Prediction of ICU admission for COVID-19 patients: a Machine Learning approach based on Complete Blood Count data

This is the dataset associated with the publication titled:

"Prediction of ICU admission for COVID-19 patients: a Machine Learning approach based on Complete Blood Count data"

and accepted for publication at Computer-Based Medical Systems (CBMS) 2021.

The dataset encompasses 4995 unique observations and 22 features (20 features from the Complete Blood Count and 2 demographics feature), along with two possible targets: ICU admission (column "Severity") and death (column "Dead). All data is de-identified, an anonymous ID field is available to associate patients with observations.

As regards the features: Sex is encoded as a binary variable where 1 represents "Male" and 0 represents "Female"; simiarly also the two target variables are binary encoded, and they both refer to a 5 day horizon (that is, the value of the target variable is equal to 1 if, within 5 days from the observation date the adverse event occurred). Full information about dataset features, processing methods, et cetera is available in the accompanying paper.

Fon any question or comment please contact: a.campagner@campus.unimib.it