August 30, 2021 | Net Health
4 min read
5 Components of Effective AI Implementation in Healthcare Analytics
The future for artificial intelligence (AI) solutions in healthcare analytics can best be summed up in two words—explosive and exponential. For key decision makers successfully staying on the cutting edge of this technology, it means the potential for big gains and revolutionary change. And for those who may feel they’re a bit behind the curve, it’s not too late to get caught up and start seeing ways to implement these solutions.
With the knowledge of what’s at stake here, what’s the first step for a healthcare leader looking to get involved? How do we go about educating ourselves on what’s possible, how to make it happen, and how to identify the make or break details to get across the finish line?
It all starts with understanding the recipe. What are the proverbial ingredients needed to create, implement, and sustain AI solutions that can drive things like predictive analytics, machine learning, and data-driven clinical decision-making?
In a recent webinar titled Modern Healthcare Innovation Leaders, several leaders in this space shared their thoughts on the key components needed for an effective AI implementation in the healthcare space.
1. Transparency
The number one component needed for success is transparency. And the reason transparency is so important is that it drives trust. If clinicians and healthcare leaders can’t trust the technology to deliver accurate and reliable insights, it’s effectively worthless.
Here’s where things get challenging. Everyone remembers their math teachers in school that always asked them to “show their work.” It was a way of tangibly demonstrating to the teacher how you arrived at your conclusion so they could trust your answer. And this is a great solution when the computations were limited to a few lines of notebook paper.
But what happens when there are tens of thousands (or more) calculations happening in real-time over many different data points to reach that conclusion? And what happens when the process of reaching that conclusion changes every time as a result of machine learning?
Algorithms with data inputs and operations that aren’t visible or transparent to the end user are referred to in the industry as “black box AI algorithms”, and they are likely to be an obstacle to provider buy-in. Champions of AI implementations need to find creative ways to build the trust between machine and clinician to drive adoption, and it starts with transparency.
2. Learnability
One way to drive that trust is by giving clinicians and end users the ability to provide feedback and be intimately involved in the refinement of the solution. First, this allows them to see more into the inner workings of the algorithm, which helps with transparency. Second, it will inevitably improve the algorithm. And third, when clinicians feel included in the development process, the chances of buy-in are higher.
3. Tied to a Value Component
Technology for the sake of technology or data for the sake of data are not the way forward with anything. Any AI solution needs to be tied to a value component—a metric that has a marked and impactful change on the goals of the organization. Often, the best way to get there to start with the problem first, develop an intimate understanding of the problem, and work backward. Instead of building it first before deciding how to use it, the solution should be based on a clearly identified problem and the data that is needed for that particular problem.
4. Targets
Not only do these AI solutions need to be tied to a value component, but there needs to be clearly defined targets. People love to talk metrics, but often forget targets and goals. The metric might be readmission rates, but the target may be a specific percentage. And while gains are still possible without this, it limits the ceiling of change and confines the solution to incremental change. When effective targets are in place, though, it pushes the limits of what’s possible and gives the team the opportunity to challenge critical assumptions. This is where revolutionary change begins.
5. Intangible Goals
It’s easy when you start to delve into the data realm to forget the intangibles. Remember, not everything meaningful is measurable, and that’s a critical thing to focus on here. What we need to do is get to the relentless obsession of understanding what really means something to people across functional, social, and emotional dimensions of progress. It’s a tall task, but the truly transformative solutions and business models are born out of a deep understanding of these measures.
For example, a rehab therapist is working with a patient with a hand injury and asks the patient what he is looking forward to when he gains the full use of his hand again. The patient replies that he’d love to laugh again and be able to play his guitar. So, what’s the measure? Laugh again and play the guitar. Discovering and taking into consideration the personal motivators behind patient behavior must also play a role in healthcare solutions.
Learn More Now
If you’d like to hear more of this discussion we’d encourage you to check out the complete webinar. AI may seem like a tall task to tackle, but with the right direction—it’s more feasible than you might think.
Modern Healthcare Innovation Leaders
How Top Health Systems Plan and Execute Innovation