Researchers looked at different computer models used to predict if patients with sepsis would develop acute respiratory distress syndrome (ARDS). They compared advanced machine learning models against standard logistic regression methods to see which was better at predicting lung damage and short-term death.
The results showed that these models have a moderate ability to identify risk. However, the study found that machine learning did not consistently perform better than the standard methods currently in use. Because the data came from many different sources, there was a lot of variation in how the models performed across different tests.
It is important to note that the evidence for these tools is currently weak. Many of the studies included had a high risk of bias or used less certain methods. Because of these limitations and the inconsistent results, these computer models are not yet proven to be superior for everyday clinical use.