Feature: Diaverum develops AI model to help prevent vascular access thrombosis

Dimitris Moulavasilis chief executive, Diaverum

Global renal care service provider, Diaverum, is set to roll out an artificial intelligence (AI) model to support the company’s physicians and nurses predict episodes of vascular access (VA) thrombosis among haemodialysed patients. Simulations using the AI model, developed in-house using Diaverum’s historical patient data, predicted an average of 75% of actual cases that were not detected by best-in-class clinical assessments and monitoring systems.

VA thrombosis is a major factor of suffering and increased mortality for chronic haemodialysis patients, with incidence ranging between 0.11 to 0.5 episodes per patient every year, according to Diaverum. It accounts for a significant increase of the total cost of care for these patients, contributing to the unsustainable trajectory of growing expenditure and disease burden impacting national healthcare systems around the world.

“VA management is central to our clinical strategy,” said Dimitris Moulavasilis, Diaverum’s CEO, who sees the AI model as a breakthrough innovation that will deliver a step-change in the ability to predict and take preventive measures to avoid VA thrombosis episodes in dialysis patients.

“It’s a big win for everyone,” he said. “Better digital tools and predictive analytics insights for our healthcare professionals; improved medical outcomes for our patients; and lower cost of care for payors and national health systems.”

The model is the result of a three-year journey in which Diaverum has been converting its know-how into a proprietary digital platform, aiming to provide the most efficient, standardised, high-quality and scalable dialysis service in the industry, according to Zoltan Szepesi, Diaverum’s Chief Transformation Officer. It was developed in house by a newly established AI team working to best practices of Responsible AI.

Validation of the model was performed with data selected from three countries, beginning with Portugal and tested across data gathered from Spain and Saudi Arabia. It is in these three countries, in which Diaverum cares for approximately 12,000 renal patients, where the VA AI model will initially be rolled out; however it plans to expand its use across the company’s entire network of 452 clinics in 24 countries over the next 18 months.

Diaverum is not stopping at a VA AI model but is building its own AI Development Factory to train, validate, deploy and monitor a series of AI models that will address important unmet clinical needs.

“Our ambition is to have an AI-empowered clinical workforce in all our clinics, improving outcomes that really matter to our patients,” said Szepesi.