The Centers for Disease Control and Prevention (CDC) acknowledges that social determinants of health – such as race, finances, housing, and employment – can lead to disparities in health outcomes. Organizations from the American Medical Association to the American Public Health Association joined the CDC in working to address the underlying issues that lead to health inequities.
Two physicians with the UAB Medicine Cardiovascular Institute (CVI) are leveraging the power of natural language processing and artificial intelligence to help address those inequities. “There are gaps in cardiac care that sometimes occur because of a patient’s clinical or social status,” says Associate Professor of Cardiovascular Disease Julian Booker, MD. “We saw an opportunity to close those gaps and ensure that all patients receive the highest level of care in a timely fashion.”
From Concept to Implementation
In September 2019, Dr. Booker and Assistant Professor of Cardiovascular Disease and Radiology Efstathia Andrikopoulou, MD, partnered with third-party vendor Mpirik Cardiac Intelligence to develop a software algorithm that provides clinical decision support to help identify patients at risk for heart valve disease who otherwise might be overlooked. “We focused on supporting echocardiography in Cardiovascular Imaging Services, cardiovascular surgeons, and the Structural Heart and Valve Clinic,” Dr. Booker says.
After defining and validating the algorithm and analyzing preliminary data, the clinical decision support system went live in February 2020. Now, the group is working on better identifying patients who have heart failure. “It’s a complex disease, and we want to connect patients with our heart failure specialists and electrophysiologists in a timely manner, so that their need for defibrillators can be evaluated,” Dr. Andrikopoulou says.
Also, the team is in the final stages of developing an algorithm that helps identify cancer patients and cancer survivors who need cardiology services. “We want to make it easier for oncologists to easily identify these patients and refer them to a cardiologist,” Dr. Booker says. “This is an exciting collaboration that’s generated a lot of interest.”
Central to the group’s work is studying the socio-demographics of UAB Medicine’s service area. “Characteristics of certain neighborhoods correlate with poorer outcomes in patients with valvular heart disease,” Dr. Andrikopoulou says. After obtaining approval from the UAB Institutional Review Board (IRB), the team examined patients’ ZIP codes. “Among people living in neighborhoods where the average income is less than $60,000 per year, we found that more than 20% of the residents are Black, and residents who have inadequate access to transportation are at higher risk of experiencing faster worsening of their heart valve disease,” she adds.
Mpirik’s data scientists, data engineers, and data architects helped Drs. Booker and Andrikopoulou extract data based on ZIP codes. Their next step is to integrate the data with personalized patient data from UAB Medicine’s electronic health record (EHR).
“Combining both personalized and aggregate data will pave the way to understanding and providing equitable care to our patients,” Dr. Andrikopoulou says. “The decision to integrate EHR data makes the possibility of using technology to support clinicians a reality.”
Recognizing that cardiovascular issues arise across the spectrum of care, Drs. Booker and Andrikopoulou anticipate future partnerships with other UAB Medicine areas, such as surgery and obstetrics. “When we think about cardiovascular-related conditions like atherosclerotic disease, diabetes, and hyperlipidemia, we realize that we have the opportunity to optimize clinical care both in the aggregate and in individuals,” Dr. Andrikopoulou says.
A Scalable System
The team sees its work both as a proof of concept and the tip of the iceberg. “The technology and approach we used is entirely scalable across all facets of the delivery of medical care,” Dr. Booker says. “There’s no limit to the services that can be provided across UAB Health System. It only requires patience and algorithm development.”
Dr. Andrikopoulou connects the potential of machine learning and artificial intelligence with the need for clinicians to expand their toolkit. “This program is a prime example of how physicians cannot do it all,” she says. “We have to recognize that we can no longer be the sole agents responsible for providing high-quality care.”
In turning to Mpirik, the team was able to process and sort a large volume of data. “With this additional information, we gained insights that can transform the way we deliver care,” Dr. Andrikopoulou says.
Dr. Booker says the technology’s power is in being able to quickly and accurately review charts to identify patients who may qualify for certain types of care. “Health care delivery is a winding path, and patients occasionally slip through the cracks, either at the patient end or the health care end,” he says. “What drives our effort is a desire to ensure that all of our patients receive equitable, high-value care.”
Referring to the Mpirik project as a labor of love, Dr. Booker says his interest in natural language processing and artificial intelligence as a diagnostic tool grew organically. “I have an interest in quality improvement and health system management,” he says. “But I took the next step because I saw a gap in our ability to identify patients with clinically significant heart valve disease.”
He quickly internalized the technology’s potential. “If we allow the system the time it needs to understand patients, there’s no ceiling,” Dr. Booker says. “We’re limited only by the guardrails of our imaginations.”