ABSTRACT
As the population ages, it is necessary to investigate how to prevent key risks. One such risk is fall risk. This is unpleasant to the patient and can lead to a rapid and serious decline in the patient's health as well as have far-reaching cost implications. Given advances in technology and IoT, it behoves to examine how one might adopt successful technology initiatives from other sectors to address this issue. Implementing digital twin technology for real-time evaluation of trends and outcomes has dominated the manufacturing sector, but this technology solution could have benefits to healthcare contexts. We develop a digital twin architecture using a Bayesian network approach for establishing the predisposition of patients to fall risk on admission. This framework provides decision support for the seamless management of inpatient fall risk acuity by interfacing patients' clinical, demographic, and psychosocial attributes with the Bayesian network for pattern classification to predict fall risk levels. By using the Synthetic Minority Oversampling Technique (SMOTE) to upsize the minority fall risk class, it was possible to enhance the prediction accuracy of fall risk vulnerability after a case study with real hospital data. Our results show that the chances of optimizing patients' experience with a digital twin framework are very high because of the benefits it can provide to the management of fall risk on admissions.
