OpenSAFELY: The New COVID-19 Analysis Technique that Could Change the Face of Medical Research
Written by Sarah Butler
A new open source analytics platform, OpenSAFELY, offers fast large scale database analysis for electronic health records without extracting sensitive patient information. Trialed in the UK through the The National Health Service (NHS), the platform played a key role in delivering prompt results during the COVID-19 emergency. Now researchers are looking at the potential of this model, beyond the pandemic, for secure, timely access to data.
On May 7th, medRXiv, an online server for non-peer-reviewed medical research, published an article about factors associated with death from COVID-19. This was the largest study on COVID-19 conducted by any country to date, and co-primary investigator, Professor Liam Smeeth was confident it would be of, “crucial importance to countries around the world”. The article, written by the UK-based OpenSAFELY Collaborative demonstrated a higher risk of death in the elderly, ethnic minorities, those with uncontrolled diabetes, and those with severe asthma, among a number of other factors. The results were widely shared in the medical community and the media.
Less spoken of however, is the story behind the article, of how a group of epidemiologists and data scientists were able to turn an idea into an audit of 17 million patients, and publish within 42 days — a seemingly impossible feat.
Traditionally, a study as large as OpenSAFELY would require a process of extracting sensitive patient information from large databases, held in individual companies for both hospitals and general practices. From there, researchers would need to link the databases together to permit the large-scale analysis required. To receive ethics approval and access to such a large source of sensitive information would be a notably onerous task. Permissions can take months, or even years, and the analysis itself can take even more time before the results are able to be published. Yet with their novel software, in conjunction with loosened restrictions in the COVID-19 era, OpenSAFELY made this all occur in just 42 days.
The OpenSAFELY team developed a software that could analyse the data in-situ, requiring no removal or copying of sensitive information whatsoever. Instead, their software could perform the specific necessary analyses from within the The Phoenix Partnership (TPP) data centre, which stores the electronic medical records of some 50 million British people. Specifically, the software was written as a number of programs, each with an interrogation aimed at the patient records through a secure connection. Only the results of such queries would be the output provided to the team. OpenSAFELY also ensured that a log of all queries was kept as part of the software, keeping themselves accountable for what data was analysed and what questions were asked.
Co-primary investigator Ben Goldacre stated in an article for The Independent, that their software permitted them to do what, “was impossible to do ... with the [readily accessible] data alone.” It was this combination of accessing the data without making copies and keeping a detailed log of all actions performed that gave OpenSAFELY the NHS's trust, and permitted such a rapid turnaround.
“Everyone wants to do things fast,” Francesca Orsini, a biostatistician at the Clinical Epidemiology and Biostatistics Unit (CEBU) at the Murdoch Children’s Research Institute told MedicGuild, in reference to how in the current climate, COVID-19 trials are fast tracked through the usual approval processes.
While Orsini’s work has also benefited from these accelerated processes— she is part of the team of researchers investigating the potential of the BCG [tuberculosis] vaccine against COVID-19 — she cautions against continuing at this speed if it potentially jeopardises the scientific and ethical integrity of research.
“It’s very dangerous for us to go at the speed we are allowing research to progress during COVID-19,” Orsini said. “We still need to ensure we are proceeding in a way that is accurate and with high level scientific evidence.”
But Francesca Cavallaro a research fellow at the Institute of Child Health, at the University College London, together with the other authors of the OpenSAFELY article, argue that it’s not a matter of making shortcuts when it comes to scientific and ethical standards, but of addressing inefficiencies in the current research infrastructure, especially in the UK.
“Researchers are required to demonstrate scientific quality and public benefit in applications to data providers and governance bodies, even when these important aspects have already been assessed by peer review and funders,” Cavallaro wrote in their article for BMJ.
“While appropriate governance is important for protecting confidentiality and preserving public trust, approval processes are not streamlined and timelines do not reflect expectations of the public.”
COVID-19 has highlighted the fundamental limitations of existing systems, and has sparked innovation for supporting data access - Francesca Cavallaro, research fellow at the Institute of Child Health, at the University College London
TELL US YOUR STORY
We want to hear from you
Do you have a story idea? Or have an experience and perspective you'd like to share?
Podcasts
Is There a Doctor in the House? What The Rise in Doctors Running for Election Tells Us About Our Health Crises
The pandemic has revealed the cracks in our public health system are gaping holes, with many...
Other Articles
The Best Jobs to Juggle with Medical School
Any university degree requires a lot of juggling, with the need to balance study alongside family...
How to Become a Surgeon: Your Questions Answered
Recently, MedicGuild wanted to hear your questions for our surgical registrar about surgical...
Cavallaro and her team of researchers argue that the OpenSAFELY Collaborative has given us an exciting glimpse into the future of medical research. “COVID-19 has highlighted the fundamental limitations of existing systems, and has sparked innovation for supporting data access,” Cavallaro wrote.
“When we reach the “new normal,” we should not return to business-as-usual, but instead take heed of lessons learned during the pandemic and rebalance the public benefits of wider data use against numerous existing barriers.”
The OpenSAFELY Collaborative hopes that others will be able to use the system they created during the pandemic as a blueprint for more efficient, secure and timely research. The team has made all of their codes and algorithms freely available as open source software to the broader research community, showing a dedication and commitment to promoting open science. With a few alterations, the code can be used to run any specific query on any database with the same "hands-free" approach.
Orsini agrees that this crisis has demonstrated the potential for historic collaboration, “something good is going to come out of this pandemic and research,” Orsini said. “In particular in the area of data sharing.”
the challenge for open safely in australia
Big datasets can provide researchers with a wealth of information, permitting greater accuracy and a wider range of approaches to analysis.
In an Australian context, however, it will be a challenge to implement something like OpenSAFELY. Dr. Karen Lamb, a biostatistician and senior research fellow who provides statistical advice on the design, analysis and reporting of medical research with the University of Melbourne sees the great potential of OpenSAFELY’s software, but also points out that for this to be of use, we need one central resource of patient data- that's not the case in Australia.
Unlike the UK system where a patient has to register with a single GP practice, Australians can visit a number of GPs and it’s up to the patient if their data is transferred across. “There is also no quality memory in our system” Lamb explained, and this is often a “huge limitation in our ability to perform such large-scale observational studies[here]” .
Lamb herself is currently experiencing this difficulty, as she examines the risk factors for cardiovascular events. Whilst she has access to Victorian primary care data, she only has access to three hospital networks to examine for cardiac events, meaning if a patient presented to any other hospital they would not know. “We are so far behind in Australia, we need the systems and networks to speak to each other,” Lamb added.
Yet if OpenSAFELY's methodology were to become adopted on a larger scale, there could be significant implications. With access to such software, electronic medical records (which are being rolled out across most Australian hospitals) could be transformed from being simply a storage of data to becoming a live portal for medical research.
In Australia, the ‘My Health Record’ system ensures every Australian citizen has an online health record that can be accessed by GPs, hospitals and health care providers. As of the 31st of January 2019, all Australians had a record set up for them unless they chose to opt-out before this date. This system can provide a vital role in emergency situations to allow healthcare professionals to find out important information about a patient such as allergies and medical conditions. However, there are still a number of challenges preventing this system from being as effective as it can be. These include a low input from providers— although 90% of Australians now have a My Health record, less than a quarter of Australia’s healthcare provider organisations are using the system — and cyber security concerns.
A 2019 report by the Australian National Audit office found the My Health Record system “largely effective,” but pointed out serious concerns about the use of third-party organisations working with the system.
"Management of shared cybersecurity risks was not appropriate and should be improved with respect to those risks that are shared with third-party software vendors and healthcare provider organisations,” the report stated.
Despite this, in an article published in the Sydney Morning Herald, a spokesperson from the Australian Digital Health Agency (ADHA) which oversees the online record system stated that the records, “cannot be used for commercial purposes...this is prohibited in legislation to protect the privacy and security of personal healthcare information.”
Currently only three apps are connected to the My Health Record, Telstra Health's HealthNow product, the doctor booking platform HealthEngine, and the health tracker app Healthi but the ADHA is looking to work with a further 20 app developers
To improve the system’s cyber security Australian National Audit office said ADHA needed to set up means to monitor compliance by third parties using My Health.
A consequence of the controversies around MyHealth has meant that it will be a significant challenge to implement similar methodologies here. “People are always hesitant when they hear they’re data is being used,” Melbourne University’s Dr. Karen Lamb said. “I don't even think it's necessarily a mistrust of researchers, but there's certainly a mistrust of the government and how they will use the data.”
Whilst the restrictions around data approval will be sure to return to their usual state in the post-COVID-19 environment, the OpenSAFELY Collaborative have shown us a new method which may make reforming data infrastructure much more justifiable and simplistic in the future.
Write for Us!
Do you have a story idea?
Or an experience and perspective you’d like to share?
All submissions are paid.
Submit your pitch here