Google Team Predicts Readmissions, Deaths Using EMR Data
Posted July 16, 2018
Researchers for the Internet search provider Google have leveraged the company’s expertise in deep-learning to accurately analyze and predict medical events using raw data from electronic health records (EHRs), according to a study published in the open-access journal npj Digital Medicine.
The technology, which Google calls an “artificial neural network,” was able to accurately predict a number of significant medical events, including in-hospital mortality, 30-day unplanned readmission, prolonged length of stay, and final discharge diagnosis (using ICD-9 codes) by examining large volumes of “messy,” unstructured data from the EHR, including text notes written by clinicians.
“These models outperformed traditional, clinically-used predictive models in all cases,” the researchers wrote. “This predictive performance was achieved without hand-selection of variables deemed important by an expert, similar to other applications of deep learning to EHR data. Instead, our model had access to tens of thousands of predictors for each patient, including free-text notes, and identified which data were important for a particular prediction.”
The system made its predictions at three points: patient admission, 24 hours into the hospitalization, and discharge using deidentified EHR data from more than 216,000 hospitalizations (114,000 unique patients) at the University of California, San Francisco (from 2012 to 2016) and the University of Chicago Medicine from (2009 to 2016).
According to the researchers, a major advantage of using deep-learning techniques is that investigators don’t need to specify the predictive variables to consider when analyzing patient data. Instead, the “artificial neural network” learns the key factors to look for in the data by itself.
“The promise of digital medicine stems in part from the hope that, by digitizing health data, we might more easily leverage computer information systems to understand and improve care,” the authors wrote. “Unfortunately, most of this information is not yet used in the sorts of predictive statistical models clinicians might use to improve care delivery. It is widely suspected that use of such efforts, if successful, could provide major benefits not only for patient safety and quality but also in reducing healthcare costs.”