Skip to 0 minutes and 13 secondsOkay, this was a brief introduction about how the disability research network using OHDSI network. And this is the study Chan performed. In this study, he tried to find out the impact of the initial combination treatment in hypertension. Here there is no gold standard for the combination treatment of the hypertension.
Skip to 0 minutes and 44 secondsSo he tried to compare the three group: ACEi/ARB, CCB, and Thiazide-diuretics (AD) Here, he used data. He, even don't have data but by using this kind of way of research, he could use that data. Like that this one. US, Truven MarketScan, it contains 135 million patient data and Optum data it contains 80 million patient data. Truven, this is about the Medicare. It contains 10 million patient data. And Medicaid data it contains 26 million patient data. And also he provides the NHIS 1 million patient data together. So for his study, he used 2.5? I don't know. 250 million patient data for his study, for free. And here he made the research analytic coders and released in public.
Skip to 1 minute and 47 secondsAnd this is the table one of the summary statistics of the data. And he got this result. This outcome was the primary endpoint what all cause of mortality and he couldn't find any difference between the three combination treatment in any of the databases he used. And for the myocardial infarction, he couldn't find any difference, but heart failure A plus D the combination is better than A plus C. And for the stroke, also A plus D combination was better than A plus C. And MACCE, there was no difference. And this is how the system works even, he doesn't have any data. There was no privacy leaking and he even didn't see any patient data, how it works.
Skip to 2 minutes and 31 secondsAnd this is another project that Chan had made. This software this is called the CIReNN. It is a kind of the AI tools that works on on CDM
Skip to 2 minutes and 46 secondsThis is the how that deep learning unit that works. And he made a package that works on normal CDM so by just providing several high parameters, you can train your own models for predicting some diseases by using the OHDSI data. So you do not need deep knowledge on neural network. This is the parameters he that required, for the running of these tools, like these. It's simple because it has basically that the polls values in there, like this. like this. Okay that's that's all while you you are what you have to do and the tools works on the CDM and it is tracts all the required data for your modeling and it's long against the data and it makes the models.
Skip to 3 minutes and 40 secondsAnd as you like, you see, you can see, that the training curves. And this is a model predicting the two-week mortality in hospital. Just he made a very crude model he got AUROC of 0.88. Very Nice. AUROC, so even even you do not know well about the technology of the models or AI algorithm, you can use this package. This is public and you can download it in the github with OHDSI. Is that right? Okay.
Two Example of Global Collaborative Research Using OHDSI Network
In this video, Dr. Rae Woong Park introduces a study using the OHDSI network. It’s a comparison of initial combination treatments in hypertension. He will explain the background information for the treatment of hypertension. Then he talks about the method by using different data sources
The whole process of analysis was packaged as software in R and released for reproducible research in Github. Another study he introduces is: Development of clinically Informing application using Recurrent Neural Network (CIReNN) based on the Common Data Model(CDM). This software, CIReNN is a kind of AI tool that works on CDM. You can see the experiment results that Dr. Seng Chen You has made.