Using Internet of things technology to construct an integrated intelligent sensing environment for long-term care service

In the current long-term care environment, there is a shortage of manpower and a high turnover rate of staff. Therefore, residential institutions are eager to build an effective Internet of Things integration mechanism to assist institutions with automatic sensor detection and early warning capabilities. Although Internet of Things facilities have enabled prompt notification and warning of emergency events, the following problems exist when implementing Internet of Things in the facilities: (1) low compatibility between sensors has led to excessive installation costs; (2) warning systems that are based on fixed threshold values and lack of flexibility can cause false or omitted reports that result in the incapability of reflecting real conditions and additional labor costs would be required. This study uses a medical-grade Internet of Things module that can calculate the environmental values with edge computing to generate different levels of alarms by combining the index-weighted moving average method to dynamically calculate the optimal threshold value for the environment. It takes 2 months to collect data from care institutions. The average F1-Score obtained in different environments is between 0.46 and 0.88. The results show that compared with using a fixed threshold, this method can effectively reduce sensor error notifications and missed notifications.