yet some of the factors that ultimately determine whether a patient does well may never appear on an X-ray or MRI.
That is where Han Jo Kim, MD, believes the next phase of patient-specific spine care is heading.
AI and machine learning are already beginning to help surgeons think through patient selection, fusion levels and alignment goals in complex reconstruction, Dr. Kim, an attending spine surgeon and director of the Complex Spine and Scoliosis Center at Hospital for Special Surgery in New York City, told Becker’s. But the bigger opportunity is combining those anatomical decisions with information about the individual patient, potentially extending all the way to genetics.
“The other part is determining whether there are patient-specific factors, beyond what we can see on an X-ray or MRI, that could contribute to a poor outcome, and how we can mitigate those risks,” he said.
For Dr. Kim, that is the promise of precision medicine in spine: not simply designing a more personalized operation, but identifying which risks matter most for an individual patient and intervening before those risks become complications.
The catch is equally important. The closer medicine gets to understanding the individual patient, the more it encounters factors that may be difficult, or impossible, for an algorithm to quantify.
Precision care could eventually begin with a blood sample
At HSS, Dr. Kim and his colleagues are working toward an effort that would use patients’ genetic information to better understand complication risk in complex spine surgery.
The concept is still prospective: obtain a blood sample, analyze genetic characteristics and study whether certain profiles are associated with substantially higher risks for specific complications.
If those relationships can be established, care could become far more targeted.
Dr. Kim pointed to blood clots as an example. Anticoagulation in spine surgery requires a careful balance because reducing the risk of thrombosis can increase concern for postoperative bleeding and hematoma. Today, those decisions are largely made case by case using the clinical information available.
A validated genetic risk profile could eventually add another layer. If one patient were shown to have a dramatically higher predisposition to blood clots than another, Dr. Kim said, clinicians could potentially tailor preventive therapy accordingly rather than treating both patients under the same broad protocol.
“If we know that that risk is much higher than the general population, then we would be able to perform an intervention to mitigate that risk,” he said.
His team also hopes to explore genetic associations with complications including infection and postoperative ileus.
Ileus illustrates why individualized risk matters. Patients can respond very differently to the same medications and perioperative stressors. If researchers can identify biological differences that make certain patients more susceptible, Dr. Kim envisions intervening earlier rather than waiting for the complication to develop.
The broader shift is away from what he described as an “umbrella” approach to treatment.
“Imagine if we’re able to give specific medications to just some people instead of everybody,” he said. “The cost of delivering that care will be much less, and the outcome of successfully treating that problem will be optimized.”
That would expand the definition of patient-specific spine surgery. The operation could be tailored to the patient’s anatomy. The perioperative strategy could be tailored to the patient’s biology.
AI could help decide where complex surgery belongs
Not every complex spine patient requires the same hospital resources. Some operations may be safely performed in less specialized settings. Others may require an ICU, multidisciplinary support and the infrastructure of a tertiary referral center.
Dr. Kim sees that as one area where AI may be particularly useful. An experienced surgeon can consider those variables individually. AI can potentially analyze far larger combinations of patient characteristics, procedural complexity and resource needs simultaneously.
“I think AI, with its ability to analyze large datasets and process multiple factors at once, will allow us to determine which resource environments are optimal for specific patients,” he said.
That could make patient-specific decision-making about more than the procedure itself. It could help determine the right operation, the right preparation and the right place to perform it.
But Dr. Kim sees a much harder challenge once AI moves from analyzing systems and populations to trying to reproduce the judgment of an individual surgeon.
Complex spine surgery still has an artistic element
There may be several technically sound ways to treat the same complex deformity. That creates a problem for algorithms built to identify an optimal answer.
Dr. Kim compares complex spine surgery to art. A Picasso, a Van Gogh and a Rothko may look nothing alike. They use different techniques and approaches, yet each can arrive at a successful result.
Experienced spine surgeons can function similarly. Two surgeons may make different choices about levels, alignment, approach or technique and still achieve excellent outcomes.
“Whether you have a Picasso or a Van Gogh or a Rothko, they’re all going to be beautiful art pieces,” Dr. Kim said. “But they all use different mediums. They all have different techniques.”
The difficulty for AI is not simply determining whether an outcome was good. It is reverse-engineering all the tacit judgment, technical experience and individual decision-making that produced it.
Dr. Kim believes that level of variability remains difficult to quantify with current AI. And surgeon technique is not the only variable that resists measurement. The patient may be even harder.
The patient remains AI’s hardest variable
Experienced surgeons gather information before they ever open the chart. They notice who accompanies a patient to the appointment. They get a sense of the support waiting at home. They hear how the patient talks about recovery, how engaged they are in the plan and whether they seem prepared to participate actively in rehabilitation.
Those observations can matter after a major operation. They are also difficult to convert into structured data.
“You could really get a sense of how motivated a patient is to do better, and that’s a big factor,” Dr. Kim said. “If they are motivated, they’re more likely to do better than a patient who has a very passive way of considering their spine care.”
Questionnaires can capture portions of that picture. Dr. Kim does not believe they necessarily capture the whole person.
He compared it to giving equally intelligent people tests in different subjects. Even with the same measured IQ, their performance could vary significantly because a single objective measure cannot capture everything that influences the outcome.
Medicine faces the same problem. An algorithm can know many things about a patient without truly knowing how that person will behave. That led Dr. Kim to an even more fundamental limitation: free will.
He offered a simple example. An algorithm could analyze a lifetime of behavior and correctly conclude that someone prefers chocolate milk. That person could still decide, at that moment, to order iced coffee.
No amount of historical data eliminates the ability to make a different choice. For AI to understand a patient with far greater fidelity, Dr. Kim said, it might theoretically need continuous information about behavior, conversations and thought processes, a prospect that immediately raises larger questions about privacy and data ownership.
That uncertainty is not necessarily a failure of AI. It is a reminder of what clinical judgment is trying to interpret.
The goal may be less variability — not identical surgery
Dr. Kim does not expect the future of complex spine surgery to mean every surgeon performs the same operation. A more meaningful sign of progress would be making outcomes more predictable.
Today, similar patients treated in different environments can experience substantially different results. AI and richer patient-specific data could narrow that range.
“I think what will happen, and I think is a good thing, is that we’ll see much less variability in outcomes for these patients,” Dr. Kim said.
That reframes what success looks like. AI does not necessarily have to discover one perfect surgical technique. It could improve patient selection, refine alignment and level decisions, flag individualized complication risks and help match patients with the resources their operations require. If those tools make excellent outcomes more consistent, the impact reaches beyond clinical quality.
“With decreasing variability, you can improve cost, you can improve value, and because of that, quality will be better,” he said.
AI may also change what patients bring into the exam room
There is another part of the patient experience AI is already changing: information. Patients have long arrived in spine clinics after searching their diagnoses online.
Generative AI has raised the stakes. People can now ask tools to interpret imaging reports, explain conditions and suggest possible treatments within seconds. Dr. Kim sees obvious potential for misinformation, but he does not believe the answer is to dismiss the technology.
“I think AI has an opportunity to be a good educational tool for our patients if it’s used the right way,” he said.
The field has not yet determined what that ideal use looks like. But patient education fits naturally into the same precision-care framework. A patient who understands the disease, treatment options and recovery expectations may be better equipped to make decisions and participate in care.
“The better a patient is informed about their condition, the more likely they’re going to have a great outcome,” Dr. Kim said.
That may ultimately define the tension around AI in complex spine surgery.
Technology is becoming increasingly capable of seeing what surgeons cannot: patterns across massive datasets, subtle risk relationships and potentially even biological predispositions hidden in a patient’s DNA. But some of the variables that matter most may remain stubbornly human: motivation, support, judgment, behavior and choice.
The future of precision spine surgery may therefore depend less on asking AI to make the perfect decision.
It may depend on using it to reduce what surgeons do not know, while preserving the judgment required for everything that cannot be reduced to data.
At the Becker’s 32nd Annual Meeting: The Business and Operations of ASCs, taking place October 29-31 in Chicago, ASC leaders, surgeons and healthcare executives will explore strategies to drive growth, enhance operational performance, navigate reimbursement challenges and prepare for the future of ambulatory surgery. Apply for complimentary registration now.
