Predictive Analytics Tool Forecasts Post-Surgical Complications

Researchers at Thomas Jefferson University have developed a predictive analytics tool that can forecast the risk of post-surgical complications, helping providers to deliver more proactive, preventive care.

“We need to be able to assess the risk of life-threatening, post-surgical complications so we can then come up with individualized ways to reduce those complications,” said Dr. Sang Woo, a clinical associate professor of medicine at Thomas Jefferson University who led the new research.

Patients are at high risk of experiencing life-threatening complications after surgery, with some individuals suffering from kidney failure, stroke, and heart attacks.

“Seeing how much suffering those patients had gone through, I wanted to figure out what we could have done differently to prevent these life-threatening complications,” Woo said.

The risk calculators providers currently use mainly evaluate patients for cardiac risks, like cardiac arrest or heart attack. These tools don’t provide risk assessment of other major complications like stroke, and doctors haven’t paid much attention to risk assessment for kidney failure.

“We wanted to assist doctors to be able to assess the risk of stroke, in addition to traditional risks,” Woo said.

Researchers leveraged predictive analytics tools to develop an accurate, easy-to-use risk assessment model for clinicians to use. The group collaborated with a multidisciplinary team including a surgeon, cardiologists, nephrologists, and hospitalists.

“Often times we do the research and publish a research paper that is too complex to translate to the bedside,” Woo said. “My goal from the beginning was to come up with a new model that is very practical and useful and that can be incorporated into routine patient care.”

In two recent studies, the team showed that the model can accurately predict the likelihood of post-surgical, life-threatening complications.

In one study published in December 2020, researchers designed a model to assess a patient’s risk of developing acute kidney injury (AKI) following surgery. AKI after surgery is associated with high mortality and morbidity – more than a third of patients who required dialysis after cardiac surgery died.

“Identifying patients at high risk for AKI and implementing preventive measures may lower that mortality risk,” Woo said.

Researchers analyzed data from more than 2.2 million surgical patients, of whom about 7,000 developed AKI requiring dialysis. The results showed that patients who required dialysis were older and more likely to have congestive heart failure and diabetes.

The team then trained the predictive model with data from over 1.4 million patients using these and eight other predictors before testing it out on data from another set of more than 800,000 surgical patients. The group found that the model was able to accurately predict which patients would go on to develop AKI.

In a separate study published in the Journal of the American Heart Association in January 2021, researchers used the model to predict the risk of stroke, cardiac event, or death within 30 days after surgery.

For this version of the model, researchers analyzed data from more than 1.1 million surgical patients and used predictors like age, history of stroke, type of surgery, and other health factors that could be measured before surgery to build the model.

The model could predict which patient would suffer a stroke, cardiac event, or die within 30 days of surgery with high accuracy. The model also performed with high predictive power, achieving an area under the curve (AUC) of 0.87 for stroke and 0.92 for mortality.

The team also found that the model can predict cardiac risk with an AUC of 0.87, an accuracy that is similar to or better than widely used cardiac risk models.

The model is also easy to use as a web-based tool, the team noted. Providers conducting pre-surgery assessments can use the tool at a patient’s bedside. With the model, clinicians can inform surgeons of potential risks and better advise patients, leading to improved care delivery.

“Now that we have a tool to assess stroke and kidney failure risk objectively, we are investigating novel ways to reduce that risk,” said Woo.