Google has been training its AI model to predict the death of patients, and the results have been really good as initial tests have shown that Google’s model is performing better than the traditional clinically used predictive models.
This all was detailed in a paper published in May in the Journal Nature, in the recent results Google’s model was 95% accurate in one hospital and 93% in another hospital in predicting the death of patients. As eerie as it sounds it is becoming reality.
Google used its system to evaluate a patient with breast cancer, in 24 after her admittance Google’s AI gave her a 19.9 percent chance of dying in the hospital while the hospital’s Early Warning score gave her 9.3 percent chance of dying in the hospital. Unfortunately, the patient passed away within two weeks.
This result was the combination of 175,639 data points from the patient’s electronic medical records, including handwritten notes.
This is not the first time Google has been working on healthcare systems. DeepMind worked with the Department of Veterans Affairs, examining its 700,000 medical records and predicting potentially fatal changes in patients’ conditions.
Google would also like to work more in the field of predictive healthcare, by predicting not only deaths but also symptom and diseases.