Claims that machines have reached parity with healthcare professionals in their ability to make clinical decisions have been treated with scepticism by the medical establishment. Given the inherent risks around patient safety, the doctors are quite right. But machine learning does have a place in healthcare: algorithms can help providers to plan capacity in pressurised services.
Our NHS partners are able to automate many of the processes and patient-professional interactions that would typically be paper-based. In doing this, vast quantities of data are generated. We use machine learning to analyse this data to yield useful insights. These can help to predict where bottlenecks might occur in waiting lists and identify better ways of delivering care.
Inhealthcare developed a machine learning prototype in partnership with an NHS trust in the North of England. We have designed the digital health service to determine which patient referrals are most likely to breach 62-day targets for starting cancer treatments. If deployed, this would benefit both the patient and the system.
We have utilised the machine learning capabilities of cloud-based computing. Not so long ago, these technologies were prohibitively expensive and needed highly skilled specialist staff to operate them. Cloud providers like AWS have democratised access to these services, bringing powerful new technologies into the mainstream. If NHS trusts aren’t using them, they will get left behind.
Of course, machine learning will develop. But big questions remain surrounding its use as a clinical tool. In a system that is non-deterministic – where the same inputs do not always generate the same outputs – who accepts the risks and responsibilities? Who is to blame when things go wrong? There are no easy answers to these questions. That is why we are focused on using machine learning to support non-clinical decision making, where we believe valuable benefits can be found.