Algorithm capable of predicting a patient’s death
It sounds pretty morbid (and not a little like something out of a sci-fi movie) but Google has developed an algorithm capable of predicting a patient’s death.
An article published in Nature, a nature research journal, indicates that deep learning can be used to help hospitals predict a patient’s length of stay and time of discharge, as well as their time of death. Indeed, a trial in which electronic health record data from two American hospitals was inputted into a machine showed that the machine could accurately forecast such things.
How it works
It’s no surprise that a huge amount of data is required for the accurate prediction of a patient’s death. The researchers used a neural network for the study, a type of AI that can handle massive amounts of information. Not only were patient vitals and medical history fed in, the neural network also took into account handwritten notes and comments when making its forecast. The algorithm created a timeline from all this information to enable a deep learning model to pinpoint future outcomes.
What’s the benefit?
But what’s the point of these predictions? How can hospitals benefit from them? Well, they may enable hospitals to better prioritise patient care as well as adjust treatment plans to improve the chance of success. For example, two treatment plans may be run through the machine, which would then predict which has a higher chance of success. The algorithm may also enable healthcare workers to catch a medical emergency before it occurs and take steps to mitigate it.
Other AI within the industry is already proving itself. There are algorithms, for example, that can recognise lung cancer and heart disease more accurately than humans. These trials, however, have been carried out on a far smaller scale than Google’s project.
The reason for this is two-pronged. Not only does UK health data exist across a number of independent databases, most companies don’t have the money to carry out large scale trials. The UK government offers R&D tax credits for software development that leads to algorithms to aid healthcare, but rarely do businesses have millions at a time to spend on research.
A major concern about this kind of algorithm development is the immense amount of data that would be held by a small number of private companies, which would in turn enable them to monopolise the system.
Another, raised by healthcare workers, is around transparency. Once AI is embedded into the medical industry, measures must be taken to ensure it meets critical criteria that will prevent misuse or bias.
How R&D tax credits can help
As mentioned above, algorithm trials require extensive financial and resource investment. The government’s tax relief initiative enables software companies in the UK to claim R&D tax credits for eligible activity that aims to overcome a challenge using scientific or technological understanding. You can calculate how much you may be able to claim using our R&D tax credits calculator.