Data can exclusively serve big tech companies to increase their profits; it can also accelerate research and strengthen science for the good of us all. After an inspirational talk at the TEDx Berlin, I interviewed Bart de Witte, a digital health leader, and the founder of HIPPO AI — an NGO focused on open source medical AI.
AI is already becoming better than a doctor at diagnosing, predicting and preventing diseases. How to convince doctors to create a team with algorithms? Is the conflict between “new, better, data-driven” and “old, traditional, intuition-based” avoidable?
We should focus on outcome-based care, where AI is the “Intel inside” of most transactions. Combined with a human-centric design, so the doctor sees less data, it would accelerate the adoption of digital health. Doctors could make faster, and hopefully, more accurate decisions. From my experience, most physicians do not fear AI because the amount of medical knowledge is growing faster then they can handle.
Secondly, there is less difference between intuition and mathematics as one might think. Daniel Kahneman wrote in the book “Thinking slow, thinking fast,” that there are two basic systems that doctors use to solve problems. System One is fast and intuitive. In this system, doctors rely on pattern recognition to make quick decisions. Healthcare is full of System One thinking, which includes cognitive biases. For example, one of the biases among physicians is over-confidence, which Kahneman calls “endemic in medicine.” He gives an example of a study that determined physician confidence in their diagnosis. Next, he compared the cause of deaths, as ascertained by autopsy, with the diagnosis the physician had made before the patient died. Clinicians who were utterly sure of the diagnosis antemortem were wrong 40% of the time.
There are practically no winning arguments for focussing solely on humans competences. Machines can process information faster, and with the support of the neural networks can detect much more complex patterns. If we allow computers to help doctors to diagnose, they could use the saved time to focus on human factors, such as compassion and kindness. The Stanford University Center for Compassion and Altruism Research and Education analyzed the impact of compassion on the quality of healthcare. Patients who feel comfortable with their care providers are willing to share more information about their health, making diagnoses more accurate. The research showed that compassionate healthcare environments are found to be more effective than aspirin in preventing heart attacks or stopping smoking habits
Human kindness, science, and technology form a powerful combination that can fight diseases, create a friendly environment for healing, and by doing so, decrease healthcare costs.
What does the development of self-learning algorithms mean for healthcare and humanity?
Without going in all the different domains, where AI is already delivering impressive results, the most significant impact of AI will be in the new insights that it will create by detecting patterns out of the phenotype and genotype data. As a result, we can develop new biomarkers and a much better understanding of the disease pathways. Potentially, we will find ways to prevent most systemic diseases (e.g., cancer). It’s an enormous opportunity for humanity if we implement technological progress into society and make it available for all. According to Hippocrates, a society where people are physically and mentally healthy is coherent and organic. A society where people are suffering from physical and mental diseases is fragmented and mechanical. Although AI is a two-edged sword, on the positive side, AI allows us to humanize healthcare again.
You are a founder of the HIPPO AI Foundation which aims at open sourcing medical AI. How do you want to convince different stakeholders with different interests to share medical data or knowledge instead of making profits from data?
The more I work in healthcare, the more I realize that inequality is a major issue. It’s not only about the gap between low- and middle-income or higher-income countries, but also differences in health outcomes between population groups. Your social conditions, where you are born, where you grew up, where you live, where you work and where you age, and which people you are connected to, determine your health status.
It took me a while to understand that the corporate structure that I was working for wasn’t the right model. If AI is here to serve us and will influence our decisions when it comes to health, medical AI should not be owned by any organization that operates for a financial interest first. Similarly, if a government or a public institution will hold medical AI, it can always be used to support the political interest. That’s why all of us should be the owners.
People who doubt it probably might have forgotten that before the enlightenment in medieval times, and from a cultural perspective, the disease was often seen as a punishment by God for sins. Literacy and access to knowledge have always been associated with power. Similarly, centralizing expertise in the few single organizations means that we allow people to gain control based on healthcare needs. Having ownership of exclusive knowledge is nothing new in healthcare, but until now, it is limited to pharma.
The question we need to ask is: Do we trust these organizations to decrease inequalities in healthcare? Perhaps the answer is yes, but then the second question we need to ask is: Do we allow them to privatize knowledge, something that always has been a public good?
Digital health enthusiasts provide future visions of efficient healthcare, patient-centered and AI-based medicine, empowerment of the patient and prevention-oriented systems instead of hospital-oriented strategies. How far are we from those dreams?
I think as we progress towards a digital healthcare system and patient centricity will become inevitable. As mentioned before, digital health solutions, if designed well, use a human-centered design. Most healthcare systems today are designed around the provider. While I was part of IBM Germany, we made a bold move together with the largest single health insurance sector, by pushing a PHR model, where the patient has access to his data, and digital services can be integrated and embedded in the digital patient journey. Neither IBM nor the Insurance had access to the two-fold encrypted personal data. This move was very different from the one that was driven by the government and focussed on a provider’s centric approach, which did not deliver any results in 15 years. Other insurers followed, and today, there are two major PHR providers.
70% of people worldwide population has no access to healthcare services. Digitalization can make healthcare accessible and affordable for everyone, which is also one of the WHO’s goals. But there are more factors to be taken into account like digital literacy, for example. How to avoid the digital divide in society?
Alvin Toffler, a futurist and author, predicted that the illiterates of this century are not those who can not read or write, but those who can not learn, unlearn and relearn. There are already many illiterates out there. From my perspective, the current digital illiterates are to be found within the older generations that are often also responsible for policy making and fear the unknown. I believe digital literacy is a crucial issue for those who are accountable for leading the transformation. That’s the reason I recently founded the DHealth Academy with Prof. Dr. David Matusiewicz, who is a health economist. We want to close the divide by educating the stakeholders within the system, so they become digital leaders, with a “digital-first” mindset. We aim to educate 100.000 people within the next five years. It might sound high, but compared to the over 4.4 Million people working in the industry, it is not a lot.
I wrote a high school paper on Artificial Intelligence. I described how it could be used in medicine to help doctors to make a better diagnosis, and save lives. I found the idea of AI integrated into our lives fascinating. With the support of AI, general practitioners in developing countries could become super-doctors.
One of the main problems with access to healthcare services is the lack of physicians and trained workers. Nigeria would need 300 years and use 20% of their GDP to educate as many physicians to come to the average OECD standard. With AI quickly becoming more accurate, there is this massive opportunity to decrease all inequalities and cross the gap between high and mid-, low-income countries. There is also an enormous opportunity for the less developed countries to leapfrog. In China, for example, Ping An a Good Doctor has rolled out 1000 One-minute Clinics across eight provinces and cities. The clinic looks like a photo booth where the patients communicate with an AI doctor and can get diagnosed. Algorithms can recognize around 2000 diseases. Currently, there is still a human physician involved who reviews the AI decisions. As the system scales and learns, what is only a matter of time, the health kiosks will get fully automated. Every One-minute Clinic has more than 100 drugs in stock. All of them are cryogenically refrigerated to ensure their quality. If a user needs a medication that is not available at the booth, he or she can purchase it online through the Ping An Good Doctor App and enjoy the one-hour drug delivery services provided by local pharmacies.
We will see more of these innovative care delivery models in the region with doctors shortages. In Europe, there will be a lot of resistance to change.
How to change the mid-shift from focusing on gadgets, innovations, startups — which are also necessary but sometimes overhyped — to new Europe’s strategy, new moonshot for healthcare to build solutions based on AI and accessible for all?
When it comes to healthcare, Europe has a much different value set than the US. It was recently very well documented in a review of a Clinical Information System from a US provider in Denmark. The Danish Healthcare system’s culture is based on trust and consensus and differs a lot from the US, where the focus is on risk and compliance.
AI will probably influence nearly every single decision that our caregivers or we make when it comes to preventing, diagnosing, and curing diseases. Values, human rights, and ethics will become extremely important. In Europe, having access to health is a human right. Which means we have to build AI that follows this principle.
But from a patient perspective, it might feel like AI is shaping us, not the opposite. We should ask ourselves the question, what we can and should do, so our children, and we benefit most from AI in medicine and inequalities will be reduced. For the last couple of years, I have been discussing this question with researchers, executives, investors, and politicians, and even with the EU Commissioner who was responsible for the digital agenda in Europe. Although a lot of work has been and is being done around ethics and medicine, nobody could give me a satisfying answer. That’s why I started an NGO, that follows all the principles important in Europe. Europe still leads when it comes to having access to proper public research; we are a continent that focusses on spreading equality, cooperation, and open science. We want our healthcare systems to be sustainable. These are all the principles HIPPO AI bases on.
Many tech companies, like Apple or Amazon, are entering healthcare. Is it a chance or a threat?
That is a threat.
Amazon, Google, Tencent, Alibaba are companies that understand very well how to build data-driven business models. Data is that what helped Amazon build an algorithm that allows storing products I want to buy tomorrow, close to my home so that they can deliver it on the same day. It makes them extremely competitive.
Google with Google Search has proven that it understands how to create and scale data science-based services that help people to find better — like anywhere else — results.
Both companies are monopolies that have been dominating their market for more than a decade. They made our lives easier, but at the same time also created dependencies. Peter Thiel, a famous Silicon Valley investor, wrote in his book “Zero to One,” a monopoly is not a disease; competition is for losers. It’s the exact opposite of European healthcare systems values. We need to seriously ask if we want to inject these values into our healthcare systems and if we want data and medical knowledge to become privatized and monetized.