Another update on the OHDSI community efforts to support and inform the Covid-19 pandemic response. Happy St. Patrick’s Day! I hope those of you observing this holiday enjoyed some tasty beverages responsibly…at least 6 feet from your fellow celebrants.
Over the last couple weeks, I received some questions about ‘what exactly is OHDSI doing for Covid-19?’ and ‘why are you waiting until March26 to being work on analyses for Covid-19?’. So let me try to address those directly:
‘What exactly is OHDSI doing for Covid-19?’: Put simply, our community is trying to do whatever we can to support efforts by national governments, public health agencies, and health-related institutions who are seeking reliable real-world evidence to promote better health decisions and better care in the current environment. OHDSI collaborators have reached out to various stakeholders in US, Europe and Asia to ask what open questions they need answered, and have volunteered to provide scientific support throughout the observational analysis lifecycle. Many folks in the community have contributed to the OHDSI forums discussion thread about research questions that we can potentially address together as well. Our explicit intention is to identify and prioritize these questions, design and implement observational analyses to answer the questions, execute the analysis packages on as many participating databases as possible, and then disseminate the results to everyone who can benefit from the findings. My hope and expectation is that we all focus and deliver on a core set of high-impact public health problems (as I briefly outline in my prior update post), but that we also use this opportunity to come together to collaborate on other valuable questions identified by the community and/or to support the needs of the community.
‘Why are you waiting until March26 to begin work on analyses for Covid-19?’: A lot of different people are doing a lot of different activities, but ‘waiting’ certainly isn’t one of them:) Since we announced that OHDSI was committed to doing whatever it could to support the Covid-19 response, several of us have been meeting on a daily basis to identify specific actions that we can take to advance the science and support those in need.
Some examples just over the last 3 days since my last post:
- @schuemie has been in rapid development mode on the CohortDiagnostics package to support the community’s phenotype development efforts.
- @jennareps has been improving the PatientLevelPrediction package to make it easier to develop and apply ‘existing risk scores’ by allowing for custom features to be based off ATLAS-defined cohorts.
- @Christian_Reich , @Alexdavv , @Dymshyts have been improving the vocabulary to accommodate new concepts for diagnosis and testing associated with Covid-19, and expect to have a release in the next couple days that should contain the necessary content.
- @hripcsa , @Daniel_Prieto , @Rijnbeek , and I have been reviewing the questions that are coming in from our stakeholders and on the forums, translating the evidence needs into potential real-world data analyses, and evaluating the feasibility of carrying out various characterization, estimation, and prediction studies across the OHDSI network.
- @Albert_Prats created a series of conceptsets to define drugs that are actively under consideration as candidate Covid-19 treatment
- Several groups have been working diligently to refresh their ETLs with more recent data. I want to give a big shout of to @JamesSWiggins and the AWS team, who have graciously provided support with technical infrastructure to make large-scale compute jobs more manageable for teams across the community.
- @aostropolets has been collating and aggregating results from the OHDSI ConceptPrevalence study. Thanks to those who have participated, and another plea for those who haven’t to please run the analysis and share back your results!
It’s amazing to me that we now have 150 collaborators who have signed up to participate in our OHDSI virtual study-a-thon. Many of you self-identified as interested in supporting literature review and evidence synthesis, and this is clearly an area where we can get started right away. So, today, we kicked off 5 sub-teams to tackle specific review activities which will lay the foundation for our future OHDSI network analyses.
phenotype definitions and validation - review the literature for prior observational studies in claims/EHRs which have studied viral diseases, symptoms, and complications, and extract the phenotype algorithms (e.g. codesets) and any validation statistics that are available. Co-leads: Matt Spotnitz (@mattspotnitz) , Ray Chen (@rchen), Gowtham Rao (@Gowtham_Rao)
prediction models for pneumonia - review all models around viral outcome prediction, including community-acquired pneumonia, that have been performed in claims/EHR/registries (or potentially other sources where observational data could be used). extract the database, T, O, TAR, features, performance measures, and validation approach. Co-leads: Cynthia Yang (@cynthiayang) , Ross Williams (@RossW)
efficacy/safety of candidate treatments - hydroxychloroquine, tociliuzumab, remdesivir - we want to summarize what is known about these products for efficacy/effectiveness for viral diseases, as well as safety, overall and within a subpopulation with viral disease. Co-leads: Nige Hughes (@nigehughes) , Dani Prieto-Alhambra (@Daniel_Prieto)
characterization of covid19 cases: Several papers are coming out that provide baseline characteristics of Covid19 cases, some stratified by death/recover. We want to extract population (e.g. sample size, geography), which baseline features are assessed, % overall and within strata, and how ‘risk factors’ were identified and modeled. Co-leads: Ben Illigens, Denys Kaduk (@Denys_Kaduk)
what trials are underway to evaluate new treatments for Covid19? We want to be actively monitoring what drugs are under consideration for either direct treatment of disease patients or prophylaxis treatment, because these can be the targets for our estimation studies. Registries of ongoing trials or other work should be summarized and updated. Co-leads: Ana Szarfman (@szarfman) , Jon Duke (@jon_duke)
If you had registered for the study-a-thon and completed the literature review questionnaire, then you should have received an email from me ‘assigning’ you based on your preference to one of the 5 groups above. If you didn’t get an email, but want to join one of these 5 efforts, please feel free to direct-message the co-leads. Also, you’ll see new OHDSI forum threads where you can follow the activities of these groups.
Amidst the flurry of activities, I have not been communicating as effectively as I should have been, so I apologize for that. Thankfully @CraigSachson has volunteered to help me with that moving forward. He’ll be giving more regular updates on our community activities, so stay tuned into the OHDSI forums, @OHDSI on Twitter, and LinkedIn.