Markus Lingman, the researcher who wants to transform healthcare data into better patient care
Markus Lingman describes, based on his experiences in clinical practice, leadership, and research, how AI and information-driven healthcare can improve patient care.
Markus Lingman is a medical doctor and researcher whose career spans clinical practice, leadership, academic research and AI. As a specialist in cardiology and internal medicine, he has seen healthcare challenges up close, but also the opportunities and value of the vast amounts of data produced by the healthcare sector. In this interview, he talks about his journey from his student days to leadership roles in Region Halland, and how his interest in AI and information-driven healthcare emerged. We also learn about concrete examples how data analysis and machine learning can improve care pathways and increase patient security—without compromising integrity and ethical principles. At the same time, the importance of collaboration and shared standards is highlighted as a decisive factor for Sweden to fully realize the potential of information-driven healthcare.
Could you tell us about your background? What led you into information-driven healthcare and AI?
- A large part of my career has been clinical, as a specialist in cardiology and internal medicine, but about 15 years ago I received my first leadership assignment. I became head of the physicians at a fairly large hospital department, and shortly thereafter I took on a role as departmental manager and soon also as area manager (which in some places is called division manager). For the past few years, I have mainly worked in healthcare management on strategic issues. In parallel with my clinical and leadership career, I’ve also conducted research. I completed my PhD at Sahlgrenska Academy in 2013, where I am still affiliated, and I also hold an adjunct professorship at Halmstad University.
- The idea of more information-driven healthcare emerged as a result of my three ‘legs’—clinical work, management, and research. In the early 2000s, healthcare became digitalized. A side effect was that enormous amounts of data started to be generated and stored. Strangely enough, this goldmine of insights wasn’t used to any great extent. Most of the metrics used for decision-making fit into an Excel spreadsheet, and we still relied primarily on basic arithmetic. I was frustrated that healthcare—a large, complex, adaptive system—didn’t have better ways to make the right decisions about organizational development. The term ‘follow-up’ was common, but follow-up is about looking backward. When developing services, you need to look forward and be able to consider multiple factors simultaneously, so you don’t create indirect negative side effects of the decisions you make.
- I saw a great need to look beyond organizational boundaries and focus on how things are going for patients in relation to everything we in healthcare do for them. Healthcare is very good at optimizing silos, even if that happens at the expense of deterioration somewhere else in the system, mainly because there is no fact-based overview of the whole. Information-driven healthcare simply means working more fact-based, holistically, and forward-looking (predictively) by making use of all the data that we in healthcare generate in large quantities every day. Among other things, we developed a resource description model called Patient Encounter Costing (PEC) to describe the resource consumption involved in patient management along complete care pathways. We now use the same approach to calculate climate impact. I just coined the term ‘information-driven’ to conceptualize and explain this. The idea came from The Institute for Information Driven Medicine at Harvard.
- In our work to utilize data in healthcare, we quickly ran into a new problem—it was just as complex and extensive as the healthcare system it was meant to describe. Traditional statistics and ‘business intelligence’ were no longer sufficient. There were too many factors to consider, and the relationships were not linear. At the same time, we were fortunate that machine learning became more broadly accessible, although still largely unknown in healthcare. We began collaborating with researchers in the field at Harvard Medical School and eventually at Halmstad University, and together we refined data into information that led to new insights we could start acting on and monitoring for patient-centric outcomes—the essence of information-driven healthcare. Over time, neural networks also became applicable, and we are now exploring the possibilities of using language models.
- Thanks to the capabilities we built, our organization today has entirely different conditions for developing healthcare and optimizing care chains in a fact-based manner than we did 10–15 years ago.
You started studying engineering but switched to medicine. What did you take away from the engineering program?
- Before studying medicine, I studied math, physics, and programming, but also national economics, corporate economics, and group economics, as well as languages within the Industrial Economics – International program. I was fortunate enough to work for short periods at European offices of Swedish corporations. Having a grasp of more advanced mathematics has, of course, been a great help when it comes to AI applications and research. Understanding economics and econometrics has been important because resource management is so central in healthcare—particularly from an ethical perspective in prioritization, where available resources must be used in the way that creates the greatest benefit. Healthcare still has a lot of work to do in this area. It is important to remember that money is just one of many resources in healthcare, and not necessarily the most constrained one.
Could you share an example where your work with information-driven healthcare has improved patient outcomes in Region Halland?
- Halland currently has fewer hospital beds than other regions. The decrease in beds is essentially due to limited availability of workforce. This reduction would have been much more problematic if we hadn’t simultaneously been able to work with the entire care chain so that the need for hospital admission was reduced. For example, we used AI to understand, in depth, what leads a person to end up in the emergency department. Many factors turn out to interact with each other, and some can be influenced. With fact-based oversight and predictive models, we have been able to focus on the right measures. The result is that we do not have more readmissions or higher occupancy levels than before, and at the same time, the patients have better outcomes. We’ve also been able to build better care pathways for specific patient groups. People with heart failure in Halland need much less inpatient care than was previously the case. Working systematically with data over a long period has contributed to Halland often performing well in national quality comparisons.
“Finding new ways to deliver healthcare is absolutely critical in light of our future demographics. Increased precision and prediction will be important factors, and these require data. I’m also convinced that many people would want to benefit from these capabilities once they themselves become ill or risk becoming ill.”
How do you view the balance between information-driven healthcare and protecting patient privacy? How have we managed this balance in Sweden?
- We have dealt with sensitive information in healthcare for all time, and one can be impressed by how rarely it is misused. The Health and Medical Services Act states that care should be good, meaning it should be safe, effective, knowledge-based, accessible, personalized, and equitable. None of this can be guaranteed without fact-based development, follow-up, and forward-looking analysis. Therefore, the law grants healthcare providers significant opportunities to do precisely this. The challenge is that protective legislation is tied more to organizational structures than to the patient (who receives care across many parts of the fragmented organization). The requirements to comply with privacy-protecting legislation are high and entail several control mechanisms and restrictions. A current issue is that much of the legislation was created in another era and therefore doesn’t provide guidance on today’s new questions. It’s clear that lawmakers, both nationally and at the EU level, want to see more data utilization to achieve healthcare goals and support research. The development of legislation and its interpretation is intense but takes time.
- In my view, Sweden has been quite cautious in order to protect privacy. However, I am equally convinced that one can protect privacy and utilize data at the same time. But you need to know what you’re doing and where the boundaries lie. The benefits for the individual and society are enormous. Finding new ways to deliver healthcare is absolutely critical in light of our future demographics. Increased precision and prediction will be important factors, and these require data. I’m also convinced that many people would want to benefit from these capabilities once they themselves become ill or risk becoming ill. But to maintain the public’s trust in healthcare, data must be handled with care and in line with the legislator’s intentions.
In 2020, you were named “AI Swede of the Year,” partly for your ability “to make the complex understandable.” Do we have a tendency to oversimplify when it comes to health issues?
- Absolutely. Many people look for simple answers to incredibly complex questions and quick fixes for difficult problems. Often, interventions are made with the best of intentions but without awareness of indirect negative consequences. This is a classic problem in governance based on indicators. It also leads to solving problems locally, even if that creates new problems elsewhere in the system or at a later point in time. With better access to facts and a culture where people can talk about things as they really are, we improve the conditions for tackling problems in the right way.
You have worked with American collaborators. What are the biggest differences in working with information-driven healthcare and AI in Sweden compared to the U.S.?
- The Swedish and American healthcare systems are very different. The possibilities for healthcare providers to use data are greater in the American system, but it is afflicted by it being tied to reimbursements. In the U.S., healthcare is an industry like any other, operating in a market. That drives certain types of AI solutions. At the same time, the U.S. has an absolutely outstanding research environment at several top universities and islands of clinical excellence. We can learn from and collaborate with them. But the applications must be consistent with Swedish priorities and how we run our healthcare. I do envy the way some leading academic healthcare organizations in the U.S. have been able to integrate research and clinical care in ways that are not feasible in Sweden.
- On the other hand, there are also enabling advantages in Sweden—not least the public’s trust and the equal access to healthcare. What we have in common is that decisions always improve with a stronger factual basis. Fundamentally, we face the same challenges with the rising healthcare needs of society and a limited supply of skilled workers, even though the demographic profile in the U.S. is more favorable than Sweden’s.
How can Sweden get better at information-driven healthcare and creating value from data? What are the most important measures?
- The first short answer is probably increased collaboration—both across organizational and knowledge boundaries. Right now, many people are sitting in their own corners trying to solve the same problems and interpret the same laws. We need more consensus. This could involve agreeing on standards, legal interpretations, and what is ‘good enough.’ At the same time, the government has some homework to do because the state agencies that healthcare providers must relate to are far too siloed from one another. Healthcare providers would benefit from more legal guidance.
- The second short answer is not to be tempted to treat this as an IT issue. Delivering patient-centric, holistic, forward-looking, and fact-based care is a matter of operations, where clinicians, managers, IT specialists, lawyers, (health) economists, and others must be involved in an integrated way.
How do you stay up to date in such fast-changing fields as AI and medicine?
- I continuously follow about 20 core scientific journals in medicine, AI, patient safety, healthcare quality, and health economics, and read them regularly. That amounts to many articles to review. At the same time, I’m fortunate to have like-minded and sharp colleagues both in and outside the country who help with monitoring. It’s very stimulating. When it comes to ethics and law, I think the WHO, the EU, and the OECD regularly produce excellent summaries and guidelines. Of course, I also read some popular media, but it often lags too far behind. Transformer models were introduced in 2017 but only gained widespread media attention in 2024. By then, all the important developments had already happened.
What do you see as the most interesting AI trends in healthcare over the next 3–5 years?
- What I’m most curious about is how we can make use of the enormous amount of unstructured data in healthcare, i.e., free text. It contains much more information than the structured data, and the possibilities to extract value from it have increased dramatically thanks to transformer models (e.g., language models). However, there are a number of uncertainties that we need to clarify before we see major effects in which healthcare professionals can be relieved on a large scale.
- What I’m hoping for is that we will be able to apply clinical decision support in day-to-day care to increase diagnostic precision, select tailored treatment more quickly for each individual, and take action based on risk rather than what has already happened. However, this will require clarity on issues of responsibility and what demands we should place on decision support tools. Is it enough if they are as good as an average doctor? What standards should we apply regarding transparency and mitigating bias? Upcoming legislation addresses this, but it needs to be translated into practice and discussed in society. We need to develop a more common view of the balance between risk and benefit when it comes to AI, similar to what we already have for pharmaceuticals.
If you could give one piece of advice to young physicians and researchers who want to work with AI, what would it be?
- I think the question is a bit misdirected. Sooner or later, a physician or researcher who does not have a decent grasp of AI will become less relevant or even obsolete. We’re currently working on how AI knowledge should be included in all Swedish medical education. It’s just a bit unfortunate that it took until 2024. You don’t need to understand every detail, but you must have a good sense of the risks and limitations, just like any other tool we use to help patients.
Note! The content on this blog reflects my personal opinions and does not represent my employer. As the publisher, I am not responsible for the comments section. Each commenter is responsible for their own posts.