Artificial Intelligence in Healthcare and Clinical Decision Support

“AI is poised to revolutionize clinical care, impacting every facet of healthcare and biomedical research, from groundbreaking discoveries to advancing clinical development and improving care delivery.”

Sean Khozin, MD, Founder of Phyusion

AI-powered tech could make healthcare more affordable, equitable, effective, and accessible. This isn’t just hype: preliminary research suggests wider AI adoption could lead to a 5 to 10 percent savings on US healthcare spending annually, translating to roughly $200 million to $360 billion per year (NBER 2023). The potential is massive.

However, some risks do remain. Building public trust in appropriately validated tools, and developing appropriate safeguards and risk mitigation strategies, will be key. As it stands, the majority of the American public remains doubtful about the use of AI in healthcare diagnostics (Pew 2023).

Some apprehension is warranted: non-clinical AI applications like ChatGPT have significant error rates that would be intolerable in an area like healthcare. But several complex AI-powered tools have been working successfully in medical settings for years, with many doctors and healthcare systems turning to them for clinical decision support (CDS).

Done right, AI has the potential to revolutionize healthcare. But it will take the work of many to integrate it safely, effectively, and equitably. Read on to learn more about the benefits, challenges, and future of AI in healthcare and clinical decision-making.

Meet the Expert: Sean Khozin, MD, MPH

Sean Khozin

Dr. Sean Khozin is a physician-executive, board-certified oncologist, and data scientist with over 15 years of experience in therapeutic development, drug regulation, and artificial intelligence and machine learning applications in biomedical research.

As the founder of Phyusion, Dr. Khozin is currently working on advancing ventures at the intersection of biology, technology, and artificial intelligence. Previously, he was a founding member of the US FDA’s Oncology Center of Excellence and Executive Director of Information Exchange and Data Transformation (INFORMED), the FDA’s first data science and technology incubator that he established under special federal authorities.

Dr. Khozin currently serves on the boards of the Society for Translational Oncology, the Alliance for Artificial Intelligence in Healthcare, the Digital Medicine Society, and the Life Sciences Council of the CEO Roundtable on Cancer.

The Benefits of AI in Healthcare and Clinical Decision Support

“AI provides numerous opportunities to transform clinical care, from improved diagnostics to tailored treatment plans and efficient resource allocation,” Dr. Khozin says.

Historically, machine-based tools have been ideally suited to handle machine-like tasks: analyzing huge volumes of data, identifying patterns, and finding the needle in the haystack. In healthcare, AI is already helping screen for breast cancer, spot early signs of sepsis, and detect irregular heart rhythms.

AI-powered chatbots might even help solve some reimbursement hassles (AHA 2023). These AI tools have been at their best when they eliminate repetitive tasks and empower clinicians to engage with more complex and critical issues. But they’re beginning to redefine what’s possible in medical science.

“One notable application is the use of AI-powered computer vision to objectively analyze histopathology specimens obtained from biopsies for cancer diagnosis,” Dr. Khozin says. “This can allow for more accurate and personalized diagnoses that surpass the limitations of the human visual inspection of the specimens using traditional microscopy.”

By spotting patterns that humans can’t, AI provides researchers and scientists with a more comprehensive understanding of cancerous cells. As cancers get classified into more precise and distinct categories, clinicians can develop more individualized treatment plans that can target specific characteristics of a patient’s cancer.

“Additionally, AI can help predict patient outcomes, identify optimal treatment plans, and tailor therapies to patients’ unique genetic profiles,” Dr. Khozin says. “This personalization can not only improve patient outcomes but also reduce healthcare costs by minimizing ineffective treatments.”

Challenges Associated with AI in Healthcare and Clinical Decision Support

The transition from paper-based records to electronic health records (EHRs) demonstrated many of the challenges facing health systems upgrading to AI today. Efficient systems can be implemented inefficiently. Helpful technology can become unhelpful quickly. Clinicians want to spend more time treating patients, while new tech requires training and a disruption of routine. In many cases, a clumsy AI-powered tool is worse than no AI-powered tool at all.

“Adapting existing clinical workflows to accommodate AI-powered tools requires both technological and cultural shifts,” Dr. Khozin says. “Moreover, consensus is needed on validation standards to ensure the safety and effectiveness of AI applications.”

Algorithmic bias remains a concern as well. In 2019, researchers discovered racial bias in a risk-prediction algorithm that helped determine whether patients needed extra medical care (Scientific American 2019). While the algorithm wasn’t filtering for race, it considered factors highly correlated to race, and heavily favored White patients over Black ones.

“To minimize unconscious biases in AI-powered CDS systems, it is essential for developers to employ diverse and representative datasets, foster multidisciplinary collaboration, and uphold transparency throughout the AI development process,” Dr. Khozin says. “Involving physicians as key stakeholders in the development and implementation stages is crucial, as they can act as safety nets to identify and address potential biases.”

It’s also possible that AI could reinforce health disparities through its uneven distribution. In more affluent geographical areas, health systems may be able to afford more AI-powered services, and the more affluent patient population may be able to afford them regardless of their insurance’s reimbursement policies.

Meanwhile, low-income and rural areas may lack AI services and an inability to pay for them regardless. Dr. Khozin notes that health plans must develop clear policies to reimburse AI-driven healthcare tools and tech.

“Equitable access to AI tools in underserved communities is a critical moral and ethical concern,” Dr. Khozin says. “Proactive policies must be implemented to ensure fairness and prevent disparities in healthcare outcomes.”

The Future of AI in Healthcare and Clinical Decision Support

AI in healthcare requires intelligent and iterative regulation and policy. This is an area where regulation needs to be particularly flexible to neither squash innovation nor become irrelevant with every software update. Policymakers must consult with AI experts to develop guardrails that benefit patients and providers, without neutering the technology’s most promising capabilities.

“Regulatory frameworks need to be agile and adaptable, with regulatory authorities having the ability to attract top AI talent to make nuanced and informed decisions,” Dr. Khozin says.

Historically, powerful technological tools have disrupted workforce dynamics, and healthcare is no exception (Technology Review 2020). Third-party payers must adequately compensate physicians for their time and expertise when interpreting AI analyses and clinical recommendations. Dr. Khozin notes that this approach would discourage a reflexive overreliance on AI tools and ensure that physicians remain actively engaged in decision-making.

There’s still work to be done in the sphere of public trust. A recent survey found that six out of ten Americans would feel uncomfortable if their healthcare provider relied on AI for diagnosis or treatment, and fewer than four believed AI-powered diagnostics and treatments would lead to better outcomes (Pew 2023). Healthcare administrators, policymakers, and AI developers will need to work together to realize a blueprint for trustworthy AI (CHAI 2023).

As AI discovers and designs more drugs, clinicians can bring new therapeutic approaches to some of the world’s most challenging diseases. Precision medicine will get more and more precise. Outcomes will improve. Trust should follow. The new era of AI-powered healthcare may leave patients and clinicians wondering how anyone got by without it.

“AI is poised to revolutionize clinical care, impacting every facet of healthcare and biomedical research, from groundbreaking discoveries to advancing clinical development and improving care delivery,” Dr. Khozin says. “By addressing organizational, cultural, and data access hurdles, we may witness a paradigm shift within the next decade where AI significantly augments or even supplants human involvement in diagnostic processes in some cases.”

Matt Zbrog
Matt Zbrog Writer

Matt Zbrog is a writer and researcher from Southern California. Since 2018, he’s written extensively about emerging topics in medical technology, particularly the modernization of the medical laboratory and the network effects of both health data management and health IT. In consultation with professors, practitioners, and professional associations, his writing and research are focused on learning from those who know the subject best. For, he’s interviewed leaders and subject matter experts at the American Health Information Management Association (AHIMA), the American Society of Clinical Pathology (ASCP), and the Department of Health and Human Services (HHS).