Data in the healthcare industry – an essay on potential and challenges
Data are the raw materials of the 21st (our!) century – one of the most frequent things people say about big data. But what actually are big data? And what do big data mean for healthcare services? Steinbeis experts Dr. Martin Vogel and Jürgen Blume of the Steinbeis Research Center for Medical Technology and Biotechnology talked to TRANSFER magazine.
If there’s one thing the two Steinbeisers agree on, it’s that the term big data is not just a reference to size. Huge volumes of data have been a common phenomenon (and challenge) in imaging science and medicine for decades. The term big data is more likely to be used to describe the phenomenon of continually gathering big volumes of data and the associated requirements this places on system design, storage systems, analysis, and displaying information. One well-known example of this is Google Maps. Maps uses positioning data supplied by countless cell phones to provide information on current and forecast traffic flows on certain days of the week or times of day.
There’s now a term for the experts who deal with all the technical issues relating to big data: data scientists. Data scientists work on a number of challenges:
- Selecting appropriate system topologies for certain issues (such as which system components should take on which tasks – a central computer, a peripheral computer, a cell phone etc.)
- Proposing storage models that also take future system developments into account (e.g., normalized storage in SQL databases or document-centered storage in NoSQL databases)
- Developing suitable evaluations (e.g., traditional analysis models, self-adapting, self-learning algorithms, or artificial intelligence)
Artificial intelligence: opportunities and threats
Dr. Martin Vogel and Jürgen Blume believe that AI algorithms offer the biggest application opportunities for big data. A computer system can access huge volumes of data – expertise based on experience – and use this to make predictions. How successful every new decision made by the system was is then used to optimize decision-making. This results in a permanently improving system, which becomes much more accurate at making predictions than classic models. In individual cases, systems used in specialized fields are already even “better” (in other words: more accurate and quicker) than human experts.
At this point, however, one needs to consider the risks: The decision-making chains or “trees” used by AI algorithms are sometimes huge (millions of mutually dependent individual decisions) and become impossible to be understood by humans. In other words: In most cases, we don’t know what computer systems base their forecasts or decisions on. This can be particularly worrying if a system has to make a decision about an “unusual” marginal case (in terms of stored experiences).
As a result, the two Steinbeis experts believe that it only makes sense to use AI systems in healthcare if they play more of a supportive role – by which they mean the final decision should always be made by an expert (although mechanisms need to be found to stop people getting into the habit of continually “clicking through” options), or risk assessments should be used to ascertain if an autonomous system could cause irreversible harm to life and limb.
Raw materials and data protection
If such factors are taken into account, AI algorithms can be extremely helpful in revealing unknown connections within data. There is a hitch, however: Some struggle with the concept that data are a raw material. Unlike traditional raw materials, data are not something that is simply gathered and sold, especially in healthcare services, where they actually belong to people – usually patients – and there’s every possibility that they are also owned by other key stakeholders, such as medical insurance companies, medical staff, or other service providers. Legislators may even have an important influence, albeit unintended, if new laws are badly formulated.
In any case, it’s not analyzing data that becomes interestingly challenging, but clarifying with certainty to whom which data belongs – and whose permission is required to use data.
This is a direct consequence of data privacy rules, especially EU general data protection regulations introduced in May 2018. Ultimately, this takes us back to the previous situation under German data protection law, seen by many as the strictest in the world.
From a general standpoint, the Steinbeis experts believe the goal when using big data for any kind of purpose must always be that people are understood, as well as their intentions, and how data affect (or will affect) them. In addition, especially in healthcare services, it must be ensured that corresponding measures introduced for legal reasons – or to give reassurance – are not merely given consideration as an afterthought. Instead, they should be put in place as soon as possible.
Thinking about dig data projects from this angle – in terms of the technology, but also legal and social aspects – and adopting an overarching approach from the beginning, is certainly a challenge, but in the end it’s worthwhile!