An Artificial Intelligence-Based Clinical Decision Support Tool to Reduce Hyponatremia after Total Joint Arthroplasty

Although clinical outcomes after total joint arthroplasty (TJA) are generally positive and reproducible, certain medical and surgical complications are not insignificant and may negatively affect patient outcomes. Hyponatremia is an often overlooked and preventable electrolyte abnormality in patients undergoing TJA that may lead to adverse clinical consequences, including nausea, dizziness, seizures, and death. Incurring such complications may alter the trajectory of recovery after a routine TJA procedure, requiring additional interventions and prolonged hospital stay, and negatively impacting the value of health care rendered. The Hospital for Special Surgery in New York City is a high-volume, tertiary musculoskeletal care center that performs more than 43,000 orthopedic surgical procedures annually; to sustain this volume and positive hospital performance metrics, optimizing value per episode of care is essential. Therefore, the authors implemented an internal quality-improvement investigation utilizing digital implementation of an artificial intelligence (AI)-driven prediction model into the electronic medical record workflow to identify patients at an elevated risk of hyponatremia presenting for elective TJA between April 1, 2022, and March 31, 2023. This was transformed into a clinical decision support tool utilizing a best practice advisory alert on opening the patient chart, raising awareness for those involved in the episode of care. Among those identified as at risk, an intervention was initiated on behalf of the anesthesiologist of record that represented a deviation from standard of institutional care by changing fluid management from lactated Ringer's intravenous maintenance rate to a Multiple Electrolytes Injection, Type 1 solution, as well as by discontinuing medications with known associations to hyponatremia (such as duloxetine, hydrochlorothiazide, and nonsteroidal antiinflammatory medications). During the 1-year trial period, comprising a total of 22,271 consecutive TJA episodes of care, the authors observed an institutional reduction in the overall rate of hyponatremia of greater than 50% (from 29% to 14%). This study demonstrates the efficacy and feasibility of integrating a scalable AI-based digital solution into the clinical workflow to help augment clinical care through risk stratification and selective interventions.