<![CDATA[Inpatient care is the largest component of healthcare service expenditure. Healthcare management plays a role in reducing expenditure costs and improving healthcare services. Identification of factors related to patient length of stay and accurate prediction of how long patients will be hospitalized becomes important to support stakeholder decision making. In this study, the length of stay for patients using BPJS insurance services was predicted using the random forest method. An experiment has been conducted to compare different numbers of trees and candidate split attributes in a prediction model. The experimental results showed that increasing the number of trees and candidate split attributes can improve prediction performance and reduce the resulting error rate. The optimal value was found when the number of trees was 100 with the MSE/Variance value of 0.3805. The main determinant variables for predicting patient length of stay were found to be the patient's disease diagnosis, participant segment, and healthcare facility type.]]>
Copyrights © 2023
