Why scaling AI is a prerequisite for success

There is definitely a lot going on in hospitals in the field of AI development; many of the applications are related to radiology. "There is a realization that AI can offer solutions to many current problems in healthcare," says Mark van Buchem, professor and head of the radiology department at the LUMC. "But we do see that each hospital is reinventing its own wheel."

Early adopters in this field also face the problem of not being able to make a business case to recoup their investment in development. An AI tool can easily cost twenty to thirty thousand euros annually. And you have to spend that amount multiple times if you truly want to change your workflow with it. Van Buchem: "Besides that, the effective implementation of AI, both technically and interpersonally, requires a level of expertise that is not always available in hospitals."

The fact that hospitals have signed the Integral Care Agreement logically leads them to seek cooperation with each other in this area. "You do see initiatives emerging to join forces," Van Buchem continues. "It also explains why we are introducing the ReplicAId initiative. With this, we want to take up the challenge of bringing AI from PowerPoint to practice."

Clear Benefits

The initiators are convinced that AI can play a role in fulfilling the IZA's missions. "The application of AI in healthcare can offer significant advantages in many aspects," says Maurice Fransen, partner AI and data & future of health lead at Deloitte. "Think of administration and file creation. Also consider supporting logistical processes, for example, to predict whether patients will show up for their appointments based on algorithms. But ensure that you also think of support in the care process itself, with image assessment in radiology as a well-known example."

Van Buchem provides another compelling example. "A well-known complaint from both doctors and patients is that the doctor has to enter all the information that the patient shares during the consultation into the EHR. To do this, they must look at their screen continuously, leaving little opportunity to make eye contact with the patient. Yet, visual information is also important to get a full understanding of the patient. As a result, patients rightly feel overlooked. That is why we are working on the digital scribe project in our hospital, where the relevant information from the conversation between doctor and patient automatically ends up in the correct place in the EHR."

Under the name ReplicAId, Deloitte and the Leiden University Medical Centre (LUMC) aim to collaborate with market parties to achieve rapid and targeted scaling of AI initiatives developed in hospitals. The premise is that AI is needed to meet the biggest challenges in healthcare, and that scaling is the only way to make AI profitable and truly labor-saving.

He emphasizes the importance of the quadruple aim: cost, quality, patient experience, and job satisfaction. "The application of AI can be beneficial in all four of these areas. Automating tasks reduces costs. Predicting which patients will and will not benefit from expensive drugs - a very relevant topic in oncology - improves patient outcomes and thus increases the quality of care. A doctor who looks at the patient instead of their screen enhances the patient experience. And automating routine jobs increases the job satisfaction of healthcare professionals."

Scaling Up

With ReplicAId, the goal is to fully utilize opportunities to achieve the quadruple aim in AI application. As mentioned, many hospitals are already developing AI applications, but scaling these up to other hospitals is virtually non-existent. The true value of AI is realized through its scalability.

Every hospital encounters implementation challenges and issues related to laws and regulations when they start working with AI, and these challenges multiply when scaling applications to other hospitals. "With ReplicAId, we aim to facilitate scaling and share lessons learned in this area. Additionally, we will collaborate with research institutes to test all developments in practice. Especially when scaling up, there is much to consider. That's why it hasn't happened yet; there is no business model for each individual hospital."

Why not use technology companies that have the capability to do this? "That does come at a cost, and moreover, many beautiful solutions have already been developed by Dutch hospitals." An alternative is to utilize what startups develop. "But then you end up with point solutions," Van Buchem adds. "Moreover, you have to integrate those applications into your workflow. That's where things often go wrong."

Role of Health Insurers

One of the strengths of ReplicAId is that it has health insurance companies on board. "This is crucial, firstly to be eligible for the IZA transformation funds for which we intend to apply," says Van Buchem. "But it's also important because many applications require an investment that only later yields a benefit for the healthcare system in the patient care process. To finance this, we really need the health insurers."

There is also momentum, emphasizes Van Buchem, and not just among the health insurers. "The Covid period has given healthcare providers a sense of urgency. When we emerged from it, it was as if we had stepped out of a time capsule and suddenly saw the looming healthcare crisis very clearly for the first time. So, this is the right time for ReplicAId; a cooperative initiative of this kind has not existed before. We are starting with a number of hospitals, both top clinical and general. The NFU, Santeon, mProve, and SAZ networks are backing this."

Fransen adds: "These are the external parties we will initially be collaborating with outside of hospitals. We have already had discussions with these networks and are receiving enthusiastic responses everywhere. Everyone understands it. And the fact that it is by and for healthcare, i.e., non-profit, also helps. Just as the fact that having access to IZA transformation resources really hinges on collaboration."

Organizational Capability

For Van Buchem, having Deloitte involved in ReplicAId is of great value. "As hospitals, we lack the organizational capability needed to achieve ReplicAId's goals," he says. "Deloitte does have that, along with a wealth of knowledge that is valuable to our ambitions."

For Deloitte, involvement in this initiative is also important, Fransen emphasizes. "We truly believe that the acceleration advocated by ReplicAId is necessary and that it has taken the right approach. We also see a social role for ourselves in helping to prevent the impending healthcare crisis."

Van Buchem views the ReplicAid approach as a pipeline. "The translation of AI into practice can be seen as a series of hurdles in areas such as ethics, law, technology, and change management. We create knowledge groups for every hurdle. Within the LUMC, AI activities are bundled in the CAIRELab, which is affiliated with SAILS, Leiden University's interfaculty AI knowledge network. We aim to expand this knowledge network with other academic partners. In this network, we mobilize and bundle the knowledge with which we hope to overcome the various hurdles in the implementation process of AI in healthcare. Perhaps with the insights gained from ReplicAid, we can even propose new ideas for laws and regulations regarding the application of AI in healthcare."

This article was previously published on ICT&health in April 2024.