Students receiving lessons from a teacher.
Getty Images
Tremendous excitement. Grand predictions. High hopes. These words describe the zeitgeist of healthcare AI today, but they also describe how many people felt about electronic health records in the 2000s and 2010s.
Since then, nearly every health system and practice in the U.S. has implemented EHRs, improving care in some ways and worsening it in others. Results have been mixed, with organizations that have invested in people and systems seeing better results overall.
AI is the next phase of healthcare’s decades-long digital transformation. Although deployment methods will differ, organizations adopting AI will be best served by applying lessons learned from implementing and leveraging EHRs.
Lesson 1: Set realistic expectations
After decades of hope and hype around the digitization of healthcare, many expected EHRs to make healthcare safer, less expensive, and more effective. But that hasn’t turned out to be the case.
EHRs are a mix of good and bad: First, while they provide information to clinicians, they also overwhelm them with unnecessary and meaningless information. For example, clinical notes are now easy to read and retrieve, but they are often bloated with unnecessary, redundant, and sometimes incomprehensible information.
Similarly, EHRs both bring clinicians and patients closer and push them apart: Portals facilitate communication between appointments, but unsightly screens and keyboards in the exam room impede human connection.
EHRs make clinicians more productive in some ways but less so in others: For example, while it’s easy to prescribe medications and communicate test results electronically, clinicians must deal with countless alerts and notifications.
Lesson 2: Put people first
Many criticize EHRs for serving billing needs rather than improving clinical care, which often leads to nurses and clinicians finding them difficult to use and contributing to burnout. However, organizations that prioritized their employees by providing clear communication, investing in implementation, and individualizing training performed better.
To leverage AI, organizations must start by winning back the hearts and minds of patients and providers who no longer believe the promise that technology will necessarily improve healthcare. To do this, they must use AI to improve outcomes and experiences (not just billing and efficiency) and make AI tools easy to use and support the change.
Lesson 3: Improving systems of care
Health IT does not function in isolation but is part of a socio-technical system involving different teams and workflows.
When adopting EHRs, many organizations left their existing paper-based workflows as they were, rather than digitizing them and updating their teams for the digital world. This resulted in many incompatible and wasteful workflows, often forcing healthcare workers to find workarounds and perform tasks that were previously done by others. However, organizations that have redesigned their workflows and restructured their teams for the digital world have achieved better results.
Organizations must avoid making the same mistakes with AI. As Bill Gates explains, “The first rule of any technology used in business is that when you apply automation to an efficient task, you make it even more efficient. The second rule is that when you apply automation to an inefficient task, you make it even more inefficient.”
So instead of rushing to automate broken processes or using AI as a band-aid for poorly designed technology, organizations should first optimize their EHRs, streamline operations, and eliminate wasteful tasks. Initiatives like the Getting Rid of Stupid Stuff (GROSS) program can help.
Lesson 4: Continue to invest in change
Many organizations treated EHR implementation as a one-time event, not realizing that it’s impossible to fully predict what a “live” EHR will look like before they go live with it and train their employees en masse.
As a result, many EHR tasks are cumbersome (for example, physicians at one health system must click 61 times to order Tylenol) and many clinicians don’t use powerful EHR features (for example, Epic reports that only one in three physicians use chart search). Conversely, organizations that have prioritized ongoing training and EHR enhancements have achieved significantly better results.
The key is that AI implementation is never done: organizations need to continually monitor AI, evaluate its effectiveness, support frontline staff, and ensure that AI-based tasks are always aligned with the work to be done.
The risk is too great to fail
Healthcare organizations can leverage AI to make care more accessible, effective, and efficient. But success is not guaranteed. Organizations that follow the lessons learned from EHR adoption – setting realistic expectations, putting people first, improving health systems, and continuing to invest in change – are most likely to succeed.
