How can AI transform healthcare?

Dr. Tal Patalon, head of the KSM Research and Innovation Center, anticipates that the transformative potential of artificial intelligence and other new technologies in the healthcare industry is on the brink of realization. As she prepares to leave her four-year tenure, she told The Jerusalem Post that it’s time to bridge the gap between innovation and implementation in her field – and she hopes to be the one to do it.

“There is the Promised Land that AI will change healthcare,” Patalon said. “I am sure it will happen, but it is not happening today.”

Drawing a parallel with a trauma patient in the emergency room facing the critical “golden hour” for lifesaving intervention, she said that the current moment is the golden hour for the health system.

“If we want to rescue health organizations, we must know how to bridge the innovation to the implementation. We are not implementing enough.”

The world has become obsessed with ChatGPT, a large language model (LLM) chatbot developed by OpenAI. Experts are focused on how systems like ChatGPT could alter the education system, redefine and automate some professional roles, improve people’s writing skills, or even serve as a more intelligent search engine than the ones in use today.

Patalon suggested that ChatGPT or analogous LLMs hold the potential to disrupt the landscape of medical research possibilities profoundly.

She explained that only 25% of medical data are “structured,” meaning they have a code or follow a standard text format in electronic health records. The remaining data are in free text, making it challenging for research.

Consider a woman going through menopause, Patalon said. Often, her symptoms are described informally, like trouble sleeping, slight irritation, irregular periods, and sweating. These details could help create an algorithm for predicting the onset of menopause, which is often underdiagnosed because doctors may not consider it. However, to develop an algorithm for predicting menopause, access to the informal descriptions in the free text is crucial.

“With the large language models that we have today, you can access free text much easier,” she said. “Accessing free text will give us access to 75% of the medical record that is not researched today.”

How can these models comprehend the content they read, especially when each physician’s notes are likely to be distinct from others’ notes?

It is called “feature extraction,” Patalon explained. “With LLM, you also put it in context.”

For example, consider a three-year-old child, Patalon offered. If they have a fever and rash, it’s not the same as having a rash followed by a fever. These are distinct diseases; looking at fever or rash alone wouldn’t reveal the whole picture.

“Understanding the context is essential, and this is where LLMs make a breakthrough,” she said.

So, why aren’t physicians already doing it?

According to Patalon, a gap exists between developing algorithms to extract information, and validating their accuracy for practical use. However, the situation is rapidly evolving.

The US Food and Drug Administration began approving AI algorithms in 1995. However, in nearly two decades, only 50 got approval. But in the last four years, according to an article published by Health Imaging, almost 700 market-cleared artificial intelligence medical algorithms are available in the United States; the vast majority were approved in the last four years.

Patalon is positioned to affect the technology transformation in her industry significantly. She is a practicing clinician specializing in family and emergency medicine and an enthusiastic entrepreneur and applied researcher dedicated to fostering the creation of innovative health technology solutions. Before joining Maccabi KSM, she spearheaded the innovation department at Samson Assuta Ashdod University Hospital. Currently, she serves as a lecturer on health, big data, and ethics in Reichman University’s MBA program in healthcare innovation.

PATALON SAID that another area that will transform the healthcare landscape is DNA sequencing. In the last decade this technique has become technologically more advanced and dramatically more accessible to the public.

She thinks that DNA will be included in electronic medical records in the next five years, finally ushering in the anticipated era of personalized medicine.

“How do you practice medicine in an era of certainty?” Patalon asked. She said most medicine today is practiced under “uncertain” conditions. “We don’t know how our body will react, which drug, for example, to use for high blood pressure. There are five different types of pills, and we test them by trial and error.

However, she continued, “If I have your DNA sequenced to fit the right treatment immediately, I don’t need the trial and error. I know what is going to work, based on your genetic information.”

Could doctors know too much?

Patalon said how doctors deal with the questions they are going to be able to ask and the answers they get back will be one of the challenges of this new era, which will come with a new set of responsibilities for medical professionals.

“What do you tell a patient if he has a gene for something that you know will happen and there is no treatment?” Patalon asked. “What do you do with the information? How do we develop the tools to deal with and return all this information to the patient?”

Recently, Patalon published a paper addressing the ethical concerns related to misattributed parenting, which involves conducting a genetic test for medical purposes and discovering that one of the patient’s parents is not biologically related to them.

“You found out by accident. Now, what are you going to do? Are you going to tell the patient or not?” Patalon asked. “A lot of times, these cases involve adult children. Do you tell him he is not genetically related to his dad? This is a huge ethical question that no one is dealing with.”

What about the cost of all this information to the health system? Here, there is an easy answer, Patalon said.

Even if it costs more at first, personalized medicine will save the system money in the long run because the patients will receive better and more effective treatment.

PATALON SAID she looks at her four years with Maccabi KSM with pride. She led the team through COVID-19, including conducting research that changed the vaccination policies in the United States, allowing people infected with COVID-19 to wait 90 days before getting vaccinated. Former chief presidential medical advisor Anthony Fauci quoted her work.

“Research is supposed to find something new – that is the essence of research and innovation,” Patalon said, though she admitted that these successes do not happen as often as scientists would want. “When you change guidelines in medicine, it’s a wow. It is not ordinary. It happens very rarely. I am very proud of it because we invested so much in it and did it against all odds.”

Most recently, under her tenure, Maccabi became the first Israeli health sector organization certified for ISO 27701, an information security standard.

Throughout her tenure, Patalon has also led many international collaborations. However, she said that since the October 7 massacre, partners from abroad seem hesitant to collaborate, are not sure how to react and what to say, and are cautious of politicizing the R & D field.

“However, as academia around the world was publicly involved, such as leading universities, our field, as well as our partners, cannot stay detached or uninvolved,” Patalon said. “We are sure [the situation] will change after this acute stage.

“This time has been very challenging; however, many Israelis believe R & D efforts are a major driving force of our country, and therefore remain committed in an admirable way.”

If Patalon closes this gap between innovation and implementation, will she put herself out of a job?

“The relationship in medicine between the physician and the patient will never be replaced by technology,” Patalon said. “Technology will only help doctors to do their jobs better.”