Predictive analytics can slash unnecessary GI lab tests by half
Predictive modeling likely could reduce unnecessary lab tests for intensive-care patients with gastrointestinal bleeding, according to a new study published in the International Journal of Medical Informatics.
Working from a database of 746 patients, researchers found that predictive modeling based on 11 measurements could accurately classify more than 80 percent of both necessary and unnecessary lab tests. They achieved an average reduction of 50 percent of eight common gastrointestinal lab tests, better than the 37 percent reduction reached in similar studies.
The researchers, from the Massachusetts Institute of Technology, said they planned to expand their work to include a number of underlying medical conditions and additional laboratory tests.
The article, "Reducing unnecessary lab testing in the ICU using artificial intelligence," was published online Jan. 2.
In other recent research, scientists reported using predictive modeling to project negative side effects from prescription drugs. Reporting in the Journal of Chemical and Information Modeling, researchers said computer modeling was a viable, less-costly way to test for drug side effects than current methods.
Researchers also report success predicting when and where flu outbreaks will peak using data from Google Flu Trends and techniques for weather forecasting. In a report published in the Proceedings of the National Academy of Sciences, researchers said their predictive analyses made predictions based on past flu trends.
To learn more:
- read the study abstract
How one hospital uses EHR data to track core measures
Tulsa supercomputer holds potential for health improvements
Computational models increasingly improving research, healthcare quality
CHIME, eHealth Initiative: Analytics still stuck in the past