Human-centered AI: A blueprint for transformative healthcare
With AI being all the buzz, it’s important that the next frontier of healthcare innovation step beyond just ‘going live’ to solve vexing challenges.
Not long ago, health systems experienced a digital transformation with the launch of electronic health records systems. At the time, there was pressure from within and from regulators to go live by a certain date. It was a huge transformation, and healthcare leaders were focused primarily on meeting meaningful use deadlines.
That transformation was driven by project-designed implementation and, while many organizations met the deadlines, it was painful for many in the industry. Many EHRs have poor user interface design and lack standard data formats, which make workflows worse and increase clinical staff work. EHRs also generate vast amounts of data, which create alert fatigue and overwhelm caregivers. In addition, the traditional patient-physician relationship was disrupted, and now clinicians spend more time with computers than patients.
Now, healthcare is at the beginning of another tech transformation driven by artificial intelligence, and the industry has a new opportunity to approach it from a standpoint of “human driven implementation.”
More than just using new tech
The goal of AI cannot be just using it – it has to be delight. Delight the patient; delight the clinicians; delight health system administrators. Adoption can’t be the end goal, because if we go through massive disruption and change and fail to improve the human experience, the change won’t stick.
To complicate matters, the speed of the AI revolution is happening fast in a more connected, integrated and digitized world. The challenges, the risks and the societal impact are much higher. Job displacement, ethical use, security and privacy are main concerns but not the only ones.
What are the unintended consequences of a poor implementation? To name a few, there’s legal risk, regulatory risk, public distrust and backlash.
The ability of our healthcare infrastructure to execute change, and do it well is more important than ever. The industry needs people who can navigate change successfully while engaging participants in the process for positive outcomes.
For AI transformation in healthcare (or any industry) to be successful to organizations and society as a whole, we have to design implementations with humans in mind. We cannot just invest in the newest technology – we have to invest in human experience and increase human skills as well.
The goal of AI should be to get out of the mundane and repetitive tasks and enable us to tend to the higher-order, higher-value aspects of patient care. And shouldn’t the goal of AI be to empower human caregivers to be more human? Only then can we finally achieve the aim of improving the healthcare experience for all.
Starting with the end in mind
Laying out the rules of the transformational road at the onset is critical. Any change effort, especially one as big as AI adoption, needs governing principles. Too often, leaders avoid dealing with challenging issues like fairness, transparency and ethics. In doing so, they are dooming change efforts to failure.
Before launching any transformation, design guiding principles and identify where humans oversee and make the final decisions. For example, if a workflow is going to be automated or an AI tool is going to do the documentation, where in the workflow does the human oversight happen? How are decisions made and reviewed?
It’s also imperative to decide upfront how the transformation will be monitored and evaluated. What is the vision for the changed state and does the application provide sufficient auditing and monitoring data for tracking both the AI use, and associated outcomes and decisions in the workflow. In other words, how will we know that we achieved what we set out to achieve? Are the metrics and measurements in place to make that determination?
The promise of improved workflow
AI has the ability to correct the real constraints in the workflow that plague every healthcare institution. For that to happen, the organization needs to analyze workflows with the goal of identifying both the pain points and the value points of the experience.
Where can pain be eliminated? Where can value be added? This is always the goal. Very often, making work simpler creates big value. What can be taken away while still achieving the outcome? What can be automated to reduce manual effort? In healthcare there are three focal points to explore when it comes to reducing pain and adding value.
Handoffs
What is a handoff? Think of a workflow as a relay race. Runner one hands the baton to runner two. All runners are responsible for the outcome and the goal is never, ever, drop the baton.
In healthcare, it’s the same. What are the key handoff points in the process where there is room for mistakes? Focus on the handoff, where information or actions move from one person to another. Every handoff should add value to the outcome. Now we have AI as a member of the relay team. We have to know when AI has the baton and when it is handing it back to a human again.
First, carefully evaluate handoffs and see how they can be improved or minimized. Research suggests that when handoffs are reduced, error rates decrease.
Second, evaluate whether the handoff adds value or contributes to the outcome. If not, this is an area where workflow can be improved.
Bottlenecks
Next, identify where bottlenecks occur.
Identify areas where available resources are limited or lacking and operations slow down significantly. Can automation be used for repeatable, mundane tasks? Can resources be added or streamlined to increase capacity? Can the bottleneck be removed and allow individuals to work at their highest levels?
Black holes
Finally, watch out for black holes. This is where information gets dropped or errors are not being corrected.
For example, if a patient submits a complaint or tries to make a follow-up appointment, but nothing gets done with it. That is a black hole.
Look specifically at outputs from databases and error reports. Is anyone following up on the errors identified and correcting them? Black holes significantly damage both the patient and provider experience and are key areas to evaluate for making workflow improvements.
The promise and potential perils of AI are too great to simply strive for implementation. As stewards of healthcare organizations and providers of care, we must be human centered in our AI transformation efforts, or we risk losing the faith and trust of the very patients we went into this business to serve.
M.J. Reiners launched Summerland Education LLC in 2011, after two decades of experience leading technology transformations at premier healthcare organizations. She is the author of Engineering an Epiphany: Master Business Evolution Using the 7 Forces.