Tyler Loftus, M.D., was awarded an NIH K23 grant to explore machine learning as a means of augmenting personalized, patient-centered decision making in surgery.
Currently, around 15 million inpatient surgeries are performed every year across the United States. Complications can increase costs as much as $11,000 per major complication.
Once a patient has undergone surgery, their risk of critical illness and death ought to be commensurate to the frequency of vital sign and laboratory measurements (i.e., ‘intensity of care’), as well as where they are treated.
In order to assign the postoperative patient to the appropriate area and subsequent level of care, surgeons depend on time-consuming manual review of health records, decision-support tools that ignore intraoperative physiologic changes, delayed provider responses to changes in patient acuity, and often biased, error-prone decision-making.
“These problems are difficult to address systematically because there is no validated, unifying definition of intensity of care,” Loftus said. “There is a critical need to define intensity of care decision spaces and generate precise, autonomous tools that can support the alignment of patients’ acuity with intensity of care.”
The grant’s aims will: Develop and validate postoperative intensity of care definitions ; develop and validate interpretable, actionable acuity assessments that elucidate decision spaces, ; and identify and predict postoperative intensity of care phenotypes.
“My goal is to become an independent surgeon-scientist with expertise in design and implementation of machine learning systems to improve how we make decisions in the clinical space,” Loftus said.