A secure analytics platform is now running inside the data centres that hold GP records for 27 million patients, soon expanding to 55 million, to answer urgent questions about Covid-19 without copying sensitive data elsewhere. The problem is speed and privacy. During a pandemic, researchers need rapid answers about who is most at risk, but moving millions of patient records to a central database creates security risks and delays. OpenSAFELY solves this by bringing the analysis to the data, not the other way around. The platform is already producing results. If sustained, this research will identify which patients face higher risks of ventilation and death—including whether higher risk among BAME groups stems from pre-existing conditions alone. It will model local disease spread to guide lockdown exit strategies. It will also measure indirect harms, such as missed cancer diagnoses and disrupted cardiovascular care during the pandemic. The findings could reshape how the NHS prioritises shielding advice, allocates intensive care resources, and catches up on delayed routine care across every local area in England.
View original technical description
We are delivering a platform for analysing healthcare data, currently including 27 million patients' GP records (with identifiable information removed), soon expanding to 55 million. This unprecedented scale is needed for rapid results during Covid-19. OpenSAFELY uses a new model for maximum efficiency and privacy: rather than copying large volumes of information from GP records into another location, which involves security risks, we have instead built a secure analytics platform inside the data centres where GP records already reside, and linked to all relevant Covid-19 outcomes data such as death records and intensive care. We have set this up rapidly and our first analyses are already complete. We now urgently need funds to sustain the platform, and to accelerate and expand our work on the following three workstreams: 1) Further research to identify which patients are at higher risk of admission, ventilation and death from Covid-19, to inform management, seclusion advice, service planning, and underlying mechanisms behind these findings (e.g. is there higher risk among BAME groups simply because of previous medical conditions?); and to explore hypotheses on certain drugs helping to protect people against poor outcomes, including ibuprofen, asthma drugs, anticoagulants, and other treatments. 2) Predicting local spread and service need, and evaluating exit strategies from lockdown, by creating models of disease transmission which combine "disease-dynamics" approaches with detailed clinical data on Covid-19 prevalence and the population at risk in every local area. 3) Measuring and mitigating the indirect health impacts of Covid-19 on, for example, cancer diagnosis and referrals, cardiovascular management, vaccinations, etc, and identifying urgent necessary actions both nationally and in smaller regions.
Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.
Is something wrong? Let us know