Patient advocate at HPI event: "Where is the data? Where is the AI?"

Contents

Artificial intelligence could fundamentally change healthcare, but a large gap still exists between technological vision and actual care. Experts criticized this at the Digital Health Innovation Forum at the Hasso Plattner Institute (HPI) in Potsdam.

The conference was opened by Ariel Dora Stern, Professor of Digital Health, Economics & Policy at HPI, and Lothar Wieler, Professor of Digital Global Public Health, both of whom emphasized the institute's role as an interface between research, politics, and practice.

Jennifer Goldsack, CEO of the Digital Medicine Society, a non-profit organization from the USA, highlighted the significant discrepancy between the visions of most participants and reality. The British native, who has lived in Florida for over ten years, described her cancer illness. For years, she had no access to medical care, repeatedly sought medical help for months but only received a diagnosis when it had already become an emergency.

“Where is the data?”

Despite access to cutting-edge research, Goldsack experienced outdated processes and hardly any visible AI use in everyday clinical practice. Her repeated criticism of the lack of use of existing data became particularly clear. “Where is the data, where is AI? Why don't we combine this data and use it?” In the hospital, the IT systems were also outdated, and sometimes simple spreadsheets were used to keep track of administered medications and painkillers.

Marek Rydzewski, Deputy Chairman of the Board of AOK Nordost, and investor Alexandre Momeni discussed investments in AI. Germany continues to be “a major construction site with individual success stories,” according to Rydzewski. While there are advances such as the electronic patient record, a lack of incentives and high regulatory hurdles are slowing down innovation.

Much shadow AI in use

A central problem is the lack of risk-taking in the system. Innovation is rarely rewarded, while potential errors are severely punished. Rydzewski referred to his experiences at the Barmer health insurance company. There, when using a chatbot, there was intensive discussion about whether they could afford incorrect answers. Instead, the focus should be on minimizing the risk of incorrect answers. At the same time, many employees and patients are already using AI informally. In this context, Rydzewski spoke of a “shadow AI” that is establishing itself bypassing existing structures.

Momeni nevertheless sees AI as a potential turning point. The technology can make care more scalable and cheaper. However, it is currently often used in administrative areas without patients directly benefiting. “We need to experiment much more aggressively with the data,” said Monemi.

From global AI to real system limits

Catherine Pollard, Director of Impact at Google DeepMind, presented AI as infrastructure for a global health system. This includes models for predicting climate-related health risks and personalized applications for chronically ill patients. Guy Tsafnat, founder and Chief Scientific Officer of Evidentli, warned of new risks. Similar to medications, different AI systems could influence each other. Such interactions have hardly been researched so far but could have significant effects on care and costs.

In the panel “How Research Scientists Build Health AI,” it became clear that the biggest hurdles are not technical. Björn Eskofier, Director of the Institute for AI in Medicine at Ludwig Maximilian University of Munich, emphasized that algorithms are not the problem but rather the lack of interdisciplinary collaboration. He described an early collaboration with a neurologist who wanted to measure Parkinson's disease using sensor systems in sports shoes. He initially found the idea absurd, but it later led to research, start-ups, and clinical applications. Innovation only arises when technology, clinical practice, regulation, and care are considered together from the outset.

Jessilyn Dunn, Associate Professor of Biomedical Engineering at Duke University in Durham, North Carolina, pointed out that wearables and digital sensors have long been generating health data that could enable preventive medicine. Nevertheless, the system remains reactive. We continue to wait for people to get sick, even though many developments are predictable. Dunn also cited Goldsack's case as “a perfect example” of how many diseases announce themselves early.

Jonathan Kolstad, Professor at the Haas School of Business at the University of California, Berkeley, said that the healthcare system itself must be understood as a complex system. Much knowledge is not stored in databases but in the practical experience of doctors. AI must learn to navigate this system instead of replacing it.

(mack)