Leading up to the AIMed ICU (Intensive Care Unit) virtual event taking place on 22 September, AIMed hosted a webinar this July to explore some of the key trends of using artificial intelligence (AI) in critical care. A subsequent webinar on overcoming the barriers to deployment of AI in the ICU was held yesterday (2 September).

Dr. Matthieu Komorowski, Consultant in Intensive Care at Charing Cross Hospital and Senior Lecturer at the Imperial College London and Dr. Bart Geerts, Consultant and Cardiac Anesthetist at the Academic Medical Centre, University of Amsterdam were invited to discuss with Dr. Anthony Chang, AIMed founder and Chief Artificial Intelligence Officer at Children’s Hospital of Orange County (CHOC) on some of the challenges that are hindering ICU from adopting AI.

Absent of infrastructure to leverage real-time data and sharing of information

Dr. Komorowski said leveraging real-time data for clinical decision making is still not happening at the moment. What they are doing now is to collect and analyze data from the very first cohort of patients and used them to make informed decisions for future patients. While it may not be ideal, this nearly real-time approach is still much faster than traditional methods. Furthermore, the severity of the ongoing COVID-19 pandemic allowed physicians to bypass some of the protocols and push for more innovations or modify existing tools.

Dr. Geerts agreed. He added the pandemic has also speed up the adoption or creation of certain infrastructure. For example, the institution where he works right now is using Cloud facilities to merge databases. Dr. Geerts also mentioned his colleagues’ work to connect ICU across 50 hospitals, a process which normally would take two years to complete but now, it was done within six weeks. Nevertheless, both speakers feel the need to share certain life-saving strategies or information more quickly. On the other hand, Dr. Komorowski regards online discussions conducted by the European Society of Intensive Care Medicine (ESICM) as a way to pick up new knowledge quickly.

He explained once in while the Society would invite prominent practitioners or researchers to share their clinical experiences or expertise on a particular topic so that new findings can be disseminated across a large group of audience at once. However, Dr. Geerts cautioned the potential backlash coming from insufficiently peer-reviewed results. “There were some pretty poor trials with some results that we took on as a kind of definite proof that wouldn’t be there at all when you have a proper review,” he says.

Governance and cultural shift as the two main barriers

Thus, balancing innovation and factchecking becomes a key especially in this war against COVID-19.  As such, the speakers imagined a centralized depository to be in place in the coming decades. One which contains highly granular and patient level data and permits international level of data sharing, so that anyone and everyone can leverage it to develop interventions. Clinicians will also be able to compare a patient with others on similarity metrices and figure out what is or is not working for present and future patients.

Yet, in reality, Dr. Geerts is afraid something like governance may hinder full deployment of AI in the ICU. “We have a lot of regulations coming at us as governance is picking up the pace and trying to bridge the gaps in technology” but since AI has the potential to be “live”; to keep monitoring and improving on its own, “we do not have the experience to govern something like this”. Dr. Komorowski believes there should be better ways to translate what has been found in the research setting into the clinical field.

“Most of us don’t have the expertise and sometimes, no interest in doing AI. Even if we like a model that has been published, there is a missing link in that circle of development and proved that the new tool is indeed beneficial. I think what we need is the right environment, something that nurtures innovation; giving the right support to the right people,” Dr. Komorowski says. “So, I think, in my opinion, the biggest challenge is a cultural change. I think we need evident to prove AI is working so everyone in the ICU will go for it because it works”.

The webinar is available on demand here. If you are interested in how technology, particularly, AI, robotics, virtual/augmented/mixed realities, and many others are influencing ICU or will like to join in the discussions to explore how new technologies can be better introduce or deploy in the ICU setting, do not miss the upcoming AIMed ICU virtual event, taking place on 22 September. Register your interest or get a copy of the agenda here today!


Author Bio

Hazel Tang A science writer with data background and an interest in the current affair, culture, and arts; a no-med from an (almost) all-med family. Follow on Twitter.