The story so far:
In a significant advancement for health data research, Ben Goldacre and his team have developed the OpenSAFELY platform, enabling access to the health records of approximately 58 million people in the U.K. without compromising individual privacy.
What data does the NHS hold?
The National Health Service (NHS) provides free healthcare to every person in the U.K. for every malady. The country is proud of its health system, and in fact, part of the opening ceremony of the London 2012 Olympics was a salute to the NHS. That the NHS is in a crisis at present is a separate story.
Every British citizen has a single health record from birth to death. Every bit of care provided by the NHS is entered into that record. The NHS health records go back more than 70 years. Since 1996, 96% of doctors’ offices have maintained electronic records.
The NHS has had its share of data leaks. Each such event reduces the public’s trust and a reduction in trust does not bode well for popular support for such a service.
Who is Ben Goldacre?
Ben Goldacre is a professor at the University of Oxford. In the last few years, Goldacre has focused his efforts on understanding the health data of the British people. He is the first director of the Bennett Institute for Applied Data Science at Oxford. His team created OpenSAFELY, a platform for accessing health record data without breaching the privacy of the person whose data it is. Goldacre stressed that they required top-notch IT professionals to develop the system. It required the skills of people who had created or run the backend of a large e-commerce company.
How does OpenSAFELY work?
Imagine a strongroom that holds all the NHS records. It used to be that researchers could request particular types of records, which would be sent to them after vetting. There were three problems with this system. First, given the backlog, it could take months if not years to vet a researcher, thus slowing research and innovation. Second, even if patient records had been stripped of identifying information, it was sometimes possible to reidentify the person. Third, it is in principle not ideal to have patients’ records in other private hands.
With OpenSAFELY, researchers are neither allowed to enter the strongroom, nor sent any records. If they have a query, they are provided with dummy data where they can test their software and check whether the data retrieved is correct. If it is, they can submit their queries to the strongroom using the same code and obtain the data they need without accessing a single record.
What else can OpenSAFELY do?
Another important feature is that every bit of code a researcher submits is visible to the world the minute it is shared with even one person. This is important. If researchers desire particular results, they can keep changing the code to access different pieces of data, for instance, until they get the desired results. As British economist Ronald Coase is reported to have said, “If you torture the data enough it will confess to anything.”
This practice is called p-hacking, where ‘p’ stands for p-value, a number that indicates whether some data is significant or just a chance occurrence.
Once the code is publicly visible, what the analyst did also becomes clear. This avoids the problem of different analysts performing similar but not identical analyses, thus producing different results from the same dataset.
How is OpenSAFELY being used?
Having created OpenSAFELY, Goldacre’s team now routinely publishes papers involving data of 20-30 million people. The team members were able to undertake many analyses related to COVID-19, including assessing factors associated with COVID-19-related deaths; identifying about 20,000 prostate cancer diagnoses that were missed in England during COVID-19, and using the number and nature of respiratory infection cases and laboratory tests to assess the impact of the first wave of COVID-19 on clinical activity.
Separately, they have investigated a range of other topics, including auditing the prescription of pancreatic enzymes for pancreatic cancer patients and assessing the safety of prescribing direct-acting oral anticoagulants to patients with mechanical heart valves.
Thus OpenSAFELY is a system that pleases both those who are concerned with data privacy and data transparency. It is perhaps a world first to do so at this scale. Such a system could be developed for many scales of data in a wide variety of domains. It is a template that we would do well to be aware of.
Gayatri Saberwal is professor, policy research, at the Institute of Bioinformatics and Applied Biotechnology, Bengaluru
Published – October 11, 2024 08:30 am IST
