The COVID-19 pandemic has claimed millions of lives around the world. Long-term hospitalization and Healthcare-associated infections (HCAI) are one of the main factors leading to death. This paper carried out a study of hematological variables associated to HCAI among COVID-19 patients hospitalized at a large health center in the metropolitan area of São Paulo, Brazil. An open database containing 499 patients diagnosed with COVID-19 and 37 attributes was assembled and given as input to different Machine Learning techniques. The objective was to predict the risk of developing HCAI based on initial status at hospital admission. Different Machine learning prediction algorithms were applied, including Support Vector Machine (SVM), Gradient Boosting, and Random Forest with 10-fold cross-validation. The SVM model performed better, with an AUC of 0.77 ± 0.06. SHapely adaptive explanations demonstrated that the indirect bilirubin and fibrinogen variables influenced the SVM model in an inversely proportional, differently way than expected for our pool of patients.