There are limitations to applying artificial intelligence to medical databases, suggesting that EHR data collection will need to be improved for AI to make a difference in predicting mortality, a March 10 study published in JAMA Cardiology found.
The study concluded machine learning models did not substantively improve prediction of in-hospital mortality compared with models based on traditional logistic regression.
The same advanced methods that were successful outside of healthcare did not meaningfully predict mortality in a large national registry. Improving EHRs may be the key needed to unlock artificial intelligence in medicine, the researchers suggested.
Researchers said in an article about the study that for machine learning to be adopted, data needs to be collected and processed better, as it currently consists of a limited number of fields that may not be relevant to predicting mortality.
"We believe that the next frontier for improving clinical prediction may be the application of these methods to the high-dimensional granular data collected in the EHR," study author Rohan Khera, MD, said.