Effect of model updating on long-term physical condition change detection system in elderly care facilities

In long term care facilities for elderly people, detecting anomalies in physical conditions is important to improve the quality of care and of life for elderly regidents In our previous study, we constructed an anomaly detection system for physical conditions in such facilities. In this study, we attempted to improve its performance by updating the anomaly detection model in scenarios where a subject’s vital data have statistically changed. Therefore, we compared the anomaly detection performance of the system before and after updating the model with 20 elderly subjects. As a result of the proposed updating of the system, the detection performance for anomaly in oxygen saturation (SpO
2
) or body temperature was improved in most subjects except for one who was in end stage of care treatment. The degree of difference in the effectiveness of model updating may be a function of the subject’s condition. In a future study, we intend to investigate the relation between subject attributes and anomaly detection performance.