AIMed UK 2020 virtual summit took place early on. In the opening keynote session: Deployment of artificial intelligence (AI) in the UK and across the world, Professor Neil Sebire, Chief Research Information Officer at the Great Ormond Street Hospital for Children National Health Service (NHS) Foundation Trust talked about some of the considerations healthcare organization need to have as they plan to deploy AI tools at scale.
Think beyond the electronic health record
Professor Sebire said healthcare organization ought to think about what is required, in terms of infrastructure, when it comes to dealing with healthcare data. Often, it’s great to have talks focusing on electronic health records (EHRs) but these data warehouses do not facilitate utilization. What the healthcare system needs is a place which not only keeps all the data but also permits algorithm development; planning the deployment and scaling of AI, and everything else. Professor Sebire believes in the coming decade, all the data important to healthcare will not be in the EHRs, there is a need for additional infrastructure.
As such, the first realization healthcare leaders need to have is to find ways to manage data and analytics beyond a data lake or data store. True personalized medicine will require data coming from different areas, like imageries, wearables, home-based healthcare devices. This new architecture has to meet the demands of all these additional data. It is necessary to also think about “Trusted Research Environments” (TREs) or reliable spaces where researchers can securely access sensitive data with robust controls to ensure that only summary statistical results, and not patient level data, can be exported.
“I am amazed continuously with the different discussions I have. Whenever I ask, ‘tell me about the IT infrastructure in your organizations?’ Healthcare leaders will start talking about the bigger problems rather than doing AI. They need to realize we cannot do AI without proper infrastructure; regardless of whether you want to scale AI in different organizations in the UK or from the UK to the US. There is a big missing link here,” Professor Sebire comments. “There are starting to have some movement, possibly through FHIR (Fast Health Interoperability Resource) (i.e., standardize data within and between EHRs) to solve the issues but that is not the whole story. We need to recognize there is an increasing requirement for new infrastructure”.
Understand what AI can do
Furthermore, Professor Sebire noted the digital health space is huge and healthcare systems should distinguish between what they will like to do and what they do not wish to do with technologies. He had come across healthcare leaders who are too eager to think about adopting certain AI tools but failed to realize half of the organization is using pen and paper to keep their healthcare records. To prevent a mismatch in vision and reality, healthcare leaders will have to define a limit so that the scenario will not be “many people are excited about something but only a few can support it”.
In general, according to Professor Sebire, there are three things that AI can do at the moment and most of it is around clinical decision support or having tools that allow clinicians to do their job more efficiently but not allowing the machine to make the final decisions. “In the UK, we are nowhere near leveraging AI around decision support, it’s more about how to run the hospital because very few healthcare organizations are able to collect data granular enough to truly support AI-driven operational decision support”.
Finally, there is patient support, which most medical professionals will like to take on, yet again, the core questions that need to be addressed are “Do you have the data?”, “Do you have the data in the format or quality that is required, in a particular platform?”, “what processes are you going to develop and test the algorithms or tools?”. “Sometimes, there are fantastic models being built but when asked how do they fair in a hospital setting, some creators will reply that’s not really their problem. We can observe a real gap here,” Professor Sebire adds.
Ask “what if it works” at the very beginning of the AI journey
More importantly, Professor Sebire feels healthcare organizations did not do enough to ask the question “what if AI works?”. “Imagine a tool is brilliant, what would the impact be? Can you do anything about it? What are you going to do, who are the people to act on that?” Professor Sebire asks. “Even if AI works, we still need the infrastructure to deal with its impact”.
In summary, organizations have to start taking responsibilities for their data and infrastructure. The healthcare space will never be able to scale the value of AI if different organizations have different data systems. Next, healthcare organization need to know that we cannot do AI alone and we cannot do things from scratch. There is a need to partner with the industry. For example, computer vision, the healthcare realm should not think about re-creating something similar but to adopt what is already available.
Professor Sebire mentioned all the need for someone to oversee the whole process, specifically, a career structure for someone who is interested in the area and want to serve the healthcare system. “We have tried to source for these individuals within the organization; trained them to understand the healthcare system and data but very quickly, an industrial competitor will come in and offer them a higher salary and they leave. We will then go back to square one.”
Dr. Anthony Change, AIMed Founder and Chief AI Officer at Children’s Hospital of Orange County (CHOC) reassured hopefully as AI becomes part of the medical education and certification process, more people will become knowledgeable in the area and the younger generation will find it relatable and want to make this for a career.