Multisource data for monitoring and intervention of loneliness in the elderly

Loneliness is harmful to health like smoking, obesity or environmental pollution and is considered a critical public health issue. Studies report that between 24.7% and 33% of people feel lonely. In the population over 80 years it has increased since 1970 from 11% to 50%. Loneliness increases a person’s risk of premature death from all causes, including dementia and heart disease. In the elderly, it causes difficulty in maintaining an independent life. This study aims to apply artificial intelligence to monitor data that comes from IoT systems with Amazon Web Services serverless architecture and online questionnaires with in-depth interviews to provide a remote service at home to detect and monitor loneliness. Along with psychologists and social workers, questionnaires and interviews were designed using non-directive conversational structures to collect data from people with and without detected loneliness. A website has been developed in such a way that it meets all aspects of accessibility and has different users such as workers/assistants and users. In this way, each user can look at the data generated for each one and the workers can carry out the interviews through the web to store the results in the database automatically. The developed IoT system is designed to monitor homes and is able to collect data from 5 sensors, presence, temperature, humidity, accelerometer, air quality, and 1 fitness tracker, the model Fitbit Charge 5. To collect data from sensors is has been used a Raspberry Pi 4 and the esp32 microcontroller, these controllers are communicated with AWS thanks to the IoT Core microservices. The data collected from the fitness tracker is requested to the Fitbit database through AWS Lambda. All data is stored in the DynamoDB Database table automatically after is communicated to AWS. To get preliminary results a series of laboratory tests were carried out to verify the correct connection with the devices and AWS. Once the test was complete, it was verified that the generated data was successfully saved to Amazon DynamoDB, more than 8,000 samples were collected. For the data obtained, clustering algorithms were applied, such as the one to find patterns in data from the Co2 and temperature sensors. 4 clusters were found for Co2 with temperature with the k_means algorithm and 3 clusters with the DBSCANS and k_medoids algorithms. For the interviews, a test was carried out with a user who reported unwanted loneliness and with another who said he did not feel alone. The next steps are to collect more data developing the designed system in some residences and with caregivers and psychologists detecting which users are lonely to later carry out an analysis by applying statistical analysis and data cleaning to develop artificial intelligence in AWS Sagemaker to detect possible loneliness alarms before further development. The developed monitoring system showed great potential and is an attractive approach for real-time monitoring in healthcare. It has the advantages of security and privacy communication with the user, as well as more possibilities for scaling and instant deployment of services

Intervention (counseling · Loneliness · Psychiatry · Health disparities and outcomes · Medicine · Nursing · Psychology · Gerontology