Regular physical activity is a key factor to a successful aging, contributing to positive outcomes in health and well-being in later life [1-4]. It can improve physical function [4], slow the progression of degenerative diseases [3], reduce risk of falls [1], and also improve cognitive performance, mood, and quality of life (QoL) of older adults [2,4]. A physically inactive lifestyle, on the contrary, can increase the risk of developing chronic diseases, one of the leading causes of death and disability in older adults [5,6].

Engaging in regular physical activity can be challenging. Older adults might suffer from reduced mobility, low self-efficacy, lack the proper infrastructures in their communities, or simply find it difficult to leave home and participate in physical activities on a regular basis [7,8]. For these and many other reasons, physical inactivity is still prevalent in older adults [9], leading to the undesired effects on health and well-being.

Intervention programs to promote physical activity have shown to be effective in increasing and maintaining physical activity [10]. In particular, group-based interventions have shown promising results in long-term settings with higher adherence compared with individual home-based interventions. Studies have also reported a preference by older adults for group exercising [11] and discussed the potential of the social context to stimulate social interactions and increase social well-being [12].

However, despite the body of literature on the topic, little attention has been paid on populations living under difficult environmental conditions and undergoing complex social changes, such as the Siberian community. Seasonal fluctuation has been found to determine the level of physical and social activities of older adults [13] leading to less opportunities to go out and interact, especially in high latitudes where winter can result in a decline of physical functions of older adults, such as ankle strength [14]. Recent history has also shaped the lives of older adults in Russia. The breakup of Soviet Union in the early 90s, and the difficult years that followed, negatively affected the social and economic well-being of the Russian population: the life expectancy of men is 14 years lower than in the European Union [15], and loneliness levels are among the highest in Europe [16]. The social, political, and economic uncertainty also deeply affected QoL, with a decrease in life satisfaction and happiness [17].

The above observations point to the need for solutions that can help older adults living under the above conditions to keep physically and socially active. Technology-supported interventions have been shown in the past to be successful in this goal [18].

Recent research has demonstrated an effectiveness of technology-supported exercise interventions for older adults in terms of physical fitness [18]. However, although there is an ongoing discussion on whether group exercising or home-based individual exercising is more effective in increasing adherence of individuals to training programs (eg, [19,20]) and despite calls for analysis focusing on understanding group-based exercising in terms of cohesiveness (frequency of contact and group dynamics) [21], no intervention has compared the effectiveness of individual and (different types of) group settings in a technology-supported intervention.

Research has also shown a preference by older adults in group training [11,12]. However, implementing group exercising can be challenging, especially in a heterogeneous elderly population, with individual differences leading to motivational issues and problems in tailoring the training [11].

Fitness apps for home-based training have been widely explored in technology-supported interventions (see [22] for a review); however, we are not aware of interventions supporting online group exercising for individuals of different levels of fitness. Consequently, there is very limited research on the effects of level of fitness, social support, and subjective well-being in online group settings. The exception comes from a recent study on an Internet-based group training intervention [23] relying on a general-purpose teleconference software to deliver real-time exercises to older adults in rural areas. Although targeting homogeneous groups, focused on physical fitness outcomes, and limited to a small sample of 10 older adults, the study highlights some interesting challenges in deploying this type of technology.

In our previous study [12,24], we made some steps to test the feasibility of a tool for online group exercising, namely Gymcentral, that allows individual of different levels of fitness to follow exercises with the remote company of others. We conducted an 8-week pilot study exploring the effects of online group exercise training in Trento, Italy, with 37 adults, 65 years and above, who followed the Otago exercise program [25] aiming at strength and balance improvement in older age. The specific focus of the study was on technology acceptance, attitude, and preference toward group training and its effects on physical and social well-being; in comparison with a traditional tablet-based individual training program implementing no persuasion strategies.

Still, despite the prior work and the extensive existing literature, open questions remain:

This paper reports on two pilot studies of an online exercise intervention with older adults living in Tomsk, Siberian Federal District (Russia). The aim of the intervention was to enable older adults of different levels of fitness to follow a personalized exercise program from home, with the (virtual) company of training companions and under the supervision of a remote coach. This was done with the support of a tablet app offering group exercising in a virtual gym while leveraging on the social context of the group exercising to enable social interactions and feedback.

The main objective of the pilot was to study the feasibility of online group exercising under different cohesion settings among Siberian older adults. We focused on the technology acceptance, on the adherence to the training (especially in light of pre-measures of social support, as well as on the enjoyment of physical activity and leg muscle strength). As a secondary objective, we also explored the effects of the technology-supported intervention on subjective well-being and loneliness.

The technology support was provided by Gymcentral, a tablet and Web app that allows trainees of different functional abilities to follow online group exercises from home, under the supervision of a remote coach [26]. Gymcentral serves the needs of trainees and coach via the trainee and coach apps (see Figure 1).

The design of the trainee app is based on a virtual gym environment that provides the following main features:

The monitoring and feedback is supported by the coach app, a companion Web app for the training expert.

Details about the features of the Gymcentral app are discussed in detail in the study by Báez et al [26] and the underlying conceptual model in the study by Far et al [27].

In this work, we studied the feasibility and effectiveness of the online group exercise intervention and its effects on the well-being of Siberian older adults by addressing the following specific research questions (RQ):

RQ1. Is the online group exercising technology usable and accepted by older adults? We aimed at exploring the perception of older adults toward the technology by measuring the usability and acceptance. More importantly, we also explored how the app was used in practice and how the usage relates to the observed effects of the.

RQ2. How do online group exercising and baseline measures influence the adherence of older adults to a training program? Previous research suggests that exercising in a group results in higher adherence and preference by older adults [7,11]. However, research also points to major obstacles when delivering group exercises to heterogeneous populations, which can make training in this setting difficult and less motivating [11]. In this study, we explored how a virtual group environment influences the adherence of older adults under different measures of known determinants of physical activity.

RQ3. Does online group exercising affect the well-being of older adults? We explored the effects of physical training via a virtual social environment on the subjective well-being and social well-being of older adults. By addressing this question, we aimed at contributing to the existing research on the association between physical training and well-being [1-4].

We explored the above questions in two pilot studies in Tomsk, Siberian Federal District (Russian Federation) that adhered to the same protocol and conditions, except for the group cohesion setting:

As seen above, we explored two group cohesion settings: participants with strong group cohesion and participants with low group cohesion. Thus, for the reasons explained above, candidate participants from Tomsk 1 had a stronger cohesion than Tomsk 2 at recruitment time, regardless of the treatment they ended up receiving. We did so to understand the effect of the prior connectedness among participants on the observed outcomes.

Both pilot studies were follow-ups to a previous pilot performed in Trento, Italy, and so they follow the same study design [12]. An overview of the study flow in consolidated standards of reporting trials–compliant format is shown in Figures 2 and 3.

In both studies described here, participants were assigned to an interaction group (online group exercise condition) or to an individual group (individual exercise condition) using a random assignment procedure, with age and participants' frailty level as random assignment variables. In Tomsk 1, the process was slightly different as to ensure a high level of cohesion after randomization: pairs of friends, identified during the informative meeting, were treated as single elements during randomization. In this modified process, we firstly followed the randomization procedure for participants without friends, assigning participants to interaction and individual treatments, and then repeating the process for the friend pair units. Thus, friends were assigned to the same treatments, contributing to the overall group cohesion in Tomsk 1.

The two studies and the two treatment conditions defined four effective groups (see Table 1). Participants in the interaction groups have access to online group exercising with social interaction and persuasion features, whereas in the individual groups, participants have access to individual training with persuasion features but with social interactions limited to contacts with the coach. Details about the group cohesion and features available to each group can be seen in Table 1.

Both versions of the app implemented the same training program, developed on the basis of the Otago exercise program [25], which includes a set of muscle strengthening and balance-retraining exercises. The training program was designed with a standard set of exercises to be performed in each training session, varying in intensity each week according to the performance of the trainees. In the app, each exercise had 10 levels of intensity based on the duration and the number of repetitions. At the beginning of the study, a personal trainer (who was also the coach in the virtual gym) performed a physical assessment, which was used to set the starting intensity level of the program.

Participants received an iPad Air tablet (9.7-inch) preinstalled with the assigned version of the app and Internet access, a case to support the vertical positioning of the tablet, an activity monitoring sensor (Misfit Shine), one pair of ankle weights (0.5 Kg each), and the telephone number of the support team.

Before the start of the training program, participants joined pretest and technology training meetings: (1) an initial meeting where they signed the informed consent and filled out enrollment questionnaires, (2) a session with a medical doctor to evaluate eligibility, (3) a technology training session in the use of tablets and the assigned version of the app, and (4) a session for the physical assessment with the coach and pretest measures. The technology training followed a workshop format and was done in small groups of 10 participants each. Participants assigned to individual and interaction conditions attended workshops separately as they were provided with different versions of the app.

In the 8 weeks of the training, participants performed the home-based training activity with the monitoring of the coach and of the support staff. The training schedule offered three exercise sessions per week, and participants were required to perform at least two exercise sessions every week. The duration of the training session ranged from 30 to 40 min depending on the intensity level. Participants were free to join the training sessions at any time. Posttest measures took place on the week after the training.

The coach guiding the participants during the training was a practicing doctor with a primary care doctor degree and had over 10 years of experience in gymnastics, rehabilitation exercises, and yoga for older adults. Before the beginning of experiment, the coach was acquainted with the Otago training program and Gymcentral app settings.

During the training period, the coach had the task of progressing the intensity of the exercise program and providing feedback. At the end of every week, the coach could maintain or increase the intensity level of each trainee according to the attendance and completeness of the training sessions in the week. The coach was also instructed to contact trainees at least once a week to provide feedback and to respond to any question from the trainees. The coach was not aware of the difference between the interaction and individual groups, and both received the same amount of technical support.

The pretest measures included the Groningen Frailty Indicator [28], the Rapid Assessment of Physical Activity Questionnaire [29], demographic information, and questionnaires concerning psychological and social well-being. The posttest measures included the System Usability Scale (SUS) [30], a set of questions on the acceptance of the app, the Satisfaction with Life Scale (SWLS) [31,32], the medical outcomes survey (MOS) Social Support Scale [33,34], and the 3-item revised University of California, Los Angeles (UCLA) Loneliness Scale (R-UCLA Loneliness Scale) [35,36]. The participants filled in all the questionnaires by themselves in pencil-and-paper format.

The study protocol received ethical approval from the CREATE-NET Ethics Committee on ICT Research Involving Human Beings (Application N. 2014-001) in Trento, Italy. The studies reported in this paper—as follow-ups to our previous study—comply with this protocol, with the informed consent and informational materials translated into the Russian language.

We considered eligible for the study: participants aged 59 years or older, independent living, self-sufficient, and with a nonfrail, transitionally frail or a mild frailty level. These criteria were measured by self-reports. All participants had to pass a doctor assessment to ascertain the absence of conditions that would prevent them from performing light physical exercises. Participants wearing pacemakers were considered not eligible as the study required the use of an activity sensor (Misfit shine monitor). The specifics of baseline measures for each study site are described in Table 2.

Participants in both studies were contacted through retirement organizations in Tomsk, Russia. In the first study, Tomsk1, participants were mainly invited through organization offering computer-learning classes and hobbies activities for seniors. In the Tomsk2 study, the recruitment was carried out through three organizations organizing social activities and events. We conducted presentations explaining the project and their expected involvement and handed out printed bulletins. Older adults interested in participating provided their phone numbers and were later on contacted by the project coordinator. Details about the retirement organizations and the number of candidates reached can be seen in Table 3.

In the Tomsk1 study, 20 participants were found eligible for the study (mean age individual group=65, SD 6.1; interaction group: mean 68.2, SD 7.8; 19 females and 1 male). In the Tomsk2 study, 40 participants were accepted according to the inclusion criteria (mean age individual group=68.9, SD 7.2; interaction group: mean 67.6, SD 6.2; all 40 female). The difference in the number of male and female participants is because of the demographics of the study location and the availability of male candidates at the retirement organizations. In Siberia, lifespan gap between males and females is one of the biggest in the world: life expectancy at birth for men is 64.7 years, whereas for women it is 76.3 years [37]. These demographics posed difficulties in recruiting male participants from the retirement organizations. The study flow for Tomsk1 and Tomsk2 is depicted in Figures 2 and 3. After the recruitment, participants in both studies signed the informed consent before participating in the experiment.

In the Tomsk 1 study, out of 20 participants, 5 withdrew before the start of the study for health problems or personal reasons; therefore, data of 15 participants was included in the analysis. In the Tomsk2 study, out of 40 participants, 2 withdrew before the beginning of the training because of travel plans. During the training program, 4 participants in the individual group and 5 participants in the interaction group dropped out because of health issues, travels, or reported lack of time for participation. Thus, in the Tomsk2 study, a total of 29 participants were included in the analysis (individual: 16, interaction: 13).

There were no statistical differences between individual and interaction groups in term of initial measures (Table 2). These baseline comparisons have been performed on participants that finished the training program.

We analyzed the difference between the interaction and the individual groups in terms of the SUS score with two Mann Whitney tests, whereas for the difference in the percentage of copresence, we use t tests.

We analyzed adherence (measured as rate of persistence) to the training program with an analysis of covariance (ANCOVA) with group (interaction vs individual) and study (Tomsk1 vs Tomsk2) as between-subject factors and leg muscle strength, social support (MOS score), and enjoyment of physical activity (PACES score) as covariates.

For well-being measures, we selected the SWLS score and R-UCLA Loneliness Scale score as dependent variables to be used in two separate repeated-measures analysis of variance (ANOVA). We used the same independent variables in both ANOVAs: time (pretest vs posttest) as within-subject factor and group (interaction vs individual) and study (Tomsk1 vs Tomsk2) as between-subject factors.

The statistical analyses were performed using the open source statistical software R (R Studio Team) [42], using the ggplot2 package to create plots [43].

A starting point to understand the feasibility of the technology for our target population was to address (RQ1) and investigate the perceived usability, acceptance, and usage of the online group exercising technology.

The overall persistence rate was of 74% (SD 27%) when considering the number of sessions available in the 8 weeks of training. Breaking down this number by group treatment, we observe a persistence rate of 75% (SD 28%) for the individual groups and 74% (SD 26%) for the interaction groups, whereas the result by study shows a persistence rate of 82% (SD 24%) for Tomsk1 and 70% (SD 28%) for Tomsk2. In the study Tomsk1, the persistence rate was 77% (SD 25%) for the individual group and 87% (SD 23%) for the interaction group; in Tomsk2, it was 74% (SD 30%) for the individual group and 65% (SD 25%) for the interaction group.

An ANCOVA was performed to compare the persistence of participants of individual and interaction groups in the two studies while controlling for the initial baseline measures of leg muscle strength, social support, and PACES. The results show neither a significant main effect for group (F_1,18<1, P=.74) or for study (F_1,18=1.46, P=.24), nor interaction between study and group (F_1,18=1.15, P=.30).

Considering the baseline measures, the results show a significant interaction between study and the initial social support score (F_1,18=5.23, P=.03). As observed in Figure 6, part A, in Tomsk2, participants with higher social support level showed higher adherence to the training, whereas in Tomsk1, the adherence is not significantly associated with by the initial social support score.

No significant effects were found for the initial scores of leg muscle strength.

The interaction between PACES score and group was also significant (F_1,18=6.001, P=.03). As shown in Figure 6, in Tomsk2, participants with higher enjoyment of physical activity had a higher adherence level (Figure 6, part B), whereas in Tomsk1, enjoyment of physical activity had a negative effect on the interaction group (Figure 6, part C).

Eight participants did not answer to one or more questions of the SWLS and thus, have been excluded by this analysis. On the subset of participants without missing answers, SWLS score was analyzed with a repeated measure ANOVA with time (pretest vs posttest) as within-subject factor, and group (individual vs interaction) and study (Tomsk1 vs Tomsk2) as between-subject factors. Only the main effect of time was significant (F_1,31=5.85, P=.02). Participants reported high satisfaction in the posttest questionnaire (mean 23.8 [SD 6.2]) compared with the pretest measures (mean 21.34 [SD 5.8]).

The same analysis was performed on R-UCLA Loneliness Scale. Eight participants were excluded from the analysis because of missing values in the pretest or posttest questionnaires. Only the main effect of study showed a tendency toward significance (F_1,31=3.55, P=.07). Participants reported a lower level of loneliness in the Tomsk1 study (mean 4.77 [SD 1.7]) compared with the Tomsk2 study (mean 4 [SD 1]).

The results point to the feasibility and effectiveness of technology-supported physical interventions, and in particular, of online group exercising among Siberian older adults. High cohesion groups are preferable for group exercising, especially to mitigate effects of low social support on adherence. Cultural differences might explain the preference of private messages over public ones. Results in terms of subjective well-being are promising, but enabling interaction has proved not to be enough to observe a decrease in loneliness. Thus, further research is needed to understand how to better enable community-building interactions.