In line with codevelopment and health technology assessment recommendations [28-30], we designed a two-phase pilot study [31] involving community-dwelling prefrail and frail older participants remotely followed at home. It consisted of, first, a noncomparative trial in a living lab and a 12-week prospective, parallel, randomized controlled clinical trial (field trial). The first trial lasted from September 2015 to January 2016. The second lasted from October 2016 to January 2019.

The living lab experiment was set up in a living lab at the University Institute of Technology in Blagnac, France (Maison Intelligente de Blagnac) located on the university campus [26]. This flat of 70 m² is equipped with a networking infrastructure accessible to valid, frail, or disabled persons. It enables testing of technological devices in an environment similar to one’s home setting but with controlled technical conditions. The living lab phase was carried out in two subphases: (1) 9 older adults were evaluated in the Maison Intelligente de Blagnac living lab and (2) 3 older adults among the 9 participants of the living lab were followed up with at home for a month.

During the first phase, participants were invited to complete a single session standard scenario in the Maison Intelligente de Blagnac living lab (45-minute sessions consisting of a walking tour in the living lab and outside, as per usual) after calibrating the insole to their own step length and after a 10-minute presentation of the product’s objectives. During the second phase, volunteers were asked to use the device over a 1-month time period in their own homes (Figure 1). Participants were invited to wear the insole and use a touchpad without any additional instructions or training.

For the field trial, 35 participants were randomized following a 2:1 allocation into two distinct arms. Randomization was conducted independent of recruitment by the hospital’s epidemiology department. The aim of randomization was to describe the secondary objectives for future effect size calculation. Participants were randomized in either the intervention (smart insole follow-up) or control group (standard follow-up) and enrolled for 12 weeks.

Participants in the living lab phase and those randomized in the intervention group in the field trial phase were equipped with the instrumented shoe insole and were given a touchpad feedback app. The technological devices tested were a pair of insoles (only one insole is instrumented), a touch pad to collect data from the insole and inform the users about their activity (Bluetooth communication), and an induction charger to charge the shoe every night (Figure 2). The insole thickness is less than 2.5 millimeters at its thickest point (arch). The insole measures according to time (day, week, and month) the number of steps and walking distance, the average walking speed during walking periods, and active walking duration (as opposed to shuffling). Active walking was defined as continuous walking for at least 5 minutes with a tolerance of 1 minute (see Figure 3) considering that health benefits of aerobic exercise begin with any increase above the lowest levels of activity [32]. Moreover, the insole is calibrated on each participant’s step length for an accurate measurement of walking speed, unlike consumer-grade devices. The walking speed measure algorithms and the lab test results of the insole have been previously described [26].

Both groups benefitted from a frailty assessment at a geriatric day hospital for frailty (GHF) as per usual follow-up. Indeed, GHFs are developed nationwide in France following the French national authority recommendations and may accommodate up to five patients per day. Patients benefit from a follow-up phone call at 3 months and at 1 year by a nurse.

Study participants were recruited through the Toulouse University Hospital GHF (France). Ethical approval for the study was obtained from the regional independent ethics committee in September 2014 (ID-RCB: 2014-A00523-440). The trial is registered on ClinicalTrials.gov (NCT02316600). All participants provided written consent. Baseline assessments were conducted in-person at the GHF. The inclusion criteria were (both phases):

The sole exclusion criteria was life expectancy of less than 12 months. We did not request any specific level of computer literacy, and none of the participants were familiar with digital tools. There was no financial compensation for participation.

At inclusion, sociodemographic data (gender, age, marital status, living place), frailty status [1], functional abilities with the ADL [33], and cognition with MMSE [34] were collected. For the field trial, other functional and physical scores were also assessed (Instrumental Activities of Daily Living [IADL] [35] and Short Physical Performance Battery [SPPB] scores [36], respectively).

In the living lab phase, semistructured interviews were conducted, focusing on technical feasibility and acceptability. We also scored the 3 home participants’ satisfaction with the device at the end of the 1-month follow-up using the main themes identified during the interviews.

For the field trial, the primary outcome measured the acceptability of the device as follows:

The secondary outcome of the field trial measured:

The semistructured interview was proposed to each participant at the end of the follow-up date (45 minutes for the living lab test, 1 month for the 3 living lab home testers, 3 months for the field test). The interview was face-to-face for the living lab group and by phone call for both home evaluations. Two researchers of the Age-Imaging-Modelization Laboratory (LC and VR; Joseph Fourier University Sociology Laboratory, Grenoble, France) conducted them. Each interview lasted 2 hours and was designed to explore key questions relating to acceptability. For the first phase of the study (living lab), nondirective exchanges were conducted to identify recurring themes. These themes guided the interviews of the field trial participants (n=10). All interviews were transcribed. The transcripts were analyzed using a conventional content analysis along with a summative qualitative content analysis [40].

For the field phase, a clinical research assistant also contacted participants at 1, 2, and 3 months to evaluate secondary outcomes. For acceptability, the following question was systematically asked: “how well did you tolerate wearing the insole during the past month?” (5 possible answers were proposed: from totally tolerable to totally intolerable).

For the field trial, we planned to include 35 participants complying with pilot study recommendations [41]. Qualitative variables were presented by effectives and percentages, quantitative variables by means and SDs. Likert-type items were handled as continuous variables.

Comparison tests were performed: quantitative variables with Wilcoxon-Mann-Whitney test and chi-square or Fisher tests for qualitative variables. Missing data was replaced with the mean values of the groups, allowing complete case analysis. A drawback of this approach is reduced variability and weakening of covariance and correlation estimates in the data.

The Department of Epidemiology of the Toulouse University Hospital in Toulouse conducted statistical analyses using Stata version 14.2 (Stata Corp).

Our 12-week field trial is the first to assess participant’s acceptability of an instrumented insole over 12 weeks in frail older adults in a real-life setting. Difficulty of physical activity follow-up in older individuals and insufficient in-person measures and self-reported data limit clinical research in this domain [9,22]. Currently, there is limited data on the use of instrumented insoles beyond a few hours of laboratory testing.

This study confirmed the importance of a multiphase design for health technologies. The living lab participants valued the study’s participative aspect and proposed modifications concerning both the software and the hardware, including the manufacturing method of the insole. Our living lab tests introduced an optimized reality between technical lab tests [26] and an ongoing real-life field trial. User feedback provided technical and acceptability issues, which were fixed before the field trial (low reliability scores, mainly associated with synchronization problems between the insole and the touchpad). However, despite these precautions, certain technical problems affected the field trial.

Participants validated the device’s design: induction charger, charging routine, battery autonomy, and data transfer automation. The insole calibration phase raised no user concern. These are important results as the device was specifically designed to ensure its unobtrusiveness. There were several reasons to choose an insole: it is unobtrusive and can be worn without disturbing or stigmatizing the person, several studies showed that inertial sensors worn on the feet allow accurate walking speed measurement [27,43,44], and the user does not need to remember wearing it.

Semistructured interviews and questionnaires at the end of follow-up reported that the insole was actually worn, unobtrusive, and well tolerated. Moreover, participants were compliant during the 3-month follow-up; this was confirmed by objective data measured by the insole. Most participants claimed that the insole was comfortable and did not cause any discomfort. This result was innovative, as it is one of the first to describe insole long-term wearing acceptability in older adults who are frail. Thus, a thin instrumentation including a flexible battery was designed. This technical challenge remains and must be considered when instrumenting an insole because it is one of the main participant complaints. Acceptability of wearable devices is the cornerstone of large implementation in real-life settings. Finally, those who did not walk regularly also expressed the fact that wearing a smart insole could encourage walking, which remains to be proven.

We found no significant differences between groups in terms of physical activity or frailty evolution. This could be explained by the low study power and by the fact that our touchpad did not offer any incentive or educational content. Information and communication technology–supported lifestyle programs and motion sensing–based monitoring can influence daily physical activity thanks to feedback [45]. However, adding educational content is more efficient [46,47].

This study included a small sample of volunteer older participants. Even if a few exclusion criteria were applied, there was probably a recruitment bias due to highly motivated participants. Moreover, a large proportion of participants in the intervention group discontinued the intervention because of technical issues.

The results obtained by this semiqualitative approach are not free of the usual biases, in particular external validity. There are other limitations to the generalizability of our results. Certain participants brought up the difficulty of integrating an insole in their shoes (eg, sandals), which could be even more problematic in countries where people walk barefoot indoor or wear outdoor shoes.

Moreover, the use of the touchpad interface was not satisfying because of lack of technical support for using it and participant’s computer literacy, web interface ergonomic issues, and the absence of educational content (eg, video tutorial). These shortcomings led most of the participants not to use the touchpad. The motivational aspect of our device was poor despite the potential interest this could have [48]. In a second development phase, an educational and motivational network should be developed to increase user’s adherence. It seems important to also include participant’s computer literacy when developing health devices in this population to ensure better acceptability.

Previous work has highlighted the difficulty of implementing such technical devices in health care practice due to the importance of material and human investment [49]. The results of this study give a glimpse into health device development and improvement, but further studies are required in a larger population to improve large-scale implementation.

Previous studies have shown the possibility to monitor mobility-related activities based on motion sensors, but few explicitly mentioned acceptability issues or used experimental research designs to evaluate clinical applications in older people [16,22,50,51]. Most systems consisting of multiple sensors or devices are difficult when applied in long-term monitoring in real-life (eg, the DynaPort MoveMonitor weighs 44.5 g and is fixed with an elastic belt [15]). A few studies evaluated accelerometer-based devices on short treadmill walks [52] on a short time period at home, ranging from a few minutes to days [15,53-56] with up to 20 older participants with unknown frailty status.

Most of these studies explore various aspects of walking such as depth posture and activity detection. Nevertheless, none of these studies evaluated long-term acceptability of such wearable devices in monitoring walking activity in an older participant sample.

Kaye and colleagues [57] conducted the most notable study. They evaluated in-home walking monitoring using embedded sensors in autonomous living older individuals during a 4-week period. The study did not specifically target frailty status monitoring and was limited to indoor activity. They were able to accurately monitor mean walking speed and even predict falls [9]. Embedded and wearable tools to monitor walking activity provide complementary information. Outdoor walking speed analysis has advantages on indoor measurements because walking distance may be longer and informs global aerobic physical activity [22,58].

Consumer-grade activity trackers (connected devices or smartphone apps) are increasingly popular. However, there is a lack of real-life research on the device’s performance in older adults. Given the inaccuracy of these applications, caution is required in promoting self-monitoring physical activity and their use for health prediction [19,23,24]. Indeed, the absence of step length calibration does not allow accurate measures of walking speed and walking characterization. Another major concern is their lack of acceptability in frail older adults. Most of them require minimal computer literacy, and users have to remember to wear it every day. Moreover, there is many constraints related to obtrusiveness (eg, device charging and data transfer). Lastly, the algorithms used to measure steps and other metrics are typically proprietary and may not be available to investigators [59].

Wearable connected sensors are promising for real-life monitoring and appear to be a solution in improving physical activity promotion in frail older adults. However, optimal deployment of wearable health devices will require further research conducted in real-life conditions to test acceptability, effectiveness, and costs. This study reports real-life acceptability data on an instrumented insole in frail older participants over a 12-week period. These results are informative in terms of technical choices for those who wish to instrument a shoe insole.

This field of research is essential and offers interesting perspectives. Along physical activity promotion, these tools would improve detection of early preclinical health transitions implicated in decreased physical performance [9,57,60,61] (eg, gait speed variability over time). Thus, continuous measurements would also enable identification of innovative “digital biomarkers” as a complementary solution to “traditional” biomarkers, leading to more personalized interventions.