An invitation to appear on BBC’s Panorama does not usually bode well. On this occasion, their questions on whether Google was correct to use health data to create algorithms to read mammograms were entirely reasonable. Explaining that it is simply not possible to create these tools without the images, I tried to provide justification and waited anxiously to see if my comments would be included, but alas the programme never aired.
The pandemic struck and suddenly something became very clear: we need data to make decisions. Specifically, we needed healthcare data to feed models and inform good policy decisions.
Data provided the cornerstone in analysing how many Covid cases were tolerable without overwhelming the NHS. This national data integration, together with an amazing collaborative effort, became imperative to our response.
Data provided the cornerstone in analysing how many Covid cases were tolerable without overwhelming the NHS
Over the past 16 months, my department at Cambridge University has seen first-hand the power data yields. Using pseudonymised data, we worked to develop and test an AI tool aimed at predicting oxygen requirements needed in medical settings 24 and 72 hours in advance. The implications of such a tool are significant, as pre-empting oxygen needs decrease the likelihood of being overwhelmed by demand. Akin to the algorithmic pace, the speed this research gained approval and established international links – with minimal governance or contracts – helped the innovation enormously.
Other AI tools like those to detect fractures, lung cancer, brain haemorrhage and strokes are in development. This ability to have AI robustly and reliably diagnosing means NHS capacity could be increased, manpower released to undertake clinical tasks and potentially save more lives.
Opportunities for novel developments like this are recognised by both the UK government and NHS. By facilitating the integration of patient data from various medical sources, algorithms can provide insight to inform better policy. In future, this type of data-led approach could help predict medicine shortages or those who will develop cancer.
Perhaps…it was too soon to allow health data to be utilised by commercial companies, but academics and the NHS cannot develop these tools alone
Over the summer, the UK government consulted on their ‘Data saves lives: Reshaping Health and Social Care with Data’ strategy. Highlighting the benefits data delivered by informing our Covid response on safeguarding and treatment trials, the strategy signals a new direction of travel, one where streamlining our approach to data will allow us to better harness the power it holds.
Perhaps the producers of Panorama were right, and it was too soon to allow health data to be utilised by commercial companies, but academics and the NHS cannot develop these tools alone – we need commercialisation to fast track these innovations. The underlying issue is a societal one – ensuring the safety and security of identifiable patient data. We need the public to permit us to use their data for the public good.
Creating a machine to match the performance of our best radiologists is complex, but a truly transformative aspiration. The more we are allowed to integrate data from different sources, collaborate with private sector partnerships, and given the space needed to innovate, the faster academics can get on with making those transformative goals a reality.
This article originally appeared in The Scotsman’s Inside Science column on Monday 30 August, 2021.
Fiona Gilbert FRSE is a Professor of Radiology and Head of the Department of Radiology at University of Cambridge, an Honorary Consultant Radiologist at Addenbrooke’s Hospital, Cambridge.
The RSE’s Fellows’ Blog series offers personal views from our Fellows on a variety of issues. These views are not those of the RSE and are intended to offer different perspectives on a range of current issues.