In their article on predicting mortality in pulmonary embolism (PE) patients, El-Bouri and colleagues developed a prediction model to improve upon two existing pulmonary embolism scores, sPESI and PESI [1]. The authors conclude that their new machine learning (ML)-based prediction model (AUC 0.71, 95 % CI 0.63–0.78) outperformed the two existing scores (AUC 0.65, 95 % CI 0.57–0.73 and 0.64, 95 % CI 0.56–0.72) and are optimistic about its potential for use in clinic. While we fully support the authors’ important attempt to develop an improved prediction model for mortality in PE patients, we will also point out some methodological concerns that make us less optimistic about the potential of the model for future use.
