Artificial Intelligence in the NHS

It has been reported recently that Lord Darzi, whose report into the future of the NHS is to be published later this month, is calling for a radical shakeup of the technology within the NHS. Lord Darzi is to call on the Government to embrace both artificial intelligence (AI) and what are described as “bedside robots”. It is thought that by automating as many tasks and services as possible, savings could be in the billions of pounds per year.

For example, on arrival at hospital a patient could be digitally triaged and AI systems would then be used to make diagnoses of diseases such as pneumonia, certain types of cancers, eye diseases, and heart conditions. Further automation could also assist in administrative tasks, such as making appointments, managing diaries, processing prescriptions and controlling access or provision to medical records.

However, the reliance on automated diagnoses raises clear questions about accountability. Put simply, if a misdiagnosis is made by an AI system, who is to blame? Is it the hospital, as the user of the software, or the developer of the software? If a patient is injured by something going wrong, is it a medical negligence case against the hospital, or a product liability case against the manufacturer or developer? Such issues may become more prevalent as technology advances.

Any software system capable of making life-changing recommendations will need to be developed in close collaboration with the end user, i.e. the NHS. The output of the software will be a result of calculations made using parameters the end user has stipulated, not the developer. If we assume that the hardware performs as expected, any errors in diagnosis and interpretation are based upon the software itself. It has been reported previously by researchers at John Radcliffe Hospital in Oxford that a newly developed AI system is able to search scans for clumps of cells in specific organs and then advise whether the cells are harmless or potentially cancerous. For the software to do so, it will require extensive input from appropriate medical specialities during the development process so that the artificial decision it is making is equivalent to the standard expected from a living, breathing doctor!

The NHS is currently heavily dependent upon cutting-edge technology and, as Lord Darzi points out, it will become increasingly dependent on new technologies in the future, if it is to provide a reliable and cost-effective service. For a patient who suffers from harm due to a device, the primary responsibility is on the NHS. It seems, therefore, that we should view AI as just another piece of hospital equipment, with the AI merely being a tool used by the hospital. AI may well be a major boon to the NHS, taking pressure from front-line resources and allowing the medical staff to focus their efforts in other areas. It may help avoid, or at least lessen, the yearly winter crisis non-urgent surgery being cancelled due to staffing issues and patient numbers.

In the short-term and with increasing pressures on the NHS with an ageing population, the appropriate use of technology to speed treatment times should be welcomed and not feared. However, the NHS will need to exercise caution – with an increasing reliance on AI, will we see the first fully robot doctor? If so, how will patients react to the lack of human empathy?