Researchers have developed a revolutionary AI system, a novel deep-learning algorithm that predicts the death of hospital patients with an unprecedented 90% accuracy.
Although the idea of predicting the death of an individual sounds unsettling and far-fetched, researchers from Stanford University have made it possible for a noble cause – providing appropriate end-of-life care to hospital patients.
People suffering from terminal diseases are in dire need of end-of-life care, specialised treatments targeted at improving quality of life while providing him relief from the symptoms and stress of the illness.
However, in order to administer the approach correctly, the first thing doctors need to know is how long a patient may live. If this goes wrong due to any unforeseeable reason, some random individual may get the care ahead of the one who is actually in need and may die sooner.
This is why Stanford researchers brought artificial intelligence to the rescue. In a paper published on arXiv preprint server, the group has detailed the results of training a deep-learning algorithm to tell which patients are at greater risk of dying in the near future.
The system learned from electronic health records – information like diagnosis, procedures performed, scans and medicines taken – of some 160,000 adult and child patients from Stanford and Lucile Packard Children's hospital and came out with flying colours when put to test, Gizmodo reported.
When asked to predict which of the 40,000 patients would die in next three to twelve months, the system correctly predicted mortality in 90% of the cases. In fact, 95% of those who had lower chances of dying, according to the system, actually lived longer than the period in question.
"The scale of data available allowed us to build an all-cause mortality prediction model, instead of being disease or demographic specific," said Anand Avati, a member of Stanford University's AI Lab, according to IEEE Spectrum.
The system did incredibly well in the pilot run and researchers hope to bring it into use. However, before any of that could happen, it will have to be fed with more data to churn out better and reliable results.
Kenneth Jung, a research scientist at Stanford University notes, "We think that keeping a doctor in the loop and thinking of this as 'machine learning plus the doctor' is the way to go as opposed to blindly doing medical interventions based on algorithms... that puts us on firmer ground both ethically and safety-wise".