In light of the current uncertainty around COVID-19, we have decided to cancel the in-person OHDSI EU Symposium, which was scheduled to take place 27-29Marc2020 in Oxford, UK. An announcement about this is available here. In lieu of this large meeting event, we have decided to coordinate a virtual OHDSI study-a-thon to take place over the same period that will be focused on generating real-world evidence that can inform the current COVID-19 pandemic response.
We expect there are many ways the OHDSI community can contribute to the current situation:
- Characterisation of symptoms and complications of viral diseases
- Prediction of adverse outcomes amongst patients with virus-related hospitalization
- Comparative safety of treatments being considered/used for potential use in COVID-19
Some of these questions can be immediately addressed with historical data from administrative claims, electronic health records and other sources already available to our community. For others, we can be prepared to validate any findings in databases as COVID-19 cases become available.
Over the past several days, several of us have been reaching out to colleagues at CDC, FDA, EMA, NHS, as well as academic experts across US, Europe, and Korea to understand the current evidence gaps and how retrospective analysis of existing observational data could meaningfully inform public health decisions. While the vast majority of the OHDSI community does not have real-time data collection to support COVID-19 case detection, we have identified several specific questions for which OHDSI network analyses of existing data can meaningfully inform. To highlight a couple examples:
- In parts of Korea, some hospitals have reached capacity, in overall bed or within their ICUs, which has resulted in symptomatic patients being turned away without medical care, some of whom subsequently died at home. Therefore, hospitals are looking for more effective ways to triage patients to determine how to prioritize their limited resources. Drs. Seng Chan You and Rae Park (Ajou University; @SCYou @rwpark ), in collaboration with these hospitals, has proposed an OHDSI network prediction study to use influenza as a model viral infection, and to create a risk score for flu-related complications (pneumonia, ICU use, oxygen and ECMO therapy) and mortality that could be used to determine which patients are at highest risk (reaching a greater level of specificity than the current US CDC guidance that ‘Older adults and people who have severe chronic medical conditions like heart, lung or kidney disease seem to be at higher risk for more serious COVID-19 illness’. While we will initially train this prediction model using flu models, Dr. You is working with Korean government officials to potential access national claims data in less than one months time which will allow us to apply and validate the prediction model on COVID-19 cases, and consider how this model can be used to supplement current hospital admitting guidelines across the OHDSI Korea network of hospitals.
- Dr. Marc Suchard (UCLA; @msuchard ) is a leader in the evolutionary biology community examining COVID-19, and has said that folks within his community particularly concerned about long-term sequelae of aggregative treatments, including ECMO and extended RICU stays, as well as safety of antivirals that are being considered candidates for widespread prophylaxis. While OHDSI will not have data to examine effectiveness of treatments on COVID-19, we can look to characterize the real-world experience of patients who have received these treatments for other viral diseases.
- Dr. Dani Prieto-Alhambra (Oxford; @Daniel_Prieto ) said colleagues he contacted in Spain are experimenting with use of hydroxychloroquine as a potential treatment, with little evidence to support its application beyond a biologically-plausible hypothesis. Given that we examined the comparative effectiveness of hydroxychloroquine vs. other DMARDs in our RA study-a-thon, we may be able to re-purpose our prior study to examine effects on the incidence of other viral diseases. Dani was told such as a study would directly impact ongoing research into these treatments.
- Multiple groups have expressed that they do not have sufficient confidence in the use of claims and EHRs to study the common symptoms of COVID-19 (including fever, cough, dyspnea, malaise/fatigue) or complications (including pneumonia or acute respiratory distress syndrome). Developing and evaluating phenotypes for the constellation of diagnostic patterns observed during prior viral outbreaks, including the past several flu seasons, could provide important context for understanding the background rate of these symptoms, which can be helpful to guide policy and risk communication. FDA and CDC have both said they will come back to me with more specific analysis needs in anticipation of the OHDSI event.
Many of you in the OHDSI community are well-connected with other public health officials and may have other insights into important questions that we can answer using the OHDSI data network. We need to use the breadth of the OHDSI community to reach out to those who can benefit from reliable real-world evidence to determine their specific needs and how that evidence could be used to impact their decision-making. We also need the depth of expertise across the OHDSI community to take these important public health questions, and translate them into scientific best practice study designs and analysis source code, and we need the power of the OHDSI international data network to execute these analyses and share aggregate summary results so that we can disseminate reliable evidence to those who need it as quickly as possible. I personally feel very compelled to do whatever is in my capacity to help the current COVID-19 epidemic; these are exactly the situations when we have to bring together the best talent and science to do whatever we can for the patients we serve.
How will the virtual OHDSI study-a-thon work?
The study-a-thon will take place from 26Mar2020-29Mar2020. We will form the virtual core team, which will include myself, Dani Prieto-Alhambra (Oxford), Peter Rijnbeek (Erasmus MC @Rijnbeek ), George Hripcsak (Columbia @hripcsa ), Christian Reich (Iqvia @Christian_Reich ), Martijn Schuemie (Janssen @schuemie ), Seng Chan You (Ajou), Rae Woong Park (Ajou), Marc Suchard (UCLA) who will lead the OHDSI community effort. We will be meeting via TC/web conference with daily sessions planned to work for every timezone to accommodate participation around the world, with particular emphasis on supporting the needs of our colleagues in Korea. We will then have remote sites across North America, Europe and Asia-Pacific regions that will participate through web conferencing and though public exchange of study documentation, analysis code and results. We encourage everyone to be responsible and not incur any unnecessary risks to travel, and expect that some individuals may opt to participate from their home, while others may congregate in small groups within their respective institutions.
What can you do right now?
If you are interested to participate in the virtual study-a-thon, block your calendar from 26Mar-29Mar2020. Please post your willingness to participate as a citizen scientist by filling out the Google form here, providing your contact information and sharing what you think you can contribute during the study-a-thon.
If you have access to patient-level data that is formatted in OMOP CDM format, and would be willing to execute OHDSI network analyses against your data and share back aggregate summary results to the global effort, please post your willingness to participate as a data partner in this Google form here. You’ll be asked for contact information and high-level overview of your database. For those who agree to participate, we’ll send out a small R script that characterizes your data, in terms of population size, years of data capture, longitudinality of follow-up, data domains covered, and vocab version.
If you have research questions to support the COVID-19 response that you think the OHDSI community can potentially answer, then post them on the OHDSI forums. I will start a thread that you can add to for this purpose. For your question, please answer:
What is the decision we are trying to inform?
Who is the decision-maker?
What type of real-world data is needed to generate reliable evidence?
How will reliable real-world evidence inform the decision?
(we’ll presume we know the answer to ‘When is the evidence needed?’ = as soon as possible!)
We will be using the responses on this thread to prioritize our collective efforts during the virtual study-a-thon event.
- If you know of others who could either potentially benefit or meaningfully contribute to our OHDSI community efforts, please send them this forum thread and encourage them to ‘join the journey’.
Thank you in advance for coming together for this important effort. Let’s use this challenge time as an opportunity to re-affirm what OHDSI is all about: to improve health by empowering a community to collaboratively generate the evidence that promotes better health decisions and better care.