Banks could be “good guys” for a change.

OpenSAFELY is a capability that provides public policy and academic researchers access to the health information of the whole population of England in a hyper- secure manner. Because of its federated architecture it lends itself to collaboration with banks whilst ensuring banks retain complete control of their customers’ data.  By collaborating, a bank (or banks) could contribute to a much better public understanding of the costs and income relationships with various medical conditions.  Banks could be “good guys” for a change.

The rest of this article explains how. It starts with a description of the key features of open safely. It then describes some of the public policy research questions that are currently open and concludes with what a bank or banks could do to help.

What is OpenSAFELY?

For a full description see https://www.opensafely.org/.  The following is a brief summary. OpenSAFELY provides a service to academic and public policy researchers; it is a not-for-profit organisation. It provides access to all the GP medical records of the population of England along with much of the hospital medical records.

The access is very carefully controlled to ensure personal medical information is never released. Some, but by no means all, of the key control features are;

• All research proposals are vetted by ethics committees before work can begin.
• All code and results from the project are made public to ensure they are in line with the original research proposal. All data is stored in secure locations (vaults) under the controls of the organisations running the operational systems that generate the data.
• Open safely uses a federated model, there is no concept of a central data lake so no large-scale transmissions of huge volumes of personal information on a regular basis via Extract Transform and Load (ETL) processes.
• No researcher has access to the data. Researchers create analytic software using model data and then hand the code over to open safely to run against the data in the vaults. Only summarised results leave the vaults; not even sued normalised records leave the vaults (as these can too easily be reverse engineered to reveal personal identities).
• Code has to pass acceptance criteria prior to execution in the vault in terms of personal confidentiality.

What are some of the current health research questions banks can help with?

Among a sea of concerns, there is currently a lot of interest in two facets of public health;

• inequality of access to healthcare provided by the NHS, for example see here
• effective use of public money in the form of adult social care and its spiralling costs, for example see here

These matters and others could be much better understood by a fine grained description between individuals’ medical conditions and their finances; what their incomes are and what they spend money on. Examples of enlightened understanding might include

• Which conditions require abnormally high expenditure on energy bills (e.g. for 24 x 7 heating or hoists and electric wheelchairs).
• How benefits payments actually lineup with costs such as care assistants and medications at a detailed level. How access to healthcare varies with actual income (currently postcode is often used as a proxy for income and health research).

How can banks help?

Banks have detailed information on the expenditure that their customers make using their debit and credit cards including the classifications of the types of payment such as to supermarkets or chemists or utility companies. Banks also have information on monthly income whether in the form of NHS pensions, benefits payments or salaries.  This information if linked to a individual’s medical records would provide a much more informed insight to allow policy formation.

Banks, rightly, are very careful with their customers’ data.  The federated model proposed by OpenSAFELY, with the data not leaving the organisation that owns it, would fit well with the bank’s position on privacy.  No customer data would have to leave the bank’s premises only summary results. This would mean bank staff (who know and understand their data intimately) would have to support and run the data extract code and hence there will be a level of cost involved to the banks.  The potential benefits to offset this cost would be;

• The positive PR of the bank helping wider society is important policy questions
• The possibility of some of the research questions being directly useful to the banks; for example the relationship between medical conditions and customer vulnerability and their needs from banks.

For more now open safely see how this would work; see their White Paper on a federated national data library.