Patient care classification using machine learning techniques

Abstract

Doctors and Specialists use the lab test results of patients to classify their medical needed care into inpatient care or outpatient care, which is a time-consuming process and needs a lot of efforts from doctors to decide whether the patient needs to be in the hospital and monitored or not. In addition, the likelihood of making the wrong decision is high, thus it may endanger the patient’s life. the purpose of this study is to utilize machine learning to classify patient care into inpatient or outpatient, in order to reduce the efforts and time expanded by the doctors which reflect on the type of services provided to the patient, also this kind of studies can help in reducing the human errors that result in risks to the patient’s life and may increase the total bill of patients which led to pay significant amounts. machine-learning was utilized to build four models: Support Vector Machine, Decision Tree, Random Forest, and K-Nearest Neighbors (KNN), that could predict whether the patient should be classified as inpatient or outpatient-based on their conditions and lab test results. The best model has been chosen based on the highest accuracy, sensitivity, specificity, and precision score, and the lowest false-negative rate, and false-positive rate. (EHR) the dataset has been used which consists of patients’ laboratory test results from a private hospital in Indonesia to build these models and test them. The results show that Random Forest achieved the highest accuracy (77%), Sensitivity (65%), and Precision (72%), respectively, the model also has the lowest false-negative rate (35%), and almost the lowest false positive rate (16%).