AI is the Best Now at Predicting Death in Hospitals

Photo credit: Mario Tama/Getty

Photo credit: Mario Tama/Getty


By Michael Spencer  |  June 18, 2018 

Michael Spencer

We know that artificial intelligence will transform healthcare as we know it. It’s science fiction come alive here, as Google AI can help doctors predict when patients might die. The AI is now better at predicting death than hospitals. 

Google’s Medical Brain team is now training its AI to predict the death risk among hospital patients, and there’s lots of buzz what this will do to our healthcare systems. In a paper published in Nature in May of 2018, from Google’s team, it says of its predictive algorithm:

These models outperformed traditional, clinically-used predictive models in all cases. We believe that this approach can be used to create accurate and scalable predictions for a variety of clinical scenarios.

Bloomberg first reported the story, and went on to explain that Google has a new algorithm that can quickly sift through thousands of digital documents in a patient’s health record to find important information and with superior number crunching the AI, once fed this data, makes predictions about the likelihood of death, discharge, and readmission.

Data is a Matter of Life and Death at Times

Google’s ‘Medical Brain’ team appears to be making significant AI advances that allows Alphabet to be more implicated in healthcare. This enables the technology in some cases to help doctor, basically augmenting them with more accurate predictive analytics, and allows them to make better predictions about the outcome of particular patients and their prognosis, for instance, how long a patient may stay in a hospital or when the likelihood that the patient may die.

Google had already released an AI tool to help make sense of our genomes. Basically AI tools could help us turn information gleaned from genetic sequencing into life-saving therapies. So AI really now is at the cusp of life and death, and perhaps ready to be set loose on our health data. 

Jeffrey Dean head of Medical Brain, Google
Image credit: Fortune
Jeffrey Dean head of Medical Brain, Google

Google AI is able to work in surprising ways, and able to interpret and add thousands of data points. For instance, according to Bloomberg, what impressed medical experts most was Google’s ability to sift through data previously out of reach: notes buried in PDFs or scribbled on old charts. The neural net gobbled up all this unruly information then spat out predictions. These may be the finest predictive algorithms in healthcare on the planet. 

Google’s system even shows which records led it to conclusions; and it’s speed and accuracy appears to be without precedent. Google is getting ready to bring the tech into clinics and use ‘a slew of AI tools to predict symptoms and disease.’

AI chief Jeff Dean health research unit — sometimes referred to as Medical Brain — is working on a slew of AI tools that can predict symptoms and disease with a level of accuracy that is being met with anything from shock, hope as well as alarm. Here is both a tool that well outperforms doctors, and yet augments how they may be able to assess patients. Trials put the accuracy of the AI’s predictions as high as 95%.

Since Alphabet Inc.’s Google declared itself an “AI-first” company in 2016, Medical Brain is showing great progress. Google’s AI uses neural networks, which has proven effective at gathering data and then using it to learn and improve analysis. Google's system is possibly, and even probably, the fastest, and more accurate technique known today at evaluating a patient’s medical history. What could this mean for the future of healthcare? Dean envisions the AI system steering doctors toward certain medications and diagnoses. How much will AI fundamentally change how doctors deal with patients? AI could help improve patient outcomes, reduce error theoretically, and also use patient data like never before.