PREDICTION OF LENGTH OF HOSPITALIZATION FOR PATIENTS WITH CORONARY HEART DISEASE IN AL ISLAM BANDUNG HOSPITAL USING GRADIENT BOOSTING ALGORITHM

• Main Author: KUSUMA WIROTAMA - NIM: 14413041, FADIL
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/27065
• Institution: Institut Teknologi Bandung

<p align="justify">Coronary Heart Disease (CHD) is one of the diseases with the biggest contribution to the needs of hospitalized patients at Al Islam Bandung Hospital. With the large number of CHD patients undergoing hospitalization at Al Islam Bandung Hospital and the limited facilities to care for patients with inpatient services, hospitals need knowledge of the length of stay of patients to be able to plan service policies well. Therefore, this study tried to answer the needs of the hospital by designing a prediction model for length of stay in CHD patients at the time of patient admission and decision support systems using the model. <br /> <br /> This research uses data mining method with data used from 216 cases of CHD at Al Islam Hospital in Bandung. The data used were obtained through the literature study and validation of experts, namely cardiologists. The length of stay of patients is classified into three classes, namely 0-2 days, 3-4 days, and more than 4 days using the equal frequency method whose results have been validated by experts. The construction of the prediction model is done using a gradient boosting algorithm. The constructed model produces predictive accuracy of 77.27% with the attributes used for construction are sodium, leukocytes, potassium, history of heart disease, ST elevation, angiographic / PCI action, pulse, chest pain, hemoglobin, creatinine, blood sugar, urea, age, systolic blood pressure, diastolic blood pressure, smoking, CHF, gender, comorbidity, history of hypertension, history of diabetes, ST depression, and exercise. <br /> <br /> This study provides a solution design in the form of a length of stay prediction system for CHD patients built using the help of a shiny package in RStudio software. The system prototype can help hospitals to develop policies and manage resources more efficiently in providing good services to patients.<p align="justify"> <br />