Our Posts

Is privacy a problem for clinical research?

No Comments

One of the most common questions I get asked around data privacy are whether patient privacy is inhibiting medical advancements and is it justified? When I get asked this question I know the person asking it expects either "yes" and "no" respectively, or "no" and "yes". Whichever it is, they are ready to argue passionately for their position. I can't blame them. More often than not, they are surprised by my answer: Patient privacy is not the problem preventing advancements in medicine.

While there is no doubt that data is essential to medical research, and that personal, identifiable information should not be shared, the actual inhibitors to research are not privacy. Removing identifying markers from data is not trivial but also not hard. Different jurisdictions have different rules about how to take data out of that jurisdiction but these are not a real problem because most of the time, data can stay where it is.

So what is actually inhibiting medical research?

Consider that we only need data about the past when it predicts the outcomes of future behaviours. To know if predictions made with data really do fortell the future, we need to try them on several independent datasets, and to see that they all give us the same results. This requires collaboration, but collaboration doesn't mean sharing data.

The problem with reproducibility is that in order to agree that one analysis is indeed a reproduction of another, we need to agree on the underlying data. The underlying data has a format and historically, no two hospitals agree on what that format should be. Today we have standards to help with that, and the biggest and best standard of them all is OMOP.

Transforming data from any format to OMOP is hard when done manually, but Evidentli uses a number of AI algorithms to simplify this problem, taking away the risk, and the cost, of manual (or even mostly-manuyal) transformation. Our algorithms do that up to 50 times faster and around 6 times more precicely than people do on their own. The best results are achieved when people and AI work together and we see this over and over again.

When the data is standardized, there is not need to share it in order to reproduce research. The analyses themselves carry no patient information (identifying or otherwise) and so has no restrictions. Data standardisation avoids the problem of patient privacy. What is inhibiting research on standardized data? Exactly nothing.

Recent Posts

Leave a Reply

Comments, questions or queries? Let us know!"