The World Health Organization has developed guidelines for different age groups on how much physical activity (PA) is needed for good health [1], whereby adults should get 150 minutes of moderate PA or 75 minutes of vigorous PA per week. Yet, on a validated PA questionnaire answered by over 2000 older adults in the European Union (mean age of 74.9, SD 4.5 years), 42% were found to not meet these global PA recommendations. These findings mirror others indicating that most older adults have a physically inactive and sedentary lifestyle [2-4]. Further, older participants tend to meet the recommendations less often [5]. Notably, people’s amount of PA decreased dramatically during the COVID-19 pandemic, when they faced isolation and social distancing. Specifically, home confinement reduced people’s total weekly PA time and intensity and increased daily sitting time in a health-threatening direction [6], as these factors increase one’s risk of mortality, obesity, sarcopenia, frailty, falls, functional impairments, and neurodegenerative diseases [7,8]. By contrast, regular PA results in a lower risk of falls and mortality as well as an improved overall health status, including better Quality of Life (QoL) [7,9]. PA is known to reduce one’s risk of dementia disease (DD) [10], and in Parkinson disease (PD) it is “a vital component to maintain balance, mobility and activities of daily living” [11]. In a web-based survey from May to August 2020, 75% (64/85) of participants with these conditions reported that receiving PA recommendations was desired, and 79% (57/72) of caregivers reported the same. Interviews with patients and workshops with health care professionals support these results [12].

Strategies for changing people’s behavior to become healthier fall under the general umbrella of behavior change techniques, which might often rely on 3 activation factors: skills, knowledge, and motivation. Research shows that previous interventions that addressed these factors could positively influence individuals’ PA behavior [13,14]. Notably, these factors can be addressed by different behavior change techniques; for example, skills can be trained with instructions or demonstrations of a specific exercise [15], and knowledge can be addressed by educating individuals on the benefits of sufficient PA and the consequences of insufficient PA and by comparing their own PA behavior with guidelines [16], and motivation can be targeted by activity monitoring, either by the user, another person or a technical device [16], and through personalized feedback [17]. All of these were incorporated into PROCare4Life mobile app and iteratively tested and cocreated with older people living with PD or dementia. A recent review showed that integrating the abovementioned activation factors results in the most benefits when aiming at influencing PA behavior. Further, addressing the 3 activation factors can be accomplished by using a variety of interfaces [18] and digital interactions, that have been incorporated into PROCare4Life mobile app. Thus, it can be expected according to previous literature that digital health systems incorporating PA recommendations can potentially benefit from incorporating the above-mentioned activation factors and using various modalities to do so. These expectations were tested through the cocreation process of the PROCare4Life mobile app, which includes PA recommendations. The PROCare4Life PA recommendations approached skills and knowledge via notifications and PDFs that were displayed on the users’ smartphones, whereas the motivation was approached by personalized gamification of the smartwatch app, which could also be read through the mobile app. In addition, the PA recommendations were developed according to the Health Action Process Approach (HAPA). HAPA was used to guide the development of PROCare4Life PA training and adherence programs from a psychological theory perspective to enhance the efficiency of PA behavioral changes among older people living with PD or dementia (Alzheimer disease and others) [19].

In the European Union, the recent project “Personalized Integrated Care Promoting Quality of Life for Older People” (PROCare4Life) funded by Horizon 2020 (grant agreement number 875221, 2020‐2023) developed an interactive care system based on digital technologies for people living with neurodegenerative diseases, their caregivers (mobile health app, smartwatch for both profiles), and health care professionals (web interface) [20]. Health recommender systems [19] have become valuable digital tools [21-23], as they provide personalized suggestions and predictions to users by analyzing vast amounts of data and using advanced algorithms [24]. Health recommender systems have shown significant potential to improve patients’ and caregivers’ healthy behaviors and QoL [25-27]. The cocreation and iterative testing supported the development of the digital interactive care system through several pilot waves. PROCare4Life pilots were implemented at 6 real-life sites including daycare, home, and rehabilitation scenarios, located in 5 different EU (European Union) countries (Germany, Italy, Portugal, Romania, and Spain). The cocreation methodology included several steps:

The PA recommendations were included in the iterative testing from the start of pilot 2, when the technology was mature enough for testing. They were embedded in the mobile app to support users in creating and maintaining an active lifestyle. The focus of this article is to describe PROCare4Life cocreation and iterative testing processes from the PA recommendations [31,32].

This article focuses on the cocreation process to support the development and iterative testing of the PROCare4Life PA recommendations and its key results. Based on the specific needs of older people living with PD or dementia, and after assessing iteratively the users’ recommendations regarding the PA contents, changes were incorporated into the digital system to improve the personalization of its end users. Overall user satisfaction assessment of the PA recommendations was also included as one of the research aims. The overall aim of PROCare4Life was to increase the QoL of people living with DD or PD, their self-management, and empowerment.

The HAPA model was used to develop the PA recommendations of the PROCare4Life digital interactive system. It includes an initial motivational phase for goal setting and a subsequent volitional phase focusing on how to implement behavioral changes to pursue personalized goals. Between these 2 phases, individuals using the PROCare4Life system can create an intention to be active [31]. As older people living with PD or dementia, PROCare4Life users needed to have basic knowledge about the PA recommendations, including the adequate volume and intensity levels of activity for each person. Accordingly, the HAPA model was amended to include the construct Awareness of Standards, which was not originally present (Figure 1).

Four HAPA constructs of the motivational phase and 2 constructs of the volitional phase were aligned with 6 performance objectives resulting in the PA recommendation contents with 3 underlying activation factors, as summarized in Table 1 below. The contents were fused into 6 PA recommendation sets. The process to develop the PA recommendations contents is displayed in the table below (Textbox 1).

The goals of the PROCare4Life PA were reinforced by the technical architecture of the digital system, which is described in the following subsection.

In terms of information collection, processing, and delivery, the PROCare4Life PA technical architecture consists of 2 types of components that transfer the PA recommendations as notifications to users: the low-level subsystem and the high-level subsystem.

The low-level subsystem encompasses the set of devices, sensors, and questionnaires, both active and passive, functions to acquire data from patients. It requires a smartwatch, a smartphone with the included mobile health app, and questionnaires. Collected data are sent to the other subsystem. The high-level subsystem, includes the recommender engine that extracts the necessary information to generate recommendations according to each user’s profile. A notification bus sends and receives notifications involving the exchange of information and messages between the user-facing part and the systems logic based on the RabbitMQ [32] message broker. The recommender system operates on an automated daily routine, sending predefined notifications to patients, independent of their previous data. While this heuristic approach may diverge from traditional recommendation methods, it ensures consistent daily monitoring [33]. At the end of each cycle, a knowledge-based approach was used to deliver customized routines and exercises tailored to each patient’s primary disease group. Identity management provides a framework for creating, modifying, and deleting users’ data. In the Web interface, health care professionals can check historical notifications from patients and examine different patients’ information. The purpose of the clerical and clinical data repository is to store the different modalities of data that are related to users, including clinical and personal information (encrypted to comply with General Data Protection Regulation as well as some preprocessed sensorial summarized data that feeds the different dashboards and reports presented in the PROCare4Life user interface (Figure 2).

All pilot phases were ethically approved by the local ethical commissions of the 6 pilot sites of PROCare4Life:

As the research included human participants, the end user pilot sites prepared, within the consortium, a first draft version of the ethical documentation. All relevant information on the ethics of the study was collected. Based on the local regulations of each end user pilot site, the gathered information was integrated into the local documents, including ethical documentation for the ethics commission including the following sections: description and procedures of the study, voluntariness, and anonymity, information about no compensation type for the participation, insurance coverage, scope of data collection and processing, legal basis, option for the withdrawal, contact data of the responsible contact persons. This information was fused into the ethics application and the connected general participant information and consent. In addition, a data handling agreement was established and signed by all parties.

PROCare4Life second and third pilots were performed in accordance with the following inclusion and exclusion criteria.

The inclusion criteria included individuals who had a clinical diagnosis of DD or PD; were 65 years or older; demonstrated a willingness to participate; had the ability to provide informed consent; and were currently living at home, in a rehabilitation center, or in a daycare center.

By contrast, candidates were excluded from participating in the study in cases of inability to learn and limitations impeding the ability to use the PROCare4Life system or key elements of its technical equipment, such as speech, vision, language, hearing, or psychiatric impairments.

Participants were recruited from the social and health care centers involved in the study (Casa di Cura Policlinico, Asociación Parkinson Madrid, Campus Neurologico Senior, University of Medicine and Pharmacy, Spitalul Universitar de Urgenta Bucuresti, and Wohlfahrtswerk für Baden-Württemberg).

The study was conducted in 3 different scenarios: neurorehabilitation center, daycare center, and at participants’ homes. People living with DD or PD received periodically personalized PA recommendations (Multimedia Appendix 1) over the 40 days that their participation lasted, via their PROCare4Life mobile app. Each day a different notification was sent. All these materials were additionally stored as PDFs (Multimedia Appendix 2) in the app allowing participants to consult them later. Table 2 summarises the PA recommendation sets by the days of their appearance, gradually covering all 6 sets.

The period in which pilot 2 of PROCare4Life was carried out was challenging due to COVID-19 and analysis of its outcomes highlighted some limitations of the system. Specifically, pilot 2 experienced installation and setup issues, including smartphone and the Fitbit. Difficulties with the initial setup prevented the patients from using the system at all, which also implied that PA recommendations were not used nor evaluated. Consequently, one-quarter of the participants were represented in the pilot 2 data.

To counteract changes performed in the solution deployment to reduce the complexity of the setup and improve the user experience. The main change was the elimination of the physical mini-PC, which was the data collection hub for pilot 1 and pilot 2, and at the same time, the main source of installation and operation issues. In pilot 3, most of the patients only had the wristband and the phone. Data collection figures increased significantly in pilot 3. Additional changes were made to the PA recommendation components of the PROCare4Life solution, enabling more data collection points and potentially a more interactive experience for the patients.

This study reported on the theoretical background of behavior change techniques and on their technical development. The results are based on multinational participants, across various settings and health conditions. The multidisciplinary team, which included technical partners (for developing the backend and front end), academic partners (for creating research-related content), an educational network (for producing relevant materials), a security company (for guaranteeing data protection and privacy requirements) as well as clinical pilot sites (for testing the end product), followed an iterative design process for developing the interfaces as recommended by Zaman et al [42].

The PA recommendations were implemented in a systematic way and addressed the activation factors like skills, knowledge, and motivation to focus on relevant behavior change constructs. Based on users’ perceptions of the system, the content was rated positively, and users were satisfied with the PA recommendations. Addressing the activation factors can be recommended for researchers and technical developers of other projects. The current implementation of the mobile health app uses daily static notifications and performs limited monitoring of patients’ activities. More features need to be implemented to personalize the recommender system to each individual user.