This webinar was the first in a series of webinars organized within the scope of the EHDEN-project. The European Health Data and Evidence Network started in 2018 with 22 partners in 12 countries aiming at mapping 100 million anonymised European health records datasets to the OMOP Common Data Model.
In this webinar, Kees van Bochove (CEO and Founder of The Hyve), interviewed Daniel Prieto-Alhambra (University of Oxford). They discussed the results of the study-a-thon EHDEN and OHDSI held at the University of Oxford last year. At this event a group of 40 scientists used an open science approach via OHDSI and produced a high-impact observational study in five days predicting the outcome of a randomized-controlled trial (RCT) that has been running for a number of years.
Highlights of the webinar
Kees: Today we will dive into an example that stems from Daniel’s practice. Among other things, Daniel works as an academic clinician at the Nuffield Department of Orthopaedics, Rheumatology and Musculoskeletal Sciences (NDORMS). Every week, he sees a number of patients in the Rheumatology department. So as an academic clinician, what do you think of the interaction between research and clinical practice?
Daniel: Research should inform clinical practice in a number of ways: providing the tools to identify patients at risk of having a health problem, informing on which medications are effective in preventing or treating a given health disease. For example, RCT is the golden standard to provide evidence that a given intervention really works. Ideally, that is paired with observational studies, so we can know how the patients react if they take drugs in the real world and in the hospital. There are RCTs that are pragmatic and run in a well-controlled environment, with a population that is more likely to be similar to the one that we will encounter in the real world. But in other trials, especially in the early stages, the groups of patients could sometimes be more diversified. The end goal is really to give patients the best treatment all the time.
Kees: Could you explain the concept of a study-a-thon? And what did you do in Oxford?
Daniel: This experiment has been done before in the USA and it was hosted for the first time in Europe in Oxford last December. It is an interesting socio-technical experiment: people with different skills and knowledge backgrounds get together for a week and try to find the answer to a clinically relevant question. From formulating the research question to having the final answer, written up in a formal manuscript to be submitted to a journal in the following days. It is an intense experience, trying to make a significant advancement in a very short term, using only advanced tools that are validated for that purpose − not testing new tools, not even mining new data.
Kees: How much data is really needed if you want to run an observational study?
Daniel: It depends. One of the key parameters to consider is the statistical power of the study which depends on a number of factors. For example, how many people are exposed to your drug, surgery or procedure of interest and how many people have the outcome of interest? Secondly, you need data to cover your statistical power but also ideally a list of different data sets to show how robust your findings are. If you are able to replicate a study in a number of different datasets, your findings can be proven to be very robust. Leveraging OHDSI, we can run the same analysis in different databases, provided they are mapped onto the same Common Data Model (CDM).
Kees: So how could the conversion from source data to OMOP potentially affect your research and the dynamics of the study-a-thon?
Daniel: For the study-a-thon we had data that had been previously mapped. I guess something that is relevant to mention is that the way we worked within EHDEN with mapping data is that data will stay at source and we will generate an ecosystem for these kinds of studies to happen rather than accessing data remotely. This enables a federated structure where you can analyze data in a number of different places.So everyone could go from the raw data to the analytics part independently, yet share the results within the group. This really helped speed up the outcome. Another interesting element of this collaborative approach was that during the study-a-thon we had Patrick Ryan from Janssen Pharmaceutical (founding member of the OHDSI community) teaming up with us: he would post questions in the OHDSI forum and people from remote locations would give their input, based on their different skill-sets and knowledge. We even had one researcher contributing from Korea!
Kees: For the success of the experiment, how important was it to have everyone in the same physical location?
Daniel: It was really nice to have all these people working on their own piece of the research all in the same location. We definitely can discuss if such an event could be run in a more efficient, more cost-effective way, for example within a virtual setting. At the same time, we cannot underestimate the importance of face-to-face interaction, for example when deciding on what code to use.
In the webinar, Daniel and Kees touched a number of other topics, including the value of the OMOP CDM and data standardization in clinical research.
The EHDEN project will be a great opportunity for anyone who wants to run a study in an Open Science manner. There are grants available to engage with SME’s in this mapping exercise to start harmonizing research data. As the leader of the Technical Work Package in EHDEN, The Hyve provides training and expertise to all the parties involved. In fact, our second appointment − a workshop hosted at The Hyve’s office in Utrecht on April 28, was already a technical training to learn about the architectural aspects of implementing OHDSI suite. The second EHDEN webinar is also scheduled at 16.00-17.00 May 22 (CET) to discuss how EHDEN will transform the way real world evidence is being generated. Register here.
Stay tuned to hear more!
Watch the Webinar again: