This post is by Keith O’Rourke and as with all posts and comments on this blog, is just a deliberation on dealing with uncertainties in scientific inquiry and should not to be attributed to any entity other than the author.

Starting at the end of March, I thought it would be good idea to let folks here know about early research efforts being launched on Covid19 by OHDSI in their study-a-thon.

So almost six months later and we are where we are.

I was an observer on that study_a_thon and some of the later work done afterwards. I did not actually work with the group, but just watched and listened and so my views may be somewhat uninformed.

However, it occurred to me that many statistician might like to be more aware of opportunities that such international research groups might offer. Most I think work at a single institution where the analyses they get to be involved in have a single data set (at least at any one point in time), a single research group is involved in the project and with the unfortunate pressure to get something published in a journal. Eventually the research encounters the usual less than adequate peer review from journal editors and reviewers. Then if post-peer review occurs, some involved in the research “demand” the wagons to be circled and all members remain inside.

Well that was my career, at some places, when I was in academia. Also on the other side, those in the research group can only (easily) work with statisticians that are in their institution or otherwise available to them. Those statisticians may not have much expertise for what is specifically needed for their project. Or be able to easily draw on expertise of other statisticians.

Recently, in a series of talks by members from OHDSI at the virtual JSM2020, some real differences in opportunities to the above for statisticians seemed apparent. Briefly, rather than a single data set there are multiple sources of data sets, summaries of the separate results are made available for contrast and comparison, the researchers are often from multiple institutions around the globe, there is a methodological group that can be drawn on for specific expertise and code on github and peer review can potentially be part of the process by others in OHDSI not directly involved. Still there seems to be that unfortunate pressure to get something published in a journal, at least for many in OHDSI.

These may be just my impressions, but I think statisticians would benefit by knowing more about groups like OHDSI. I am sure there are more and I am expecting more in the future.

There are three talks at this link listed below. For those who signed up for JSM, the accompanying talks should be available until the end of August.

August 2020
2020 Joint Statistical Meetings
Session: The OHDSI Collaboration: Generating Reliable Evidence from Large-Scale Healthcare Data

Patrick Ryan – Janssen, Columbia University
The OHDSI Collaboration: Mission, Accomplishments, and the Road Ahead
Presentation Slides [Main message: Scientific harmony is achieved through collaboration, not randomization?]

David Madigan – Northeastern University
OHDSI Methods for Causal Effect Estimation
Presentation Slides [Main message: A new approach that is reproducible, systematized, open source and at scale?]

Marc Suchard – UCLA
Large-Scale Evidence Generation in a Network of Databases (LEGEND) Methodology and the Hypertension Study
Presentation Slides [Main message: A substantive case study?]