Referral Hospital Recommendation System for Patients Based on Diagnosis Data and Severity Level Using the Random Forest Method at Puskesmas Sigambal

Abstract

The process of determining referral hospitals for patients in primary healthcare facilities is often conducted based on subjective considerations without systematic analysis of patient data. In fact, diagnosis data and severity levels play a crucial role in determining appropriate referral decisions. This study aims to develop a referral hospital recommendation system for patients based on diagnosis data and severity level using the Random Forest method at Puskesmas Sigambal. The dataset used in this study consists of 120 patient referral records, including attributes such as age, gender, diagnosis, severity level, and referral hospital as the target variable. Diagnosis and severity level serve as the main variables in the classification process. The research stages include data preprocessing, categorical data transformation, splitting the dataset into training and testing sets with an 80:20 ratio, and building the Random Forest model. The evaluation results using a confusion matrix show an accuracy of 91.4%, precision of 89.8%, recall of 90.5%, and F1-score of 90.1%. These results indicate that the Random Forest method is capable of classifying referral hospitals with good performance based on patients’ diagnosis data and severity levels. Therefore, the developed system can assist healthcare personnel in providing referral hospital recommendations in a more objective, fast, and data-driven manner at Puskesmas Sigambal.