For patients facing complex adult spinal deformity surgery, the traditional way surgeons describe risk has not changed much in decades. Christopher Ames, MD, director of spinal deformity and spine tumor surgery at San Francisco-based UCSF Health, said that approach has relied on a lot of guesswork, and that his team has been building something more precise.
“In fact, we’re lying to most of the patients, both good and bad, that we’re over- and underestimating their recovery because it’s not specific enough to them,” Dr. Ames said, referring to the industry-wide practice of relying on published averages — rather than a patient’s own data — to predict surgical outcomes. “It’s based on maybe a publication that the guy read on vacation by the pool a couple months ago.”
From risk scores to digital twins
Dr. Ames’ work began nearly 10 years ago with a charge from the Scoliosis Research Society to develop risk stratification tools for adult spinal deformity patients — a population with long lengths of stay, high complication rates, and outcomes that vary widely by center. His team built one of the first frailty indices for the condition, then moved to multivariable predictive models combining patient data, procedural data and pre-op disability scores.
But scaling those models revealed an issue: Adding thousands of more patients to the registry did not make the predictions any more accurate.
“We realized we weren’t collecting all the variables we needed,” Dr. Ames said.
That sent his team on what he called a “hidden variable hunt,” and it turned up factors that had not been widely considered: biological age (measured through telomeres, rather than chronological age), gait and load-bearing dynamics, body composition, and “digital phenotyping” data pulled from patients’ phones, such as activity levels and social engagement.
That process led somewhere unexpected for his team: Rather than simply refining a risk score, it ended up building the foundation for a digital twin — a data-driven virtual model of an individual patient that can be used to simulate and optimize a surgical plan before a surgeon ever touches the patient. Dr. Ames said the concept was borrowed from other fields, including aerospace and cardiology, in which digital twins are used to plan procedures such as atrial fibrillation ablations.
“This represents a tremendous step forward,” Dr. Ames said. “The computer might not be perfectly accurate, just like Tesla auto-driving. It might not be perfect, but it’s a hell of a lot better than a human driver just kind of making it up.”
He pointed to a basic limitation of human cognition as part of the rationale: Physicians can meaningfully process only a handful of variables at once, while an accurate prediction may depend on 50 to 100 factors specific to that patient.
Where the model fits
Dr. Ames was clear he sees this technology as suited to complex, heterogeneous cases.
“Digital twins are going to have the most impact in complex conditions,” he said. “Not grade one spondylolisthesis, not a vasectomy, not simple fusion.”
In practice, Dr. Ames described a pipeline extending beyond the simulation itself: Once a digital twin generates an optimal surgical plan, that plan can be sent to custom implant manufacturers to produce patient-specific rods and interbody devices. Intraoperative measurement tools then check the actual surgical execution against the plan, a step he calls “reconciliation.” Afterward, tracking patient outcomes over months and years feeds back into the model, refining its accuracy over time.
“It’s a cyclic feedback loop,” he said. “It’s part of every manufacturing process in the world. … Until very recently, nobody was doing that” in spine surgery.
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.
