The Rise of AI-powered Health Care

Albert Katz (WG23) embraces a fundamental truth: nothing matters unless you’re healthy.

That simple philosophy, along with undergrad and master’s degrees in finance and computer science, and a Wharton School MBA in health care management, motivated him to launch Flagler Health, which combines patient data and the power of AI to improve health care.

Flagler’s AI algorithms analyze patient data, including medical records, to provide doctors with insights and recommendations for treatment and to optimize patient care. “I wanted to save money for the health care system,” notes Katz, whose startup was a finalist in Penn’s 2023 Venture Lab Startup Challenge and continues to innovate in areas like remote therapeutic monitoring and behavioral health. “Let’s try to gather that data so eventually we can help specialists adopt value-based care – providing optimal care to patients.”

It’s been called the health care AI revolution – an abundance of algorithms analyzing health care industry data to make medicine more efficient and effective. What does that look like? Flagler Health’s machine-learning for health care is just one of many approaches. (Btw, if you want to go deep on how that works, check out this podcast featuring Will Hu (GED19), Flagler’s co-founder and chief technology officer).

AI in the ER?

Flagler Health illustrates that this era of AI is inspiring entrepreneurial problem-solving in the health care industry. And it is also sparking innovation in research and pioneering algorithmic tools to improve health care on a global scale.

The application of AI in health care is widespread – and a fundamental focus of the Wharton Healthcare Analytics Lab, a research center launched at the Wharton School in October 2023.

Since then, co-directors Hamsa Bastani, a Wharton professor of operations, information and decisions who develops machine learning algorithms for learning and optimization in health care; and Marissa King, a Wharton professor of health care management, have led research and discussions about where AI is being used successfully in health care and where the challenges lie.

In conversation with Eric Bradlow, vice dean of Analytics at Wharton, Dr. King helped define the health care-AI landscape. “Machine learning and artificial intelligence have touched almost all aspects of health care at this point. If you think of everything from how you get reminders to pick up prescriptions, from who’s reading your radiology reports, to even how you’re being triaged in the emergency department, machine learning plays a key role in all of those facets,” she noted. “If you think about radiology reports, that’s arguably the place where AI has had the greatest penetration. Many, many of our radiology reports are read now by machines.”

The Wharton Healthcare Analytics Lab is collaborating with different stakeholders – patients, providers and policymakers — to design better algorithms across health care. Here’s a snapshot of where Wharton-led analytics are helping to inform the new health care AI revolution:

🩺 Resource allocation refers to equitably allocating health resources in environments that struggle with access to treatment or medicines. Angel Chung, a PhD student in Wharton’s Operations, Information and Decisions Department, is working with the Sierra Leone government in West Africa to use machine learning and optimization for essential medicine distribution across thousands of health facilities in the country. In a profile of her work published by the Wharton AI and Analytics Initiative, Angel said that 40% of patients were being turned away without receiving the medicines they needed. “We used a synthetic difference-in-differences model to evaluate the impact of our approach,” she said. “Our result shows around 20% improvement in people’s access to essential medicines and medical supplies by the second quarter of 2023. While introducing an innovative change into an existing government system is tremendously difficult, we have successfully incorporated AI technology on a national scale and showcased improvement in this resource-constrained setting.”

🩺 Workforce wellbeing addresses issues affecting the health care workforce, such as burnout. Large language models (which process vast amounts of text data) provide a way to mine data from sources that haven’t been tapped before, like the clinical notes kept on patients. “One thing we’re trying to do is to utilize data from electronic health records to understand where there’s likely to be a high risk of burnout or emotional overload in clinicians, especially nurses,” said Dr. King in an article posted on Penn’s Leonard Davis Institute of Health Economics. “There’s immensely rich data within clinical notes.”

🩺 Innovative trials are an area where algorithms can drive innovations in medical practice by improving the design of trials, or research studies that test the safety and effectiveness of new medical treatments. “Clinical trials tend to be statically designed,” Dr. Bastani told Penn’s Leonard Davis Institute of Health Economics (where she is also a senior fellow). “They’re not actually personalized or dynamically customized in any way. We’ve been thinking about leveraging data from historical clinical trials or pilots to warm start these predictive models.” The Wharton Healthcare Analytics Lab is collaborating with Penn’s Health Incentives and Behavioral Economics for Better Health on this project.

🩺 Treatment and care form the heart of health-care delivery, and large-scale health systems data can help researchers identify promising treatment strategies. The Wharton Healthcare Analytics Lab says that “Big data and the use of new algorithmic methods have an enormous potential to inform everything from treating patients with substance-use disorders…to establishing best practices for human decision-making in medical settings.” Recent data-driven research has looked at everything from trends in Buprenorphine treatment in the U.S., opioid prescribing and overdose deaths, to building algorithms to help fight sex trafficking.

🩺 Health equity is a key challenge as algorithms and AI are more prevalent in the design and administration of medical care. Datasets trained by AI and machine learning systems are often biased and can compromise systems and lead to imbalances in care. For example, an algorithm might improve the outcomes for one population of people, but not others. Dr. Bastani’s research on Rethinking Fairness for Human-AI Collaboration in part addresses this issue. She regularly discusses health equity as a priority for the analytics lab.

“Any health care dataset encodes biases because patients that face structural barriers are underserved by the health system and so we don’t have high-quality representative data on them,” noted Dr. Bastani, who has talked about biased datasets often in her research. “We’ve been thinking about addressing that by leveraging proxy data – for example, cheaper but more available data like Google Search terms and so on. A big challenge of machine learning is that the training datasets are very large and so it’s not always feasible to debias the training data. But what we’re really interested in are the downstream decisions…by leveraging auxiliary sets we can never perfectly debias the model, but at least the decisions that are coming out will be less biased.”

Conversation Starters

How are the entrepreneurs at Flagler Health using AI to make an impact in health care?

What is resource allocation in health care and how is Wharton PhD student Angel Chung using AI to impact this health care challenge?

Why are datasets often biased in health care and what is being done to make them less so?

Which area of health care-AI focus do you find most compelling? Share your thoughts in the comment section of this article.

The hero image was provided by canva.com.