A New Machine-Learning Approach to Prognosticate Poisoned Patients by Combining Nature of Poison, Circumstances of Intoxication and Therapeutic Care Indices

Estimating patient prognosis has always been a challenge for physicians. One of the first questions asked by a patient entourage during his treatment is related to his vital prognosis. Thus, practitioners, even with several years of experience and the great progress of technology still have difficulties to estimate prognosis. They have tried to build prognostic allowing them to make an estimate as close as possible to the real future of each patient. However, these scores still have many limitations today. In particular, they are confronted with the individual complexity of each patient (history, physiological age, etc.). These imperfections tend to limit their use in the daily practice of medicine. In this work, we propose a machine learning based prognosis-prediction tool for poisoning cases. We use the data of poisoned patients delivered by the Poisoning Centre of Morocco. The Selected features will be fed into five machine learning techniques, including XGBoost, Support Vector Machine (SVM), Naïve Bayes, Decision Tree and Random Forest.