How AI can help healthcare providers with patient risk stratification
Deep learning can be used to predict the future hospitalization of pediatric patients, according to new research published in the American Journal of Managed Care. And by predicting the diagnosis of these patients, the study’s authors explained, health systems can also more accurately predict resource utilization.
“Within healthcare, deep learning is in use in a variety of health information technology contexts, including genomic analysis and biomedical image analysis,” wrote lead author En-Ju D. Lin, PhD, Research Institute at Nationwide Children’s Hospital in Columbus, Ohio, and colleagues. “The application of deep learning to patient-level risk prediction is a new area of exploration.”
The team turned to Skip-Gram, a three-layer neural network that “learns” the complex relationships between different medical codes, to analyze data from more than 112,000 pediatric patients from a single accountable care organization. The patients had a mean age of 8 years old. While 2014-2015 data was used to develop the researchers’ deep learning model, data from 2015-2016 was used to validate it.
Overall, the authors found that a model using demographic and utilization features alone resulted in an area under the ROC curve (AUC) of 73.1% for predicting next-year hospitalization. Adding deep learning to that equation, however, increased the AUC to 75.1%. “Small but meaningful improvements” were also noted for the model’s specificity and negative predictive value when adding deep learning to demographic and utilization features.
“Although deep learning analytic approaches are steadily increasing, the application of these models to risk stratification has been slow to gain traction in healthcare delivery systems,” the authors concluded. “Fortunately, some forward-looking accountable care systems continue to support and conduct research that demonstrates the ‘value-add’ of deep learning methods that may enhance predictive modeling capability and thus ensure better treatment and financial outcomes.”
Michael has more than 18 years of experience as a professional writer and editor. He has written at length about cardiology, radiology, artificial intelligence and other key healthcare topics.