The practical use of AI and machine learning are too often just buzzwords and cool phrases that catch the eye or ear but don’t really resonate. But there are organizations that are acting and using AI to deliver real-world outcomes.
In the healthcare industry, we are seeing huge strides with the usage of AI to create a patient-centric approach to personalized care. Recently, leaders from Duke, Mayo Clinic and UC Berkeley, along with a few others, will announce a new Health AI Partnership. This is truly a powerful combination of healthcare leaders coming together to create a path for AI partnerships, and to change how healthcare is understood and delivered.
To help me explore this further is Paul Swider. Paul is a fellow analyst for Acceleration Economy, a healthcare technology expert, and Chief Philanthropy Officer for RealActivity.
01:58 – There has been an increasing amount of AI adoption within the healthcare industry. For instance, leaders from Duke Mayo Clinic, UC Berkeley, and a few other partners recently announced a new health AI partnership to make innovative and impactful change.
03:35 – With this partnership, the group is aiming to help with the procurement of solutions. They want to prevent an inbound salesperson from being able to influence critical decisions on an ML or AI model that could have a great impact on the organization. By implementing standards and education initiatives, the group can take steps forward with clean data and without undue bias.
05:53 – In 2022, their first goal is to use this education process to help with the procurement process. For example, they’re going to use a peer-reviewed model then publish content for educational purposes. This will include official courses and information on how to deploy these models.
06:41 – Later in 2022, their second goal involves centralizing the models for triage—using machine learning to make urgent decisions. In the beginning, these models mostly provide care management, but they have a growing potential to be used for research studies and more.
07:35 – Although it’s often by mistake, bias still makes its way into algorithms. This includes machine learning algorithms, which don’t have inherent bias but there’s no process to prevent it. So, these models could be a solution. Having a documented process of the best method to build, test, and deploy these models opens a greater possibility to eliminating bias.
08:17 – Companies that work with labs, medical devices, and drugs have much in common with the FDA—including the way that they’re regulated and have strong management control processes in place. Therefore, they are looking at the learning models with these processes and applying it to machine learning
10:12 — There’s a reason that controls are put in place. By streamlining processes and decentralizing the high concentration of technology and regulatory expertise, data won’t be as scattered. The models are dependent on underlying data. So, it enables the AI and ML processes to harness the power of that data to work towards their goals.