Design of intelligent monitoring of loneliness in the elderly using a serverless architecture with real-time communication API

1.Introduction

Loneliness can be defined as negative subjective personal feelings due to low-quality or non-social connections [1]. Older people are at greater risk of loneliness and social isolation as they are more likely to deal with living alone, loss of family or friends, chronic illness, and hearing loss [2]. The Centers for Disease Control and Prevention recognizes loneliness and social isolation in the elderly as critical public health issues [3]. Prevalence studies report that there are countries in Europe where between 18.7% and 24.2% of middle-aged people feel lonely [1]. In the case of the United States, 33% of middle-aged people manifest it and 25% of those over 65 are socially isolated [3]. Loneliness increases a person’s risk of premature death from all causes, including dementia, heart disease, and stroke [4].

It is mostly believed that loneliness only affects older people as they age but is experienced across all age groups as O’Sullivan et al. conclude by reviewing the systematic review and meta-analysis of 113 countries during the COVID-19 pandemic [5]. Variations in loneliness by age and region require further exploration and data collection because not everyone is at the same risk of feeling lonely. The state of loneliness is affected by situations of poverty, poor physical or mental health, few community connections, and living alone. These increase the risk of loneliness. A better understanding of the situation and impact of the loneliness experience is required. Also in different cultures and geographical variations.

It is important to see the increase in loneliness in the general and older population since 1970. As indicated by Crowe et al. [6]. The incidence has increased from 11% to 50% in people over 80 years of age. They suggest a multisystem approach strategy that can be developed from the health and care systems at a global level.

The use of innovative technological tools in homes and nursing homes is giving positive results. As Waycott et al. [7] analyzed with 11 interviews, immersive virtual reality (VR) is being used as an enriching experience for people living in care homes, with healthcare staff playing a key role in helping clients use virtual reality. It is an opportunity where users have fun and allows them to travel virtually where it generates opportunities to make memories while exercising among all residents. Although they conclude that the possibility of customizing the technology to be used for each user is important.

Passive sensing and Wearables-Based Solutions have been used to monitor and manage loneliness. For instance, Site et al. [8] studied a combination of the domains of gerontology, social psychology, architecture, and portable wireless technology to monitor and attempt to predict loneliness. Elmer et al. [9] use passive sensing collecting data from smartphone data to analyze time-stamped sensor data of social interactions. With that, they generate multistate survival models.

The main contribution of this study is the design and development of a portable smart real-time loneliness monitor. Therefore, it is important to mention the role of using innovative tool technologies to implement these systems in homes and residences. Cloud computing is growing as serverless architectures are used. The design of architectures through cloud services is considered to function as a service (FaaS) that allows to develop and run applications or backend as a service (BaaS) as storage [10].

Serverless architectures have been used for a blockchain hyper ledger fabric-enabled consortium architecture providing security, integrity, transparency, and provenance to health-related transactions and exchanges of sensitive clinical information [11]. Also for creating an AI-powered smart healthcare framework to reduce heart disease-related deaths and prevent financial losses by reducing misdiagnoses [12]. Moreover, serverless architectures have great advantages such as including deploying applications simply without the necessity of using a dedicated server, the combination of security with performance, reducing operating costs and development times, having no scaling restrictions, streamlining internal processes and facilitating continuous improvement [13].

Finally, the main objective is the design of a customizable loneliness monitoring system capable of storing using Amazon Web Services (AWS) and the commercial activity bracelet, specifically the Fitbit Charge 5 model.

2.Materials and methods

In this section, the methods and materials used for the design of the proposed monitoring architecture are shown. In the case of hardware, the commercial Fitbit Charge 5 bracelet has been chosen and in the case of software development with the database, it has been selected to develop it through cloud computing thanks to AWS. Figure 1 shows the design of the serverless architecture implemented using AWS services.

Figure 1.

AWS serverless architecture.

3.Results

3.1Serverless architecture description

3.1.1Fitbit connection for request

Once the access tokens and renewal of access to the data have been obtained through the Fitbit Dev application registration, two string-type parameters have been created in the systems Manager service parameter store. This way the tokens can be modified when they are updated and are protected with additional security.

To make requests to the Fitbit database, the AWS lambda function called “fitbit_connect” is called through an event, as shown in Fig. 2. The function obtains the tokens stored in the store and calls the internal function “get_activity” by passing it the desired tokens, date, and activity. This function makes the requests following the examples that Fitbit proposes. If it detects an error in the request because the access token is expired, call the “get_tokens” function and then redo the request changing the old security access with the new tokens in the correspondence parameter store. Once the request has been made successfully, the data is filtered to obtain a JSON that allows it to be correctly saved in the database. To this JSON the data of the company, professional and user are added. Finally, the internal function “storing_data” is activated, which sends the JSON to the Dynamo DB database table.

3.1.2Continuous data collection

To request the data from the Fitbit server, we have created a rule within the default event bus of the Amazon EventBridge for the Lambda function activation from time to time to check the evolution during the day. The selected time interval is 15 minutes.

3.1.3Database table model

Figure 2.

“Fitbit_connect” lambda function functionality and connection between other services.

We designed a DynamoDB database table with a combination of partition key and sort key as seen in Table 1. The parameters from left to right have been ordered depending on the hierarchy of the data to increase the number of queries from lambda. There are five types of possible data for the table taking into account that in the project there will be different companies that manage the professionals who will monitor the users of the devices.

Table 1

Partition and sort keys for each type of data for the DynamoDB table

Type	Partition key	Sort key
Company	COMP	COMP#<company_name>
Professional	PROF#<professional_name>	COMP#<company_name>
User	USER#<user_name>	PROF#<professional_name>
Device	SENSOR	USER#<user_name>
Measurement	USER#<user_name>	Sensor_type(Fitbit)#<timestamp>

3.1.4Database request

We implemented the read, upload, and delete operations in lambda to make requests to the database to be able to analyze the data collected and visualize them on the web. This function receives information on the action and variables to execute the demanded request. Below are summarized implemented actions of the lambda function:

• • User actions: Create and delete individual users for a specific professional, also read all the existing users or users for a professional.

• • Company actions: Create and delete individual companies, and also read all the existing companies.

• • Professional actions: Create and delete individual professionals for a specifically created company, also read all the existing professionals or professionals for a company.

• • Device actions: Create and delete individual devices for a specific user, also read all the existing devices or devices for a user.

• • Measurement actions: Save, and delete a measurement requested to Fitbit and also read all the existing measurements for a user.

Figure 3.

Example with part of the data collected and saved to the DynamoDB table following the partition key and sort key combination.

4.Discussion

The developed monitoring system showed great potential and is an attractive approach for real-time monitoring in healthcare. Wearable consumer activity trackers have the advantage of security and privacy communication with the user using mobile phone authentication sessions in the communication app. As Perumal et al. [22] conclude, the communication with wearable devices with the encryption of mobile phones provide secure and private communications. It also has applications in detecting more possible medical episodes than conventional monitoring [23].

As for the serverless architecture proposed, it has the advantages of scaling and instant deployment of services which allows the increase of users using the monitor which could help increase the use of possible monitors in the health system [12]. Also, the communication system using API complies with the requirements of cross-border transactions involving massive IoT application data and considers the privacy of users [12, 24]. Storing the data in the DynamoDB database allows cost-efficient and secure IoT systems supported by AWS security systems for health data taking into account that third-party auditors regularly test and verify the effectiveness of AWS security [25].

This design also has limitations, this are the data collection capacity and the precision of the same of the selected commercial bracelet. It must also take into account the rigorous control that should be in the residences. A bracelet change could occur and not be notified, which would result in the data being misclassified.

This research showed the potential benefits that a serverless solution coupled with the use of sensors could provide in health care and third age.

5.Conclusions

In this paper, a new system has been proposed by combining wearable devices and Serverless Computing technology architecture to monitor loneliness. The developed architecture shows great potential for easy data collection, analysis, security, personalization, real-time inference, and scalability of sensors and actuators in the future. It has powerful benefits to apply in the health sector and reduces cases of depression and loneliness.

The limitations of this study are the difficulty of carrying out an extensive test with real users in residences and homes for the elderly and that with this method of extracting data from Fitbit some of the data obtained through the calls is accumulated for the day and not very detailed throughout the day.

In future work, we want collect more data developing the designed system in some residences and with caregivers and psychologists detecting which users are lonely to later carry out a analysis by applying statistical analysis and data cleaning to develop artificial intelligence in AWS Sagemaker.

Later, the artificial intelligence models will be programmed in AWS to detect possible loneliness alarms before further development. In addition, using AI to make predictions and generate warnings to professionals of possible risks of loneliness in users.