Article content
Artificial intelligence in healthcare was one of the topics discussed at last year’s Tallinn Digital Summit. Estonia has implemented several e-solutions and has a vibrant health and wellness start-up scene which has helped to identify many challenges in developing digitalisation.
I was able to talk about the opportunities and challenges of developing new e-health solutions with Cardiac Intensive Care Unit cardiologist at North Estonia Medical Centre and e-health expert Dr Eno-Martin Lotman. He is a co-author of a recent article on the same subject published in “Cardiology”.
I asked him to name the main challenges in digital health. “Too little data and financing,” Dr Lotman listed. “And too much data,” he added to my puzzlement.
Can e-health really help the patients?
“We don’t have enough data on the effects of several otherwise promising ideas,” Dr Lotman explains. “When we are developing new e-solutions through start-ups, the business model hopes to be lucrative, but the primary requirement for health care professionals is that they actually make the life of patients better. In healthcare to have this kind of evidence requires meticulous testing, but testing for e-solutions can be tricky for several reasons”.
One of the reasons according to Dr Lotman is that digital health services are often “auxiliary” services in combination with others, such as video consultations or digital medical history. This may render pinpointing the benefits of digitalising difficult. Another set of difficulties arises from the lack of cooperation between developers, healthcare professionals and scientists.
“Often people who go into health start-ups are from outside of the medical field. When they run into the testing requirements in medicine this overwhelms them. Many start-ups have dwindled because of this. Hence we need funding for cooperation between them and other specialists,” claims Dr Lotman.
He points to positive developments in Estonia, where advancing cooperation and testing is one of the tasks for newly founded Innovation Foundation at the Estonian Health Insurance Fund. The first call for projects is for treatment of stroke patients with the aim of measuring the influence of novel treatment solutions to patient quality of life.
Reimbursing for digital health
The issue of financing of e-solutions is closely connected to evidentiality. This concerns both newcomers to digital health as well as reimbursing for applications in use. In their article, Dr Lotman and Dr Margus Viigimaa argue that there should be a system in place that both adequately reimburses for digital services (which may include such things as licensing of software or updating it), and also protect from innovations that do not provide expected results. Developing financing models is also one of the tasks for Innovation Foundation. Dr Lotman brings a simple example:
“If instead of in-house visit the doctor and patient meet online through a designated video platform, how should it be billed? It would be logical to assume that these video platforms could save a lot of time and money in sparsely populated areas, but we have to put it in numbers. We have to make sure that when we develop these services we are not lacking in quality elsewhere, such as the thoroughness of check-ups.”
Too much data?
“As in many fields nowadays,” Dr Lotman elaborates “the problem is the overload of unstructured data. For example, one bed in intensive care creates 1 GB of data daily, and wearables may provide us with months of pulse rate and movement data for thousands of individuals. A doctor may look at some numbers and use them in treatment, but the rest is actually put in no good use.”
Dr Lotman identifies two interrelated directions for advancement: interface-based structuring of the data and AI-based analysis. “In cardiology, for example, we use a lot of monitoring devices. These devices rely on algorithms that give rough boundaries for physiological processes. When these are exceeded – for example the heart rate goes too high or low – the monitoring instrument flags it as red. The problem is that in the end we have too many red flags, but we don’t know what actually happened. Therefore, data collection should be integrated with the patient either though voice-based or a visual interface, so that we could have some feedback about activities or how the person was feeling at the moment. This would give the AI cues for reference.”
For the AI -based analysis, Dr Lotman brings again the example of digital medical history, widely use in Estonia. “For each patient this combines information from all clinical tests, treatments, prescriptions. But this is largely unstructured and may be in text form. It can be tens of pages. So, similarly to monitoring data, AI could really help doctors with analysis.”
Could we have too much digitalisation, I wonder? Dr Lotman thinks not, but adds: “In the end it is the doctor who needs to make decisions. Because more often than not there are no simple “right or wrong” options we need all the help we can. Even from an AI.”