As Cedars-Sinai explores new uses of AI, it is balancing the rapid development of this emerging technology with responsible and ethical implementation.

“AI systems have the power to transform healthcare,” said Mike Thompson, vice president of Enterprise Data Intelligence at Cedars-Sinai. “If implemented properly and responsibly, AI can be deployed to enhance patient experience, improve population health, reduce costs and improve the work life of healthcare providers.” 

Thompson sat down with the Cedars-Sinai Newsroom to examine the uses of AI to improve healthcare and to detail how the academic medical center is pursuing this fast-evolving technology in an ethical manner.

Why do ethics matter when it comes to the use of AI in healthcare?

The integration of AI into medical technology and healthcare systems is only going to increase in the coming years. As technology continues to develop, the push toward safety, soundness and fairness occurs at all levels. This effort will require checks and balances from innovators, healthcare institutions and regulatory entities.

As technology advances, the medical community will need to develop standards for these innovative technologies, as well as revisit current regulatory systems on which physicians and patients rely to ensure that healthcare AI is responsible, evidence-based, bias-free, and designed and deployed to promote equity.

If AI systems are not examined for ethics and soundness, they may be biased, exacerbating existing imbalances in socioeconomic class, color, ethnicity, religion, gender, disability and sexual orientation.

Bias disproportionately affects disadvantaged individuals, who are more likely to be subjected to algorithmic output that are less accurate or underestimate the need for care. Thus, solutions for identifying and eliminating bias are critical for developing generalizable and fair AI technology.

Are AI ethics in healthcare different than AI ethics in other fields, like consumer goods?

While many general principles of AI ethics apply across industries, the healthcare sector has its own set of unique ethical considerations. This is due to the high stakes involved in patient care, the sensitive nature of health data, and the critical impact on individuals and public health.

It is critical that AI in healthcare benefit all sectors of the population, as AI could worsen existing inequalities if not carefully designed and implemented. It’s also critical that we ensure AI systems in healthcare are both accurate and reliable. Ethical concerns arise when AI is used for diagnosis or treatment without robust validation, as errors can lead to incorrect medical decisions.

What is an example of “AI ethics in action” at Cedars-Sinai?

As an example, consider an AI system that is used to assist in a patient’s risk for diagnosis. One question to ask is whether the AI algorithm performs equally for patients, regardless of race or gender.

In the same vein, an algorithm trained on hospital data from the European Union may not perform as well in the U.S., as the patient population is different, as are treatment strategies and medications.  

To combat these challenges, bias mitigation strategies may require us to implement mathematical approaches that help an AI model learn and produce balanced predictions.

At Cedars-Sinai, we also believe that critical AI algorithms should augment the expert, not replace that individual. Keeping the “human in the loop” to review a recommendation is another important strategy we use to mitigate bias.

Does Cedars-Sinai approach ethical AI differently than other academic medical centers? If so, how? And why?

To support our AI strategy, we created a framework for the ethical development and use of AI. The framework and policies are designed to ensure that the evolution of AI in medicine benefits patients, physicians and the healthcare community. It advocates for appropriate professional oversight for safe, effective and equitable use.

The framework starts by identifying who might be impacted and how, and then takes steps to mitigate any potential adverse impact.

How does the ethical use of AI evolve over time as Cedars-Sinai progresses in its use of these new technologies? 

The most powerful—and useful—AI systems are adaptive. These systems should be able to learn and evolve over time outside of human observation and independent of human control. This, however, presents a unique challenge in AI ethics, as it requires ongoing monitoring, review and auditability to ensure systems remain fair and sound.

Recent booms in AI technologies have been decades in the making. The most relevant and recent advances have accelerated the growth of AI algorithms and concepts—an evolution that will continue.

Now more than ever before, we must ensure that AI algorithms are trustworthy and deserving of trust. In healthcare, this entails systematically accumulating evidence, monitoring systems and data that are based on ethics and equity. 

Read more from the Cedars-Sinai Blog: The Human Factor of Artificial Intelligence

The algorithm, which identified hidden signals in common medical diagnostic testing, may help doctors better prevent strokes and other cardiovascular complications in people with atrial fibrillation—the most common type of heart rhythm disorder.

Previously developed algorithms have been primarily used in white populations. This algorithm works in diverse settings and patient populations, including U.S. veterans and underserved populations. The findings were published today in the peer-reviewed journal JAMA Cardiology.

“This research allows for better identification of a hidden heart condition and informs the best way to develop algorithms that are equitable and generalizable to all patients,” said David Ouyang, MD, a cardiologist in the Department of Cardiology in the Smidt Heart Institute at Cedars-Sinai, a researcher in the Division of Artificial Intelligence in Medicine, and senior author of the study.

Experts estimate that about 1 in 3 people with atrial fibrillation do not know they have the condition.

In atrial fibrillation, the electrical signals in the heart that regulate the pumping of blood from the upper chambers to the lower chambers are chaotic. This can cause blood in the upper chambers to pool and form blood clots that can travel to the brain and trigger an ischemic stroke.

To create the algorithm, investigators programmed an artificial intelligence tool to study patterns found in electrocardiogram readings. An electrocardiogram is a test that monitors electrical signals from the heart. People who undergo this test have electrodes placed on their body that detect the heart’s electrical activity.

The program was trained to analyze electrocardiogram readings taken between Jan. 1, 1987, and Dec. 31, 2022, from patients seen at two Veterans Affairs health networks. The algorithm was trained on almost a million electrocardiograms and it accurately predicted patients would have atrial fibrillation within 31 days.

The AI model was also applied to medical records from patients at Cedars-Sinai and it similarly—and accurately—predicted cases of atrial fibrillation within 31 days.

“This study of veterans was geographically and ethnically diverse, indicating that the application of this algorithm could benefit the general population in the U.S.,” said Sumeet Chugh, MD, director of the Division of Artificial Intelligence in Medicine in the Department of Medicine and medical director of the Heart Rhythm Center in the Department of Cardiology. “This research exemplifies one of the many ways that investigators in the Smidt Heart Institute and the Division of Artificial Intelligence in Medicine are using AI to address preemptive management of complex and challenging cardiac conditions.”  

The study was a collaborative effort between physicians and investigators at Cedars-Sinai and the San Francisco and Palo Alto Veterans Affairs hospitals. First author Neal Yuan, MD, is an investigator with the Smidt Heart Institute at Cedars-Sinai. Cedars-Sinai investigators Grant Duffy and John Theurer also worked on the study.

The investigators plan to continue to study the algorithm as part of prospective clinical trials to learn if it helps identify those at risk for heart attack and stroke. They also plan to develop more AI algorithms.

Funding: The study was funded by the National Institutes of Health and the U.S. Department of Veterans Affairs.

Read more on the Cedars-Sinai Blog: Treating Atrial Fibrillation: Risks and Complications

Through new “Idea-thons,” frontline clinical and administrative leaders and AI experts across the organization come together to cultivate ideas for using AI to enhance patient care, administrative systems and employee wellbeing. Participants then present their ideas to senior leaders, who vote on which projects to advance, support and, eventually, put into use across the organization.  

“Idea-thons empower our frontline caregivers and staff to shape the future of healthcare, tapping into their talent, creativity, and hands-on experience to craft tailored solutions for patients and the organization,” said Craig Kwiatkowski, PharmD, senior vice president and chief information officer at Cedars-Sinai. “We are leaning in to create our own road map of imaginative projects using generative AI technologies, centered on enhancing patient care and organizational efficiency.”

The Idea-thons are driven by Cedars-Sinai’s Artificial Intelligence Council, which is charged with guiding and coordinating the organization’s AI strategy. Cedars-Sinai’s AI strategy is built on three strategic pillars: investing and planning, transitioning innovation into adoption, and supporting the ethical, responsible, and scientifically sound use of AI, including generative AI.

Generative AI is a type of artificial intelligence that can create a series of outputs—like video, audio or texts—in response to human requests. These generative AI systems are based on language models like ChatGPT that can be trained to generate content.

The first Idea-thon, in June, brought together more than a dozen physicians from various medical specialties, including cardiology, emergency medicine, surgery, OB-GYN and primary care, as well as AI experts in data science, computational biomedicine and technology.

The session produced solutions with themes centered around care coordination, triage, medication management, communications and clinical trial recruitment. Some of the projects potentially overlapped with initiatives already underway at Cedars-Sinai. Others presented fresh opportunities to develop anew.

“These novel projects can be used alongside our customized generative AI tools and technologies within our health system,” said Mike Thompson, vice president of Enterprise Data Intelligence, who works with Kwiatkowski to steer the organization’s AI strategy.

Thompson said the Idea-thons are informed by the experience and wisdom of frontline healthcare clinicians and other Cedars-Sinai staff who understand firsthand the problems encountered in the delivery of healthcare.

“Our physicians and medical teams have direct insight into the roadblocks within their daily jobs,” Thompson said. “By bringing them into the discussion on generative AI technologies, they can help identify meaningful solutions to these hurdles—which will likely improve patient care, advance the frontiers of medical research and increase organizational efficiency.”

As these ideas transition into tangible solutions, Cedars-Sinai is dedicated to conducting rigorous testing and impact assessments to ensure alignment with the organization’s mission to deliver the highest-quality care. The collaborative teams involved in the Idea-thons aim to refine and tailor AI solutions to address the unique needs of both patients and the organization, while remaining steadfast to their strategic principles.   

The next Idea-thon, scheduled for October, will bring together an even broader group of healthcare professionals, including nurses, pharmacists and other clinicians. Another session will follow to include administrative and operational staff from Finance, Human Resources and other departments.

"We are truly inspired by the ingenuity of our staff and by the potential of the generative AI ideas we've heard,” Kwiatkowski said. “There are limitless opportunities for innovation within patient care, operational and administrative areas. We will continue listening to and seeking input from our frontline staff as we further explore how to best use these new tools.” 

Read more from the Cedars-Sinai Blog: Artificial Intelligence Advances

Building on a legacy of innovation, Cedars-Sinai is harnessing rapidly evolving breakthroughs in AI technology to enhance patient care, improve efficiency, advance scientific discovery, and cultivate greater physician and staff wellbeing.

AI at Cedars-Sinai already is having an early impact on clinical and research initiatives. Investigators, for example, are using the technology to identify the earliest signs of pancreatic cancer, predict sudden cardiac arrest and uncover genetic predictors of Alzheimer’s disease risk.

Cedars-Sinai leaders are excited by the possibilities afforded by AI—including the potential to reduce healthcare disparities and costs—as they guide the organization through this dynamic revolution in healthcare.

“AI extends and augments human capabilities and intelligence,” said Craig Kwiatkowski, PharmD, senior vice president and chief information officer. “It holds the potential to transform the ways we envision, plan and deliver care. Because of the vast opportunities, we are moving deliberately in these early stages of the journey.”

The Three Pillars of Artificial Intelligence at Cedars-Sinai

Cedars-Sinai’s AI strategy is built on three foundational strategic pillars: investing and planning, transitioning innovation into adoption, and supporting sound, ethical use of new technologies.

Through its first pillar, Cedars-Sinai is investing in state-of-the-art technology, infrastructure and services while fostering AI fluency across the workforce. The intent is to lay a foundation to meet the organization’s future healthcare needs.

The second pillar calls for the use of AI research and innovation to solve critical, real-world healthcare challenges—accelerating the integration of AI discoveries into clinical practice and delivering benefits to patients, physicians and the healthcare delivery system.

The third pillar focuses on the ethical and responsible use and governance of AI by adhering to regulatory requirements while ensuring that AI tools are used in fair and unbiased ways that protect patients and their privacy.

“The AI journey doesn’t have a finish line, and we must keep our eyes on the road ahead,” said Mike Thompson, vice president of Enterprise Data Intelligence, who works alongside Kwiatkowski to steer the organization’s AI strategy. “We are building a strong foundation for a future of limitless possibilities.”

Artificial Intelligence Council

An essential component of success involves the creation an Artificial Intelligence Council. It brings together cross-functional leaders—from patient care, research, data and technology teams—to review, guide and coordinate AI strategy. The council provides a forum for open dialogue and an ongoing exchange of ideas while setting priorities, evaluating the use of AI tools and identifying measures of success.

“The AI Council is a core piece of our commitment to using AI responsibly,” said Jason Moore, PhD, chair of the Department of Computational Biomedicine and a founding member of the council. “It is critical that we promote responsible AI principles, helping to ensure that AI is deployed safely, effectively and in an unbiased and transparent manner.”

Artificial Intelligence in Action

AI can accelerate scientific discovery by advancing the understanding of disease states and potential treatment pathways. At the same time, AI can improve the efficiency and accuracy of clinical data, freeing providers to spend greater face-to-face time with patients.

Early adoption of artificial intelligence already is making a difference in research and clinical programs at Cedars-Sinai. Examples include:

• Pancreatic Cancer: Cedars-Sinai investigators have leveraged AI to identify the earliest signs of pancreatic cancer, a disease notoriously difficult to diagnose in its early stages. By using advanced machine learning algorithms to analyze medical imaging scans and patient records, the AI system may help prevent deaths through early detection, leading to timely interventions and improved patient prognoses.
• Heart Health: Research led by investigators in the Smidt Heart Institute and the Division of Artificial Intelligence in Medicine in the Department of Medicine is helping clinicians get closer to predicting two common heart conditions: sudden cardiac arrest, which is often fatal, and increased coronary artery calcium, a marker of coronary artery disease that can lead to a heart attack.
• Brain Cell Modeling: Investigators from the Anastassiou Lab—members of the Departments of Neurology and Neurosurgery, the Board of Governors Regenerative Medicine Institute and the Center for Neural Science and Medicine at Cedars-Sinai—have created complex computer models of individual brain cells, unlocking new avenues for understanding brain function and neurological disorders.
• Alzheimer's Disease Research: An $8 million grant from the National Institutes of Health to study Alzheimer’s disease is enabling investigators to study new, leading-edge artificial intelligence methods and to use these to identify genetic predictors of Alzheimer’s disease risk.
• Liver Disease: Cedars-Sinai experts are investigating how ChatGPT may help improve outcomes for patients with cirrhosis and liver cancer by providing easy-to-understand information about lifestyle changes and treatments.
• Obstetrics and Gynecology: AI is helping physicians make headway in predicting the need for cesarean section delivery. By analyzing electronic health records, the Cedars-Sinai AI model can help physicians assess factors influencing the need for C-sections, potentially leading to better outcomes and informed decision-making for parents.
• Spine Surgery: The Department of Computational Biomedicine, in collaboration with Cedars-Sinai’s AI Council and spine surgeons, is using AI and machine learning to predict which patients are most likely to successfully manage their pain post-surgery and which ones might need additional assistance.
• COVID-19: During the height of the COVID-19 pandemic, Cedars-Sinai developed an AI model that helps physicians diagnose the severity of COVID-19 pneumonia. This technology already is being used in multicenter clinical studies.

Although AI initiatives are already having an impact at Cedars-Sinai, those leading the organization’s strategy are busily planning to expand the scope of applications.

“We are only at the very beginning of understanding what AI can do to improve healthcare and quality of life for our patients, our physicians and our staff,” Kwiatkowski said. “We are committed to building a firm foundation as we head into the future.”  

Read more in Discoveries: The Human Factor of Artificial Intelligence