Andrew Johnson is CTO for AIMed with responsibility for database management, web development along with client management. A highly experienced publishing executive with a passion for technology.
“I am formally trained as a cardiac-electrophysiologist and for the past 40 years I have prescribed a medication known as Amiodarone. This medication interacts with almost every other drug; especially the blood thinner, Warfarin, which is meant to prevent stokes from atrial fibrillation. Despite prescribing this combination over 10,000 times, I get the same alert about a potentially a serious drug-drug interaction from our electronic health record (EHR): “Do I know and really want to prescribe amiodarone and Warfarin?” and two clicks later do “I really want to prescribe Warfarin and amiodarone”. This is what leads to alert fatigue and this is where I see artificial intelligence (AI) coming in.
Ideally, the system should realize I regularly prescribe amiodarone or Warfarin and I have indicated previously that I am aware of a possible drug interaction. However, I should get an alert if I prescribe a novel medication or combination of medications. I hope AI will help us overcome these simple and mundane chores in the EHRs to make a real difference,” Dr. John Windle, Professor, Internal Medicine and Holland Distinguished Chair in Cardiovascular Sciences at the University of Nebraska Medical Center describes to AIMed at a recent interview.
The need to go patient-centered
The EHR was developed and evolved to support billing and administrative purposes. While there is no doubt EHRs have significant advantages over the paper records, they have persisted in maintaining the paper-based metaphor by storing unstructured documents that are not readily digested and shared by the computer. There is a lot of enthusiasm for a new standard called FHIR (Fast Health Interoperability Resource). FHIR tries to standardize data within and between EHRs, but it is simply a pipe and does not necessarily capture what the clinicians want or intended.
What Dr. Windle advocates is to have clinically meaningful data. He encouraged everyone, including patients to be able to enter data and have clinicians to verify them. “If we think about data rather than documentation, I think that will lead us to better AI,” Dr. Windle says. Involving patients as a member of the healthcare team will not only ensure the collected data have their interests in mind but also to keep their needs at the core of an AI design. In the long run, this will equate to a revamp in patient engagement and education.
A key to success is to create a data dictionary that includes terminology that is understandable to patients that can be refined into medically relevant terminology and ultimately verified by the clinician. Clinicians will be freed up from data collection to data verification and information synthesis; moving away from simply clicking boxes to maximize billing and collecting capturing the patient narrative. “I think there is a large overlap between what is considered good for AI and what is good for patient care. A format of structured data plus an emphasis on capturing the patient narrative will serve as a very rich foundation for AI development”.
The need to have a proper structure for interoperability
What Dr. Windle and his team are doing now is to try to work with different organizations, EHR vendors, healthcare systems and clinicians to have “one term means one thing” across all domains. “I think this is what we are going to need for interoperability. Everyone is talking about FHIR but what they need to know is if you don’t have a good definition or something good to begin with, it will just be garbage. Every IT person knows what is ‘garbage in, garbage out’. Until we solved the ‘garbage in’ problem, we are not going to have an interoperability system”.
“The HITECH Act of 2009 pushed interoperability without understanding clinical medicine and the need for good data. It pushed a few large vendors to the front and we lost our ability to innovate”.
Even though EHRs are not perfect, Dr. Windle believes we will not be getting away from it. “AI still needs data to operate; that’s fundamental, but at the same time, I think the sources of data are going to get more interesting and more robust. Sooner or later, we may start including social determinants of health, GIS (Geographic Information System) and wearable data and so forth”. These individual patient level information captures variability at a more granular level, contributing to a more personalized intervention in the long run.
The need for individual patient level data
“One thing humans do very well is pattern recognition. So, I can look at 25 blood pressures of the same patient over a period of time and go, okay, your blood pressure is not controlled and we need to adjust your medication. I don’t need an EHR that stores this information as a document under the media tab. I need is the data pushed to me and a useful display of the data, perhaps an algorithm that displays blood pressures relative to medications over time.
Ultimately, we need to get to precision in medicine. What I need is where this patient stands in comparison to others who have similar demographics and in similar condition and how will the patient respond to therapy and progress in the next five years. We need to build these rich and individual variabilities into the guideline and quality metrics to further assist clinicians in prevention and early detection”.
Besides, Dr. Windle noted that there are e-commerce platforms with search engines and databases that serve in the interests of users by presenting information that they may find useful based on their previous experiences. However, EHRs do not have such function at the moment and he hopes it will come. “I think we are going to see EHRs that create unique, personalized views of the data across a spectrum. There are going to be the ones for public health which look at huge, population level data and then there is going to be the personalized ones that is going to help in giving unique patient experiences”.
Dr. Windle will be joined by fellow clinicians, healthcare leaders, C-suite executives, technical experts and many more in the upcoming AIMed Cardiology virtual event in association with the American College of Cardiology on 4 November. You may register your interest or obtain a copy of the event agenda here.