Frailty is a common geriatric syndrome characterized by age-related declines in both physical and cognitive reserves, as well as physiological function, leading to increased vulnerability to adverse health outcomes (eg, [1]). According to the physical phenotype proposed by Fried et al [2], frailty refers to individuals meeting at least three of the following 5 criteria: weakness, slowness, low level of physical activity, self-reported exhaustion, and unintentional weight loss. In the community-dwelling older population, frail and prefrail individuals represent approximately 40% of people aged >65 years [3]. Frail individuals (FI) are at a higher risk of disability, hospitalization, and institutionalization [4,5] and are recognized as an optimal target population for the implementation of effective programs to prevent dependency [3] or even to reach a successful aging path [6]. As highlighted by the World Health Organization in its healthy aging concept, environments and their interactions with intrinsic capabilities play a substantial role in developing or maintaining functional abilities that enable well-being in older age [6]. In this FI population, environmental support can be a relevant approach to encourage and facilitate the instrumental activities of daily life (IADLs). In turn, this contributes to reducing or delaying functional degradation and fostering and extending aging under good conditions [7].

Assistive technologies (ATs) for activities of daily life (ADLs) refer to all technical forms of environmental support that provide an adaptation of the environment to make it more accommodating for persons with impairments [8]. Despite enthusiasm for ATs, several issues remain to be resolved. First, all the studies published so far on AT lack empirical evidence of their efficacy, mostly because of shortcomings related to study design (small sample size, irrelevant or nonstandardized measures, short follow-up duration, no control group, etc) [9,10]. Second, despite the many technological innovations available to assist older adults in their daily life [9], their silo-based nature (one technology per need) makes them challenging to integrate, as older adults require more services to assist an increasing number of ADLs because of multiple, various, and evolving task needs, particularly in the frailty context [11]. Indeed, the restricted capability to perform ADL remains extremely patient-dependent (individual and need variability) [6]. Therefore, personalized multiple intervention programs are more efficient in reducing the impact of frailty progression (cognition, autonomy, and quality of life) than the usual intervention programs [12]. A third limitation related to the silo-based approach is context awareness of assistive services that are too restricted. Hence, such services are not flexible and are delivered irrespective of the actual person’s needs in a given situation, rendering them unsuitable or even obstructive for performing ADL [13]. Finally, informal and formal caregivers are important resources for community-dwelling FI, acting as human support for ADLs [14,15]. Thus, the assessment of ambient assisted living (AAL) benefits should integrate measures related to caregiver efficiency [16].

To advance the field of ATs, we designed a clinical trial assessing AAL-based multiservice assistance, called HomeAssist (HA), dedicated to supporting FI in their 3 main needs domains: ADL, safety, and social participation. We expected that HA use would result in better everyday functioning (main efficacy criterion: IADL scale score) compared with older adults living in the community not benefiting from HA. This paper describes the clinical trial design.

The objectives of the HA trial will be to perform a 12-month field study for testing in real conditions the use the HA-AAL in a sample of older persons (frail or not) living alone at home and then, to assess the impact and efficacy of HA in terms of aging in place and the efficiency of the caregiving environment.

Consequently, the primary outcome will be aging in place-related measures through functional status (IADL scale measure). Secondary criteria of HA efficacy will also be studied, including institutionalization and hospitalization as well as measures of health scales (self-efficacy, quality of life, routinization, etc), general cognitive status, memory, and executive functioning. In addition, caregiver-related measures will be used to assess the impact of HA on the daily delivery of services (eg, feeling of burden assessment and psychological health of caregivers) and reassurance regarding the situation of the older person.

Therefore, we mainly hypothesize that (1) the benefit of an AAL with a user-centered design is effective for aging in place of frail older adults in terms of everyday functioning (IADL scale score). As previously observed in pilot studies [17-20], we anticipate that the intervention group will maintain or improve their IADL scale score, whereas the control group will decline significantly. In addition, our secondary hypotheses are that (2) such an AAL decreases or neutralizes the frailty process and reduces the rates of institutionalization and hospitalization and that (3) it improves the psychosocial health of participants and their caregivers when compared with the control condition. Given these assumptions, we also assume that a large proportion of participants will want to continue the study at 24 months, despite the HA subscription fees.

The assessment is divided into 2 parts: health (Multimedia Appendix 1 [22,38-59]) and HA device (Multimedia Appendix 2 [17,18,60-73]) parts (Figure 7).

For the HA condition, an external trial monitoring board has been specifically set up and met once a year or as needed. It included 12 members gathering multiple expertise (gerontology and geriatrics, clinical trials for behavioral interventions, home services, public services for older adults, user-centered gerontechnology, IoT and software orchestration, epidemiology and statistics, and innovation transfer). This board provided trial oversight and monitored participant safety and well-being for the entire trial duration.

A detailed manual of operations has been developed for all study protocols, and HA-related implementation and training protocols have been scripted. All study activities are discussed at weekly HA coordinating team meetings with the project coordinators, the data management team, and technical staff around issues related to data collection, transfer, or HA technology.

Descriptive analyses were conducted to characterize the recruited HA sample compared with the others (excluded and refusals) on the data available at the eligibility visit and to describe the characteristics of the recruited sample at baseline as well as those of the dropouts. We also performed group comparisons to obtain a baseline description of the HA and control groups for the primary and secondary outcomes.

To test our main and secondary hypotheses, all measures of the study will be tested based on an intention-to-treat approach using a 2-tailed level of significance set at α=.05. The quasi-experimental study design will allow the comparison of primary and secondary efficacy outcomes in HA and control conditions. The 12-month follow-up data being not available for the control sample, imputations will be performed according to a linear hypothesis of the evolution of the scores between T0 and T24. For each outcome, we will compare the mean score at T12 between the HA and control groups. In addition, to control for confounders, three successive linear regression models will be used: (1) Model 1: Y_{(at T12)} = DomAssist (vs control) + age + Sex, (2) Model 2: Y_{(at T12)} = Model 1 + dependent variable at T0 (Y_{(à T0)}), and (3) Model 3: Y_{(at T12)} = Model 2 + MMSE T0.

Finally, we will also describe the evolution of the primary and secondary efficacy outcomes over 12 months of HA use and conduct comparisons by age, gender, initial level of technology acceptance, and user satisfaction with HA technology (collected after 6 months of use).

For the HA condition, the French Southwest and Overseas Protection of Persons Committee deemed the study to be outside the scope of the provisions governing biomedical research and routine care (DC 2015/2066). The study protocol was also approved by the National Commission of Informatics and Liberty (Comission Nationale Informatique & Libertés-French name) and the Ethics COERLE (Comité Operationel d’Evaluation des Risques Légaux et Ethiques) committee of the French National Institute of Informatics and Mathematics (Inria), as protecting participants and data accordingly. For the control group from the 2 cohorts, an ethics committee approved the research according to the principles embodied in the Declaration of Helsinki: for 3C, the Ethical Committee of the University Hospital of Kremlin-Bicêtre (Paris, France) and Sud-Méditerranée 3 (Nîmes, France) and for AMI, the committee of the University Hospital of Bordeaux (France). All the participants provided written informed consent.

As indicated in Figure 5, a total of 490 individuals received a prescreening visit for the HA condition. Of these 490 individuals, 260 (53%) were excluded because of ineligibility (n=18) or a lack of interest in participating (n=242). A total of 230 older adults underwent the baseline assessment, of which 99 (43%) were excluded (16 ineligible; 83 refusals). Finally, 56.9% (131/230) of individuals were enrolled: more than half of the participants completed the 12-month follow-up (73/131, 55.7%), from whom one-third of participants (24/73, 33%) were requested to continue the experiment for 6 additional months (18 months), and 18 participants were enrolled for 12 additional months (24 months). Figure 8 illustrates the recruitment flowchart for the 24-month duration of the HA condition.

The HA sample (n=131) at T0 is primarily female (114/131, 87%) and ranges in age from 70 to 93 (mean 81.9, SD 6.0) years. The participants enrolled in the HA condition have varied living spaces: 35.1% (46/131) individuals in rural areas, 27.4% (36/131) in semirural areas, and 37.4% (49/131) in urban areas. Of the 131 participants, 63 (48.1%) do not have a senior high school degree, and 57 (43.5%) do not have a junior high school degree. The MMSE scores range from 23 to 30 (mean 25.9, SD 2.4).

Of the 131 participants, 58 (44.3%) dropped out before T12. Compared with T12 completers (73/131, 55.7%), these individuals were slightly older (26/58, 47% were over 85 years vs 16/58, 22% of T12 completers; P=.004) and less educated (mean 8.2, SD 3.2 vs mean 9.8, SD 2.3; P=.005), while no significant difference was observed in gender (P=.80), living environment (P=.26), frailty-Short Emergency Geriatric Assessment score (P=.10), and MMSE score (P=.18).

From the 3C and AMI cohorts, 474 control participants were selected according to the eligibility criteria (Figure 9).

When comparing baseline characteristics between the HA sample and the control group, we observed no differences in terms of age, education, living condition (all living alone), IADL disability, depressive symptoms, and cognition. For the MMSE test, the control group tended to have a higher mean score, but this difference was not significant (P=.08). However, as indicated in Table 2, participants in the HA condition were more often women (63/75, 88% vs 372/474, 78.4%; P=.054).

Regarding our 2 primary outcomes, the sample size provides a reliable statistical power of 0.80 with a significance level set at .05, attesting to a significant difference in means of the IADL scale score between the 2 groups for a difference >0.87 units.

The HA project proposes an ecosystemic human-centered approach for introducing a personalized AAL platform dedicated to FI and their caregivers. Hence, our main outcome is to observe the benefit of our HA approach on this capability (improvement or nondeterioration) due to IADL follow-ups performed at 0 and 12 months. Our secondary expectation is to observe a positive effect of HA intervention for prolonging aging in place of FI living alone through improved cognitive and mental health, leading to reduced institutionalization and hospitalization. We also anticipate a positive effect on caregivers’ health. In addition, owing to the personalization of HA services and their user-centered design, we also expect that equipped persons will have a positive user experience on pragmatic criteria (perceived usability, usefulness, ease of use, learnability, etc) and hedonic criteria (pleasure, satisfaction, perceived attractiveness, etc). These expectations are directly based on our previous pilot study involving 17 HA-equipped FI compared with 17 nonequipped FI to assess the feasibility of the HA concept and collect preliminary results regarding its acceptance and elicited user experience. The results showed that HA was well adopted (highly accepted and usable) by FI and their families or caregivers [18]. Moreover, after 6 and 9 months of follow-up, the benefits of measures of self-determination behaviors and IADL scale scores [9,19] were higher in equipped FI [18-20]. For instance, with normalized scores, at 9 months, the equipped group did not change its IADL scale score, whereas the control group lost 1 SD. In addition, the objective burden of caregivers to support care receivers’ ADL increased in the control group after the follow-up period, while it did not increase in the equipped group, even though the subjective burden remained unchanged [19,20].

A total of 131 users and their families or formal caregivers participated in our proposed field study, and 73 accepted to continue the experience until 12 months. Such a clinical validation study is rare in the field of AAL-based interventions in terms of experimental design quality. Indeed, most studies in the AAL field remain at the prototype level (because of technological and usability or acceptability challenges) or at best pilot studies (because of the challenges of a field study) with small samples that do not provide the onset of the ground truth of AAL-based interventions [75].

Technological barriers include issues related to the deployment of the technologies, such as inaccurate sensors; battery power or bandwidth reliability issues, restricting users within the monitoring area; and lack of interoperability [76]. The lack of user-centered design approaches in AAL development also contributes to the low usability observed in some instances because of very limited experience for older adults in using advanced technologies [77,78], and even more for FI with physical or cognitive exhaustion leading to a lack of motivation to accept actively interacting with such technologies [79]. Taken together, these barriers have led field studies with large sample sizes to adopt a silo approach offering specialized services in one domain of need for older adults. For instance, the Personal Reminder Information and Social Management system only provides services for communication and leisure [80], and the platforms designed by Rantz et al [81] and Tomita et al [17] focus on activity monitoring. In addition, the efficacy of such specialized AAL is always estimated by a focus on older adults, without involving an impact evaluation of caregivers’ well-being. However, as highlighted by Fadrique et al [76], AAL benefits must be measured for the social care environment while protecting (near and far stakeholders) against risks in privacy and security related to data sharing in the AAL scope.

Therefore, multidomain AAL is still only in the design phase, as recently proposed with the next system study protocol [82].

The HA platform has already completed all the design, user testing, and pilot study phases to reach the empirical evidence step based on this study. Furthermore, our field study includes evaluations that cover other critical indicators for home care more broadly, such as the rate of institutionalization in nursing homes, the impact on mental and physical health, and data reported by HA (actimetrics, use, and use of services). For the equipped group, we gathered data on the factors that influence usability, technology acceptance, and use. These multiple secondary indicators aim to respond to the needs expressed by all the project’s stakeholders (aging policy makers, home service organizations, caregivers, older adults, and researchers from different domains) for successful participatory achievements while also moving on to more daring, innovative plans; for example, new research areas such as the study of relationships between AAL acceptance and AAL effectiveness for aging in place. In addition, the field and participatory nature of this study imply combining a traditional top-down approach (ie, expectation relative to specific primary criteria for a prescriptive purpose) with a bottom-up approach definitively driven by the field expertise of the main stakeholders of aging in place. The growth of citizen science supports that this combined approach (albeit risky) is a successful way to bridge science and practices with the result of rapid societal impact for targeted people and for emerging new research issues [83].

From our combined approaches, the study consortium encountered a few issues that reinforced study-related challenges, especially with older adults and with technology-based intervention. First, major challenges are related to the quasi-experimental study design, where the HA condition is compared with the control condition of a subsample of an existing population-based cohort on aging. We paid great attention to matching as well as possible the 2 groups with respect of the factors known to influence the IADL scale score (main criteria outcome). These factors include demographic factors, frailty scores, and cognitive status. Nonetheless, we are aware that such a matching method does not neutralize possible biases related to differences in recruitment methods, the number of home visits on 1-year deployment, or the minimal level of technology acceptance that is probably shared by the participants in the HA condition. It should be remembered that out of almost 500 older adults contacted, only 46.9% (230/490) met our selection criteria, and only 26.7% (131/490) agreed to participate in the study. Half of the participants dropped out after 12 months of experimentation, notably regarding technology-related challenges.

A second challenge is a great variability in participants’ technological skills in the HA condition. As recommended [84,85], we set up a standardized training phase for the use of HA, and each participant was guaranteed to keep the 2 tablets provided for the study after the experiment to boost their interest in such technology training. In addition, the interoperability of HA with a variety of technologies (Bluetooth, UPnP, ZWave, Web services, etc) and its unique point of user interaction are major assets for ensuring the acceptance of HA at the organizational level (bypassing the technical disparities across territories) and the individual level (reducing the demand on cognitive or learning resources for using assistive services).

Another study challenge is to provide 24/7 technology interventions requiring reliable internet access and good-quality bandwidth in areas offering disparate digital infrastructure. Technologically, any elaborate assistance service is based on the internet, and it is impossible to guarantee its continuous operation. To minimize service interruptions, we ensured the use of a single internet operator to avoid multiple service interruptions owing to operator-dependent updates. Covering a wide spectrum of needs and their evolution relies critically on the ability to populate the HA service catalog with a range of applications that match these needs. To do so, the HA leverages a dedicated integrated development environment while relying on needs analysis and human-centered design. This strategy is of paramount importance to ensure that the proposed assistive support is personalized to fulfill the needs of the participants and caregivers.

Despite these challenges, the outcomes of the HA study will yield significant insights into the benefits of AAL in frail older adults. It will also yield some insights into the factors influencing technology acceptance and use in FI. Indeed, it will be possible to explore how they vary according to participant characteristics, such as gender, age, cognitive abilities, and frailty level. The study will also inform the feasibility of technology-based personalization of multidomain assistive services and will probably teach some useful lessons for future technology-based field studies.

The results of this study will be disseminated through scientific publications and conferences. In addition, as we previously included in the HA design phases, the decision makers of the French public policy of assistance to the autonomy of older adults ensured that the solution is financially sustainable (<€ or US $1200 per home) by the public services so that a gain of only 1 month of aging in place produced by HA is more profitable than the monthly cost of a nursing home (>€ or US $1800 in France). This will provide a solid foundation for a business model when there is a start-up to market the technology.

Although further research is needed, the findings of the HA project will provide the very first evidence-based study in IoT-based AAL in-home environments involving user-centered design for frail and nonfrail older adults.

To move forward with evidence robustness, it will be necessary to go further and consider conducting a trial using strict randomized controlled trial methods. Similarly, there is the question of scaling up for large-scale intervention. As also mentioned, our ecosystemic and participatory approach has led us to work directly with the key stakeholders in frailty and home care for older adults in each territory, to both design an upstream HA and evaluate it downstream. Consequently, scaling up other territories for the evaluation of HA will be required thereafter for better proof of the interest of an AAL system such as HA.