Risk analysis is part of the way cardiothoracic surgeons determine which patients would benefit from surgery to replace a faulty heart valve, and artificial intelligence (AI) now gives them an advantage when weighing risk vs. benefit.
Dr. Robert Hagberg, Chief of Cardiac Surgery at Hartford Hospital and part of the Hartford HealthCare Heart & Vascular Institute, was part of a team of researchers that published “Machine learning models for mitral valve replacement: A comparative analysis with the Society of Thoracic Surgeons risk score” in a recent edition of the Journal of Cardiac Surgery. It’s among the first studies examining the use of machine learning, or AI, to predict surgical success.
The researchers mined an online database of mitral valve replacement cases and their outcomes, introducing several AI techniques to generate algorithms to evaluate patient results. They then compared the results to those generated by an existing predictive tool made available by the Society of Thoracic Surgeons (STS).
“The proposed risk models complemented existing STS models in predicting mortality, prolonged ventilation and renal failure, allowing healthcare providers to more accurately assess a patient’s risk of morbidity and mortality when undergoing mitral valve surgery,” reads the conclusion section of the study publication.
The results were a 3 percent increase in the accurate prediction of surgical outcomes.
“The STS tool looks at things in a certain linear way and we speculated that we could do better,” Dr. Hagberg said. “The biggest thing we did was to use AI to tweak the models and make them more accurate. They can analyze data with different predictive models to more accurately forecast mortality during and after mitral valve replacement surgery.”
Sicker patients, he said, are at higher risk for such complications as stroke, prolonged ventilation, renal failure and death after mitral valve replacement surgery. For those reasons, he said they should, perhaps, reconsider the decision to have surgery.
“This can help us guide patients when they are trying to decide whether to do the surgery at all,” Dr. Hagberg said. “It helps us guide the patient better. Not everyone who needs a valve replacement should have the surgery. We need to weigh the risk of the operation to the benefits of it.”
Use of the new predictive models might eventually become standard work for surgeons, but Dr. Hagberg said at the very least the research showed “we can make improvements.”
“Maybe we can incorporate this into a huge database maintained by the Society,” he said, predicting there will be more validating research in the future.
