Interview on research project: "Electronic patient file not a savior"

The electronic patient record (EPR) is seen by some as a "game changer" in the healthcare sector. Thanks to artificial intelligence, rapidly growing volumes of data should enable "in-depth analysis of individual health data" and provide new and previously inaccessible insights at population level, according to a press release from the Technology and Methodology Platform for Networked Medical Research (TMF). "In order to exploit the potential of AI, we need to tap into existing data sets such as real-world data, ePA data or data from registers," said Sebastian C. Semler from the TMF. In order to achieve this, experts such as Prof. Ariel Dora Stern from the Hasso Plattner Institute are calling for a "greater willingness to take risks and more competition for the best AI tools".

Just recently, Susanne Ozegowski, Head of Department for Digitalization and Innovation at the BMG, also spoke about the opportunities of electronic patient records, the Health Research Data Center and the European Health Data Space in the podcast "Healing with Data". However, it was not clear from the podcast episode how data can heal. On the other hand, the electronic patient record did not receive any attention with regard to research in the latest statement "Effective prevention needs reliable health data" from the Federal Government's Expert Advisory Council.

We spoke to Prof. Jürgen Windeler for an assessment of the hopes associated with electronic patient records and their interaction with AI. He is the former head of the Institute for Quality and Efficiency in Healthcare (IQWiG) and a proven expert in the field of clinical epidemiology.

Will the electronic patient record be a savior?

As far as communication is concerned, an ePA could bring benefits, although many processes are already digitized in practices. Even without the new ePA, there are already practices where everything is completely electronic.

However, it is unclear to me where the "dramatic improvements in care" promised by the Federal Minister of Health, Karl Lauterbach, are supposed to come from. Certainly not through the ePA. The new ePA is completely unsuitable as a savior.

Will the "ePA for all" enable personalized medicine?

Personalized medicine is already possible and is practiced where it makes sense. The ePA is not needed for this.

Based on research projects that are becoming possible more quickly, significantly accelerated approvals are also being considered, such as by the US Food and Drug Administration (FDA). What is your assessment of this?

It is obvious that the ePA is useless for research projects with drugs that are not yet on the market and therefore cannot appear in the ePA. However, the trend towards ever earlier approvals, which can only mean with increasingly immature data, is fundamentally not a good idea.

Can the ePA prevent drug scandals such as the Lipobay scandal or other incidents based on drug interactions?

Certainly not. One could imagine that such correlations would be recognized sooner, but that is grey theory.

It is often said that Germany does not have its own data, but is dependent on data from other countries. Is that true?

Germany is indeed the beneficiary of studies, but not of data. This has something to do with the fact that clinical research in Germany is underdeveloped, to put it mildly. It is often said that many studies are carried out in Germany, but in relation to the size of the population, this is not true – and we are a distant second. And it's often just about taking part in pharmaceutical industry studies. This has little to do with our own research. Unfortunately, Germany does not have a good culture of clinical research.

All the studies that have been carried out on mammography screening, for example, come from abroad. German system decisions on the introduction of screening or, more recently, on changes to screening are based on these studies. In this respect, the benefits are true. However, this is not because we have not yet had an electronic patient file in which a lot of data is collected. It's because we don't have a culture of research. The EPR data will not improve this. Research that improves care and is internationally competitive requires specifically collected, quality-assured information, not just any old pile of data.

The billing data is already available at the Health Research Data Center. Is this data suitable for research?

Data from the system could help, for example, if you want to evaluate the consequences of legal regulations or G-BA decisions (editor's note: the Federal Joint Committee (G-BA) is the highest decision-making body of the joint self-administration of doctors, dentists and psychotherapists). There are also other research questions, such as the frequency of a particular diagnosis, the incidence of depression in Germany. This would make it possible, if at all, to describe the status quo. However, it is well known that billing data is poorly suited for such research because it is naturally documented with a specific objective in mind, and this objective is called "billing".

It is at the discretion of the doctors to write down certain things or not. In particular, the question of payments in the DRG system in hospitals depends on documented diagnoses. In the outpatient sector, billing information can have something to do with the fact that a patient is only reimbursed by the health insurance company for the costs of a medication in connection with a specific diagnosis. Billing data is optimized for billing modalities. You can't do proper research with that.

What do researchers need to bear in mind when working with ePA data?

The EPR data is subject to various distortions. It starts with the fact that the data of privately insured persons, i.e. ten percent of German insured persons, are not included in the PPR and that there are no plans (for the time being) to pass on the data of privately insured persons in the future.

Then there is the fact that the people who have the ePA can deselect information from the ePA. So you get a bias in that, for example, all or half of HIV patients don't want their HIV diagnosis to be in the EPR.

Researchers will therefore have to work with incomplete, biased data, a situation that is best avoided in research. It is then said that this is "exploratory research". But this only means that people feel entitled to publish unreliable results and leave it to others to check them using careful methods. This will certainly not make Germany the research world champion.

In debates on the subject, AI has also been compared to the pocket calculator or other technical achievements. After all, people also use things that they don't understand, such as a computer. What do you say to that?

No pocket calculator has been given such adventurous promises as AI or the ePA. Recently, for example, you could read that AI can increase life expectancy by 1.5 years. Strictly speaking, that's quite a small amount –.You could extend it by 5-10 years by giving up smoking. But a colleague commented on this aptly: If he looked at all the innovation announcements of the last 30 years, we would all have to be 150 years old by the time it all came to pass. The dramatic improvement through ePA that Mr. Lauterbach is talking about will not happen.

Against this background, I find it very questionable that there is an objection-based solution for the electronic patient file, i.e. you get it "automatically" if you don't object. There is nothing fundamentally wrong with an EPR, but the German concept would have to change drastically for an opt-out regulation to even be considered. The issue of research cannot currently be used to justify an opt-out.

What do policyholders need to know in order to make a decision?

They need to know that many of the benefits highlighted are nothing more than vague announcements. This applies in particular to health improvements and research results. And comprehensive and comprehensible information must be provided about the risks –, primarily about who can gain access, as well as about the disclosure and security of the data. It is almost a joke that, on the one hand, people rave about the great capabilities of AI and, at the same time, are assured that pseudonymization will be "absolutely secure" for the next 100 years.

Health insurance companies should actually explain the risks, just as banks do with investments. Only then should an electronic health record be created with the consent of the insured person. However, the consent should specify over what period of time which data may be collected and passed on, to whom and for what purpose. This would be a scenario that meets the legal and ethical requirements that apply in other, less sensitive areas of business. The fact that the data is available for 100 years is probably not known to the public.

Most people with statutory health insurance are probably not aware of this either. Ms. Ozegowski recently announced that there are no budget funds available for an awareness campaign on the subject of data extraction. The majority of the population will not be able to cope with the authorization management for accessing data. At the same time, the health insurance funds point out that insured persons should ensure that their file is complete and call on them not to exclude too many service provider institutions from their patient file.

Yes, an interesting message: insured individuals should please not make use of the decision-making options laid down by law. It is incomprehensible to me why there is no audible criticism from the German Medical Association or its ethics committee or from patient organizations. This applies to various provisions of the "new" ePA, but there is one point where all the warning lights should actually go on: the relativization or violation of the duty of confidentiality.

Let's assume we have all the health data at the FDZ. Now AI models from Google, OpenAI and the like are to be trained with it. In what way does this ensure medical progress that is based on research for the common good?

I have no idea. The common good was removed in the course of the legislative process. Now anyone whose imagination is sufficient to create a "purpose" can "research" with the data. Perhaps this will secure jobs for companies, but the extent to which these activities will result in health benefits for insured people in Germany remain to be seen.

(mack)