Update 4 - Another week, another preprint!
Another week, another Studyathon study available as MedRxiv preprint: Can we trust the prediction model? Demonstrating the importance of external validation by investigating the COVID-19 Vulnerability (C-19) Index across an international network of observational healthcare datasets. You can read the preprint here. External validation is critical for any prediction model, and through OHDSI analytics and global collaboration, this study suggests that C-19 should not be used to aid decision making during the COVID-19 pandemic.
Update 3 - Two more preprint articles have been published!
The first preprint, available here, indicated no clear increased risk of COVID-19 diagnosis, hospitalization, or subsequent complications for users of either angiotensin-converting enzyme inhibitors (ACEs) or angiotensin receptor blockers (ARBs) among a multinational cohort of more than 1.1 million patients using antihypertensives.
The second preprint is designed to inform individual behavioral choices and help design shielding strategies during de-confinement using a nine-predictor COVID-19 Estimated Risk (COVER) model that was validated using more than 43,000 COVID patients (following initial development and validation using more than 6.8 million patients with influenza or flu-like symptoms). This model predicts hospitalization, intensive services, and death, and can help provide reassurance for low-risk patients, while shielding high-risk patients, as many start to enter the de-confinement stage of the pandemic.
Update 2 - Preprint article on the characterisation of patients hospitalised with COVID-19
The preprint of the second OHDSI Study-a-thon paper is now available on medRxiv. Again, great to see many EHDEN colleagues contributed to this important and timely work in light of the ongoing COVID 19 pandemic!
The characteristics (demographics, prior conditions and medication use) of more than 6,800 COVID-positive patients from four databases (Columbia, Stanford, the Department of Veterans Affairs and the South Korean Health Insurance Review & Assessment) were reported in this study, and were compared to more than 52,000 patients hospitalized with influenza between 2014-19. Compared to 52,422 individuals hospitalized with influenza, patients admitted with COVID-19 were more likely male, younger, and, in the US, had fewer comorbidities and lower medication use.
Update 1 - Preprint article on the Hydroxychloroquine safety assessment
The very first preprint of the study-a-thon, on the preliminary findings from the largest study ever completed on the safety profile of hydroxychloroquine, has just been published. The paper has also been submitted to a high impact journal for peer review and many EHDEN partners have contributed to this study.
The preprint describes the safety of hydroxychloroquine, alone and in combination with azithromycin, using data comprised of 14 sources of claims data or electronic medical records from Germany, Japan, Netherlands, Spain, UK, and USA. The main finding indicates that short-term hydroxychloroquine treatment is safe, but addition of azithromycin may induce heart failure and cardiovascular mortality.
Peer review is currently ongoing, but this study will inform policy makers and healthcare professionals to be cautious if such combination is to be used in the management of Covid-19.
OHDSI uses healthcare data to inform COVID-19 healthcare decisions
- More than 330 researchers, from 30 different countries, fast-tracked research using healthcare data to inform healthcare decision-making.
- 37 healthcare databases with COVID-19 patient data participated, and this number is still growing.
- Two high-impact papers assessing the safety of hydroxychloroquine (the largest ever assessment) and characterising COVID-19 patients will soon be submitted for peer-review.
March 26-29, the Observational Health Data Sciences and Informatics (OHDSI) community, supported by EHDEN partners, hosted a COVID-19 virtual study-a-thon to inform healthcare decision-making in response to the current global pandemic.
In lieu of the cancelled OHDSI 3rd European Symposium , a core group of researchers from Oxford University (UK), Erasmus Medical Center (Netherlands), Columbia University (US), UCLA (US), Ajou University (South Korea), Janssen Research and Development, and IQVIA took the lead in organising this four-day online event. It was the core group who realised very quickly that this cancellation also offered a tremendous opportunity to rally the OHDSI community to contribute to the worldwide effort to combat this pandemic. The biggest challenge, however, was the very short timeframe to prepare as study-a-thons normally require many months of preparation and there were now only three weeks to get everything up and running. The OHDSI website, forums, Twitter and LinkedIn page were pivotal tools in this effort, rallying over 330 researchers with very different backgrounds and from thirty different countries.
“The OHDSI COVID-19 study-a-thon is a major milestone and serves as a great example of why collaboration in an open science framework is necessary. None of this work would have been possible if not for the diverse expertise of the participants, their willingness to share data, the availability of standardised analytical tools, and the OMOP common data model” said Peter Rijnbeek, Associate Professor Health Data Science ,Erasmus Medical Center, Rotterdam, The Netherlands and Coordinator of EHDEN.
Much of our current understanding of COVID-19 comes from a discrete set of questions asked on case report data. A key strength of the OHDSI community, however, is the ability to perform a rapid retrospective analysis of existing healthcare data across an international data network. Taking this into account, the aim of this study-a-thon was two-fold: 1. to generate immediate real-world evidence to inform healthcare decision-making now and 2. to design COVID-19-specific, standardised study packages that are pre-validated and can be run on COVID-19 data when such data becomes available.
The three weeks leading up to the study-a-thon were buzzing with activity, already involving a large part of the OHDSI community beyond just the core team:
- To ensure efficient communication, an MS Teams environment was set up to help discussions and file sharing. Interestingly, throughout the study-a-thon there were twelve global community calls to update the community on the progress, over 100 collaborator calls within the fifteen workstreams were held and over 13,000 chat messages were sent via MS Teams.
- The systematic literature review of more than 10,000 scientific publications was initiated by a group of volunteers, aiming to complete this task before and during the study-a-thon. In doing so, this information could inform the study designs for characterisation, population-level estimation and patient-level prediction studies, as well as manuscripts.
- The core team reached out to national governments, public health agencies and health-related institution to learn which research questions should be prioritised. Additionally, a forum thread was set up to collect relevant research questions from the community.
- A global survey was launched to identify and capture information on potential healthcare data sources containing COVID-19 patient data. An OMOP response team was set up to help institutions and organisations with converting their data into the OMOP common data model format.
- The vocabulary was improved to accommodate new concepts related to diagnosis and testing of COVID-19. In addition, several OHDSI analytic packages were further improved to accommodate the potential research questions.
- And much more…
The study-a-thon began on 26 March. Following an introduction to the concept and objectives of the four days by the core team, all 330 participants split up into one of the fifteen workstream groups for which they had signed up. There was a natural cross-over and collaboration between these groups as there were five major milestones that needed to be established for each research question so that the question could be translated into meaningful evidence:
- Review relevant literature and develop the study protocol.
- Develop and evaluate phenotypes.
- Develop the study packages (these are freely available via the OHDSI GitHub).
- Execute these study packages across the available data network.
- Review of findings by clinical experts and dissemination of the results.
The characterisation, prediction and estimation study groups were supported by a team focussed on defining phenotypes & cohorts. A literature review of thirty-six relevant phenotypes has led to the creation of 355 cohorts in Atlas, of which 114 were reviewed and validated for use in the study-a-thon studies The results of COVID-19 Cohort Evaluation are available via data.ohdsi.org and the ‘COVID-19 Cohort evaluation’ package is available via GitHub.
The characterisation studies descriptively summarise cohorts of influenza patients, or COVID-19 patients (both adults and children) as data becomes available. This includes, for example, demographics, prior conditions, drugs used, treatments and procedures during the thirty days after the initial presentation. A characterisation against stratified subpopulations at risk will also be undertaken to provide background context. Throughout the study-a-thon, research questions were converted into working study protocols and run across the network, already making some very preliminary results available before the end of the study-a-thon.
Of interest is the large-scale characterisation of COVID-19 patients in both the United States and Asia. Six databases with COVID-19 patients located in both the U.S. and South Korea already started running data on this project, and other databases are being sought to collaborate in this network study. A major advantage of the OHDSI approach is that on top of the peer-reviewed publication, all of these results are being made available online in an application which allows researchers to screen the results and which is being updated with new data as it becomes available.
The patient-level prediction studies will help to better triage patients by informing the discussion on measures taken to ‘flatten the curve”. These patient-level prediction studies are developed based on influenza data but will also be applied to COVID-19 patient data. To run these studies, the OHDSI patient-level prediction framework was used. Models and preliminary results are available via the data.ohdsi.org (COVID19). The questions we asked were: 1. Among patients showing COVID-19, influenza, or associated symptoms, who are most likely to be admitted to the hospital in the next 30 days? 2. Among patients at GP presenting with virus or associated symptoms, with/without pneumonia who are sent home, who are most likely to require hospitalisation in the next 30 days? 3. Among patients hospitalised with pneumonia, who are most likely to require intensive care or die?
And finally, the population-level effect estimation studies will examine the effects of hydroxychloroquine (HCQ), IL6 and JAK inhibitors, HIV protease inhibitors and hepatitis C protease inhibitors. While the initial efficacy analyses will be done using influenza as a viral model, the subsequent analyses will use COVID-19 patient data when this becomes available.
One of these effect estimation studies also focused on the safety of HCQ in Rheumatoid Arthritis patient data, as HCQ is an approved treatment for Rheumatoid Arthritis. This study was run over four databases from the USA, England, Germany and South Korea, and included more than 950,000 patients that have taken HCQ. To the best of our knowledge, this is the largest study to date looking at the safety of HCQ. The preliminary safety results can be seen at data.ohdsi.org and the study packages can be found on the OHDSI GitHub. The estimation of the anti-viral efficacy is awaiting COVID-19 patient data.
In addition, we sought to understand the implications (both susceptibility and severity) of the ACE-2 pathway and the use of ACE inhibitors and Angiotensin receptor blockers on COVID-19 incidence and complications. The impact of these therapies is currently debated, as SARS-CoV-2 targets these ACE-2 receptors. The result of this study will have a profound impact on patients currently taking this medication, as many are now stopping their treatment regimen in fear of it being a risk factor for COVID-19.
Throughout these four days, the team designed all of these studies and developed the necessary packages to run them concurrently in the OHDSI network. In total, thirty-seven data partners spread over ten countries and three continents had signed up to the study-a-thon, making their data available within the network. Eight of these data partners already had COVID-19 patient data available and this number is still growing. For about 35% of the data partners, the concept of a study-a-thon and the OMOP CDM were completely new and the OHDSI community rallied together to fast-track their onboarding and implementation of the OMOP common data model.
The ‘OpenData4Covid19’ initiative is driven by Ajou University (South Korea) and aims to make the South-Korean HIRA database, which has data on COVID-19 patients, accessible for research. For the purpose of the study-a-thon, the OHDSI community worked with the South Korean HIRA to run studies on its COVID-19 data as part of the OpenData4Covid19 initiative. They have also compared clinical outcomes between hypertension medications and the prediction of hospitalisation amongst patients with viral infection symptoms together with HIRA.
“It was a humbling effort to lead the OHDSI community in making a meaningful impact during this COVID-19 crisis,” said Daniel Prieto-Alhambra, MD, MSc, PhD, Professor of Pharmaco- and Device Epidemiology at the University of Oxford. “Prioritised questions from governments, health care agencies, and institutions helped direct our efforts, and it was inspiring to see how our community rallied together to make important progress on this research effort.”
“I am extremely proud to see what our community accomplished, but we are well aware that this is merely the beginning stage of a long research agenda,” said George Hripcsak, MD, MS, the Vivian Beaumont Allen Professor and Chair of the Columbia Department of Biomedical Informatics. “Our international network is committed to continuing work in this area until this pandemic has ended.”
- Recording of the kick-off call of the study-a-thon
- Playlist of the recordings of all intermediate calls of the study-a-thon
- Recording of the closing call of the study-a-thon
- Slide deck of the closing call
- Study-a-thon studies GitHub repository
- Preliminary study results in Shiny Applications
- OHDSI website page with all updates on the study-a-thon
OHDSI is a multi-stakeholder, interdisciplinary collaborative to bring out the value of health data through large-scale analytics. All solutions are open source. OHDSI has established an international network of researchers and observational health databases with a central coordinating center housed at Columbia University.
The European Health Data & Evidence Network (EHDEN) has received funding from the Innovative Medicines Initiative 2 Joint Undertaking (JU) under grant agreement No 806968. The JU receives support from the European Union’s Horizon 2020 research and innovation programme and EFPIA.
We are grateful to Erasmus MC, IQVIA, Amazon Workspaces, EvidNet and Odysseus for providing infrastructure, software and support which made this study-a-thon possible.