Reducing Unnecessary ED Utilization with Artificial Intelligence and Machine Learning
It has been estimated that up to one third of all emergency department (ED) visits are not emergencies. SWHR collaborated with ClosedLoop to create an artificial intelligence/machine learning (AI/ML) model that identified individuals at high risk for costly, preventable events, then intervened to keep them out of the emergency room. The results were encouraging.
The problem
The negative impact of ED overuse is significant. Healthcare costs soar even as the quality of care declines. Crowded emergency rooms and long waits to see a physician interfere with the needs of those who truly require urgent attention. At the same time, the emergency department is unable to provide the consistency of care which a primary care provider can achieve, resulting in an overall diminishment of health outcomes for frequent ED users.
The opportunity
Between 2021 and the launch of ED Diversion efforts in 2022, Southwestern Health Resources saw a decline of 5.20% in costs per member per month, due to improved utilization. Utilization of the emergency department in 2021 averaged 48.4 visits per thousand. After interventions and education launched in 2022, that number declined to 44.7 visits per thousand, an improvement of 5.40%, demonstrating that SWHR is making progress to ensure the right care at the right time and in the right channel of care. It is reasonable to believe these percentages will continue to improve as SWHR providers and patients grow in their understanding and appropriate utilization of the primary care provider and urgent care centers for non-emergent needs.
The solution
SWHR built and deployed an AI/ML model to predict ED high utilizers (EDHUs). The model predicts an individual’s risk of experiencing three or more ED visits in the next six months and explains each individual’s unique risk factors.
With a goal of delivering the right care at the right time in the right place, ED diversions included reductions in visits due to urinary tract infection or general respiratory care, headaches and abdominal pain, among other ailments. Redirecting patients to their primary care physician or urgent care center created a more positive experience for the patient while reducing total healthcare costs.
The results
The AI/ ML model was able to identify 21.8% of all Emergency Department High Utilizers (EDHU) in the riskiest 1% of the SWHR Medicare Advantage population as sorted by the model. While this represents a 30-fold improvement over randomly targeting 1% of the population, continual refinement of the model will provide an opportunity for even greater impact.
Within the first six months, using this model with SWHR Medicare Advantage plan patients reduced avoidable ED utilization by more than the expected 2%, resulting in $2.5 million in savings that could be reallocated for appropriate care.
