Google Is Training Machines To Predict When A Patient Will Die
Source：Bloomberg
From：Taiwan Trade Center, Los Angeles
Update Time：2018/06/21
Google has created a new algorithm for medical applications that can forecast patient outcomes better than present predictive models. Patient outcomes include how long people may stay in the hospital, odds of readmission, and chances that they will die. The Google algorithms for medical applications uses artificial intelligence (AI) software that is good at using data to automatically learn and improve.
Much of Google’s AI work has gone into improving existing internet service. Google has now found a new market to break into (healthcare) with its AI expertise. Entering new markets is something the founders of Google Sergey Brin and Larry Page have always been trying to do.
At present, software in healthcare is mainly coded by hand. Google with its AI system lets machines learn how to parse data on their own. Google found that existing predictive models sometimes miss obvious data such as whether a patient had prior surgery.
The Google research unit called Medical Brain led by its AI chief Jeff Dean wants to move this predictive system into medical clinics. Google’s Medical Brain unit is working on many AI tools to predict symptoms and disease with a level of accuracy that is both at the same time scary and hopeful.
Google’s Medical Brain unit stresses that medical experts are also seriously involved in counseling in the development and use of AI software algorithms to better patient healthcare. Impressive to medical experts was Google’s ability to look through data, even handwritten notes on medical charts, to make predictions faster and more accurately than present techniques.
Healthcare professionals have been trying hard for many years to better use electronic health records and patient data to reduce paper work time so it can be better spent saving lives. At present, mining health data is costly, difficult, and time consuming.
Trying to monetize AI in medicine has been found to be a large challenge. Most notably, IBM’s Watson unit has had only limited success in saving money and integrating AI technology to improve healthcare outcomes.
Google (and Microsoft as well) looks to monetize by licensing its AI systems to clinics or selling its AI services to them through the company’s cloud-computing division. However, to do this, Google needs more healthcare records and patient data. And Google buying this data may not sit well with regulators and the public due to privacy concerns. Google has not decided on a business model for its AI healthcare services because as Dr. Lily Peng, a member of the Google Medical Brain unit, says “I want to emphasize that this is really early on.”
Since Google already has vast amounts of information on the public, privacy officers such as Andrew Burt at data company Immuta as well as the public in general have concerns with Google making deals with research hospitals and the like such as UC San Francisco and the University of Chicago to obtain medical records and patient data. Burt says, “Companies like Google and other tech giants are going to have a unique, almost monopolistic, ability to capitalize on all the data we (public) generate”
Google is being more careful with patient information as public scrutiny over data collection has been rising. Last year, Google was fined by British regulators for testing an app that analyzes public medical records without informing patients. Although Google insists that its data collected is anonymous, secure, and obtained with patient permission, privacy officers believe it is too difficult to maintain data rigor especially if expanded to smaller hospitals and healthcare clinics where there is minimal oversight.
Although data privacy is an issue, officers still believe that the Google AI algorithms can save lives and money. There are few better companies than Google that can analyze the vast amount of data obtained to improve healthcare outcomes.