An artificial intelligence (AI) system has been developed by the tech giant to predict whether hospital patients will survive more than 24 hours after admission, and trials proved to be 95 per cent accurate.
The system was developed by a team of researchers from Stanford, the University of Chicago and UC San Francisco. It uses patient data such as age, ethnicity and gender, combined with hospital information such as medical history, current vital signs and test results.
Google took the AI system and used machine learning to 'teach' it to become more accurate, by using the anonymous data of 216,221 adults from two US medical centres.
Accuracy was increased by the system's ability to read data usually unavailable to machines, such as doctors notes on charts or in PDFs.
After taking in 46 billion data points from the information provided the AI learned to associate words such as "life" or "death" with a data outcome.
This means it can understand just how likely someone was to die.
Stanford professor Nigam Shah told Bloomberg that around 80 per cent of time developing predictive models is spent making the data readable for the AI. But Google's system is able to use any form of date it is fed, due to its advanced machine learning capabilities.
And the system can also predict the length of a patient's stay in the hospital and their chances of being readmitted.
Google's system was accurate at predicting patient mortality 95 per cent of the time, compared with 86 per cent accuracy from traditional methods.
It also beat clinical predictions for patient's length of stay in hospital, and likelihood off re-admittance.
Google's Alvin Rajkomar said: "These models outperformed traditional, clinically used predictive models in all cases."