I feel the title of the submission and article's title are a bit misleading. What was done here was predicting with ML if a given patient would be re-hospitaliz

I feel the title of the submission and article's title are a bit misleading. What was done here was predicting with ML if a given patient would be re-hospitalized in the next year. The ML does not predict Heart failure, only the probability of re-hospitalization. A much better submission would have been physionet 2016[0] or the several Kaggle competitions on heart diseases. Being 60 years old and having accompanied a few parents on their path toward their end of life, I know this article does not describe a great feat. Heart failure is a progressive disease, it lasts one or two decades, starting slowly, and having a dramatic end. In the last three years there are one to three hospitalizations per year. A simple Bayesian analysis gives the answer far much quicker than any ML. Disclosure: I started to develop a true heart failure early detector[1], it works by analyzing the heart sounds: [0] https://physionet.org/challenge/2016/ https://physionet.org/challenge/2016/ [1] https://github.com/Hjertesvikt/Hjerte https://github.com/Hjertesvikt/Hjerte