Presenting a Two-Stage Hybrid Model for Allocating Advanced Ventilators Using Machine Learning Methods: A Case Study

Abstract

In the health systems field, a significant amount of data is generated and collected, which can be used to extract valuable information using machine learning techniques. The Intensive Care Unit (ICU) is one of the most critical parts of hospitals. This study investigates the ICU of cardiac surgery, where one of the critical topics is the ventilation of patients. In other words, patients’ breathing using ventilator is a crucial need in the ICU, which has become a significant issue due to the limited number of ventilators in the hospital. One of the concerns of treatment staff is determining which patients require ventilators more urgently than others. Therefore, we predicted the patients’ need for ventilators in the ICU of cardiac surgery. Among the innovative aspects of this study, we can mention the use of a two-stage hybrid model. In the first stage of modeling, we used machine learning and deep learning models to implement 15 classification models and make predictions. In the second stage, patients were prioritized using a combination of clustering and medical expert opinion. The results showed us that the Artificial Neural Network (ANN) performed better than other models in the first stage with remarkable accuracy (80%) and specificity (95%). So, the ANN predicted that out of 328 incoming patients, 69 of them needed ventilators. Then, in the second stage, these 69 patients were divided into 6 groups by clustering and prioritized based on the expert’s opinion that determined the acuteness status of the patients to assign limited ventilators to critically patients first.

Keywords

• Machine learning
• prediction
• intensive care unit (ICU)
• ventilator
• cardiac surgery
• hybrid model