A Model for Prioritizing Covid-19 Vaccination Distribution Using Electronic Health Record

Abstract

Covid-19 vaccination drives in most African and less developed countries have been very slow due to the limited number of vaccine doses available. In order to optimize the vaccination drive, it becomes imperative to create a priority list for vaccination based on the risk of mortality. The study applied Machine Learning (ML) techniques to develop a model for prioritizing Covid-19 vaccination. The dataset for training and validating the model was collected from the Kaggle datasets repository which consists of medical information for 4711 patients with confirmed Covid-19 infection. Four Machine Learning (ML) models were trained, tested and validated on the 4711 patients’ health records. The performances of the models were compared based on their precision, sensitivity, accuracy and area under the curve (AUC) scores. Performance of the four models on the datasets indicated that Multi-Tree XGBoost was the best performing model [precision: (Survival: 0.86, death: 0.74), accuracy: 0.83, AUC: 0.89], followed closely by Random Forest [precision: (survival: (Survival: 0.81, death: 0.83), accuracy: 081, AUC: 0.88]. The multi-Tree XGBoost model was therefore chosen for the model creation. This model can be adopted by health authorities and partners to make informed decisions in the Covid-19 vaccine administration.