This study is embedded as a nested cohort within the 3-armed randomized controlled trial SMART-AGE (P2019-01-003) conducted at Heidelberg University, Germany [19]. It is registered with the German Clinical Trials Register (DRKS00034316).

The process evaluation and its findings were reported following the CONPHES (Consensus-Based Process Evaluation Reporting Guideline for Public Health Intervention Studies) guideline, developed for public health interventions conducted alongside effectiveness trials [20].

Inclusion criteria for participation in the SMART-AGE study were age 67 years or older, living in the community, having at least basic knowledge of using PCs or tablets, and being German-speaking. Exclusion criteria were severe cognitive impairment, severe medical conditions (eg, Parkinson disease with use of a walker or wheelchair), and severe visual or hearing impairment.

The SMART-AGE study was approved by Heidelberg University Medical Faculty’s Ethical Committee (S-672/2022). All participants gave written, informed consent prior to participation. All data were collected and stored in pseudonymized form. Participants were allowed to keep the tablet used in the study as a form of compensation for their time and effort. All KOKU users were covered by a trial insurance during the study period.

The tablet-based digital program “KOKU Health–Keep On Keep Up” is a strength and balance training program designed for older adults to prevent physical decline, frailty, and reduce the risk of falls [14]. After an initial introduction by a physiotherapist or qualified exercise specialist, users are able to undertake a structured exercise program independently at home without supervision. For this study, the KOKU program was translated into German by 2 native speakers, ensuring that its design, structure, and functionality were fully preserved. A professional voice-over was created for the explanatory videos to maintain the quality and accessibility of the content.

Before commencing the training program, participants completed an initial assessment within KOKU: the Short Falls Efficacy Scale-International (FES-I) [21], EQ-5D [22], and questions about their history of falls and general well-being. The FES-I appears automatically again after having reached the end of level 3 and level 6 to evaluate changes in perceived concern about falling. Additionally, participants are asked weekly to report any slips, trips, or falls, including their causes and consequences. The KOKU training program comprises 7 levels, each lasting a minimum of 2 weeks. Each week includes 3 training days, with 3 exercises per day. Participants were advised to perform 12 to 15 repetitions of each exercise, or as often as possible. Every exercise is accompanied by written instructions and video demonstrations with voice-overs, providing step-by-step guidance to ensure proper and safe execution. The initial level introduces simple exercises, such as heel raises while seated and sit-to-stand movements. After each exercise, users input the number of repetitions completed. Upon finishing a session, participants provide feedback on their physical exertion. Based on these data, the training progression is automatically adjusted and advances to higher levels with more challenging exercises, such as the tandem stance or walking backward. In addition to physical exercises, KOKU incorporates 4 health literacy games that address potential tripping hazards and promote healthy nutrition. Participants were encouraged to use KOKU 3 times per week over a 3-month period. KOKU sends automatic reminder messages after 1 week of inactivity to promote adherence.

Home visits for the introduction of KOKU and assessments were conducted by trained assessors with professional backgrounds in physiotherapy or sports science, following a standardized protocol to ensure consistency and reliability.

Pragmatic considerations, including logistical and financial constraints, limited data collection to a subset of participants from the SMART-AGE study, resulting in a final subsample requirement of 116 participants.

Descriptive statistics, including means and SDs or median and range, were calculated for all outcome measures. The distribution of the variables, including assessments of normality, is presented in Multimedia Appendix 1.

To assess differences between premeasurement and postmeasurement, 2-sided paired t tests were conducted. Effect sizes were determined using Cohen d, with values interpreted as small (0.2), medium (0.5), and large (0.8) effects [36]. A significance level of α less than or equal to .05 was set for all statistical analyses; P values were Bonferroni-Holm adjusted for each instrumented measurement, that is, the instrumented Timed Up and Go Test (iTUG), the instrumented 30-Second Chair Rise Test (iCRT), and real-world walking outcomes.

In addition, regression analyses were conducted to identify potential determinants of physical capacity in instrumented assessments. For each instrumented measurement, we chose the variable with the highest effect size to conduct a regression analysis. A change score was calculated by subtracting the premeasurement from the postmeasurement for the total time in the iTUG and the percentage of inactive time in the iCRT, respectively. A negative change score meant a decrease in time, which signifies an improvement in iTUG performance, and a positive change score in iCRT score indicated an improvement. The change score was then entered as the dependent variable in a multiple linear regression model. Since we expected KOKU usage to contribute to an improvement in performance, we included the highest reached KOKU level at postmeasurement and the average days the participant had been active on the app per week as predictors. In addition, we wanted to investigate the influence of baseline performance as well as its interactions with the KOKU usage.

All statistical analyses were performed using the statistical software SPSS (version 29.0; IBM Corp.) and RStudio (version 2023.09.1; Posit Software).

In total, data of 113 participants were included in the following analyses (Table 1). The number of dropouts was lower than assumed (n=5, 4.2%). Two participants went on holiday without taking the tablet with them, and 3 participants were excluded because of missing postmeasurement data after 3 months.

The average duration between premeasurement and postmeasurement was 95.62 (SD 8.49; range 77‐146 days). Throughout this period, no falls or other adverse events were reported while or due to using KOKU.

Participants trained with KOKU for an average of 24 (SD 15; 1-71 days). On these training days, participants spent an average of 12.90 (SD 7.36; range 0.08‐64.32 min) using KOKU. The average break duration between training days was 3.75 (SD 6.99; range 1‐91 days).

The total number of exercises performed during the study period ranged from 2 to 213 exercises, with a mean of 75.9 (SD 47.3) exercises. On each training day, participants performed an average of 3.17 (SD 0.98; range 1‐18) exercises with 22.1 (SD 24.9; range 0‐250) repetitions for each exercise leading to an average time spent on exercises within KOKU of 6.95 (SD 4.16; range 0.13‐33.03 minutes).

About half of the participants (n=56, 49.6%) had reached level 5 or higher, with the largest proportion at 28.3% (n=32) of the participants reaching the maximum level 7. Sixteen (14.2%) participants did not complete the first level. Out of these, 7 (6.7%) participants did not use KOKU at all.

Regarding the health-literacy games, 67 (59.3%) participants played at least 1 health literacy game within KOKU, playing an average of 7.82 (SD 8.95; range 1‐53) games per participant.

In the first 3 weeks, participants exercised on average more than 2 days per week, with the highest median training frequency observed in weeks 2 and 3 (Figure 1). From week 4 onward, a gradual decline appeared with around 1.5 days per week between weeks 5 and 10, reaching the lowest point in week 12. The total time spent using the KOKU program per week averaged 32.18 (SD 3.48) minutes and showed a slight increase during the initial weeks, peaking at 40.62 minutes in week 10 before declining to 26.45 minutes by week 14. The actual time spent exercising per week followed a similar trend, reaching a maximum of 13.71 minutes in week 10 before decreasing thereafter (Figure 2). The number of exercise repetitions rose from 24.55 in week 1 to 33.85 in week 4 and then remained relatively stable, ranging between 30 and 34 repetitions per week.

The aim of this study was to evaluate the extent to which KOKU was implemented as intended. This process evaluation assessed both adherence to the protocol and participant engagement. As a digital intervention, KOKU represents a complex health care initiative that requires process evaluation to determine the fidelity and quality of its implementation, as well as the influence of contextual factors. It was anticipated that KOKU might require adaptation across different settings or populations and that the evaluation would reveal both strengths and weaknesses in its delivery.

While pre-post outcome measures were included to explore potential changes over time, this study was not designed to provide definitive evidence of effectiveness. Therefore, the observed outcome trends should be interpreted as exploratory and descriptive rather than causal. The combination of implementation-focused evaluation with preliminary outcome data provides insights into how the intervention was delivered and received, while also generating hypotheses for future controlled effectiveness studies.

No falls or other adverse events were reported by the KOKU participants, indicating high program safety. These aspects are of indispensable importance for any digital intervention targeting populations in unsupervised home settings.

Participants engaged with KOKU on 2 to 3 training days per week, which is in line with the recommendations [14]. Engagement levels varied among participants, with some using the system only once while others used KOKU extensively. General adherence trends suggest that engagement with KOKU showed good initial engagement in the first 3 weeks, followed by a gradual decline over time. This pattern of declining engagement is consistent with previous research on digital health interventions [38]. However, despite this decline, the total number of exercises performed remained relatively stable, which could suggest that those who continued using KOKU maintained a consistent level of exercise activity. Moreover, the increase in exercise repetitions in the first 4 weeks, followed by stability, reflects the observed usage pattern in the logged training data rather than indicating a specific behavioral mechanism. As no subjective measures (eg, participant-reported reasons for engagement or disengagement) were collected, the underlying reasons for these trends cannot be determined. However, the collection of such subjective data is currently being implemented in a follow-up study with KOKU to better understand factors influencing user engagement and adherence [39].

The time spent on KOKU exercises remained relatively short. The recommendation by the KOKU developers is at least 10 minutes per session for inactive older adults with prefrailty to moderate frailty. In comparison, the conventional Otago program recommends training sessions of 30 minutes or more [30]. This limited training duration may have contributed to the limited improvements, alongside recruiting more active older adults.

By the end of the study, nearly half of the participants had reached level 5. To reach this level, participants had to use KOKU regularly for at least 8 weeks without feeling unstable, indicating that a substantial proportion of users demonstrated sustained engagement and progression in their training.

A substantial proportion of participants did not complete the first level or did not use KOKU at all, highlighting potential barriers to engagement. It is important to note that the study cohort was relatively fit at baseline, as indicated by the motor assessment data reported [40]. This characteristic is likely related to the study’s recruitment criteria, which required a certain level of digital literacy to participate [19]. This requirement may have introduced a selection bias toward individuals who were healthier, more functionally independent, and potentially more motivated than the general target population. As a result, this may have led some participants to perceive KOKU as insufficiently challenging, resulting in minimal or no use. However, other barriers such as individual differences in motivation, usability challenges, or others may also have played a role.

Compared to traditional analog assessments, which often rely on subjective interpretation, instrumented measurements offer a more comprehensive and standardized way of capturing movement data. This can be particularly valuable in clinical and research contexts where even small changes might indicate early functional decline or emerging effects of an intervention such as KOKU.

The iTUG results indicated an increase in total time needed to perform 1 repetition. This increase in duration may reflect performance variability rather than a meaningful decline in capacity. Analysis of individual gait segments showed no significant differences between premeasurement and postmeasurement. Further investigation is warranted to determine whether the observed trend of an increase in the number of steps during the iTUG represents a robust effect that reflects a compensatory adaptation or a trial-to-trial variability. Overall, the iTUG results suggest that mobility performance measured with the iTUG remained relatively stable between premeasurement and postmeasurement, with only minor variations that are unlikely to be of significant clinical concern. Future studies could explore whether these small changes persist over time or whether they are influenced by factors such as fatigue, learning effects, or external conditions.

The results of the iCRT indicate that participants performed slightly better in postmeasurement compared to premeasurement, as evidenced by small increases in the number of fully completed repetitions, the number of repetitions based on the complete 30 seconds, and the frequency of repetitions. However, these differences were small and are unlikely to represent clinically meaningful improvements. Rather, they may reflect minor variations in performance, familiarization with the task, or day-to-day variability. The most notable finding was the reduction in inactive time, with a medium effect size. This suggests that, during the postmeasurement, participants spent less time pausing or resting between sit-to-stand and stand-to-sit transitions. While this change may indicate a difference in movement execution, its clinical relevance remains uncertain in this sample. The iCRT findings highlight the importance of considering not just repetition counts as in conventional 30-second CRT, but also more detailed metrics of movement execution, such as inactivity between movement segments. Future research should explore whether the observed reduction in inactive time during iCRT translates into clinically relevant functional improvements in different populations, particularly in rehabilitation settings or frail populations. Additionally, investigating potential factors influencing inactivity, such as fatigue, motivation, or motor control, could provide further insights into optimizing performance.

In line with the capacity measures, the real-world walking measures revealed minimal changes between premeasurement and postmeasurement across most parameters, indicating that participants’ everyday mobility patterns remained largely stable. A small reduction in walking speed during walking bouts of >30 seconds was observed; however, this change was small and unlikely to reflect a clinically meaningful decline in functional mobility [41].

Overall, changes in both capacity and real-world measures were minimal, likely due to the characteristics of our study sample. About 82% (n=93) of the participants performed below the 12-second threshold in the TUG, reflecting overall good mobility [42]. Their real-world walking metrics also aligned with those of healthy community-dwelling older adults from the inCHIANTI cohort [40]. This relatively high baseline fitness may also help to contextualize the small observed changes in motor outcomes. Larger effects might be expected in a frailer population with more potential for improvement through KOKU usage.

Nevertheless, the use of instrumented assessments to analyze premeasurement to postmeasurement changes revealed more nuanced and detailed insights that cannot be captured by traditional analog methods. The findings open up new avenues for understanding changes in motor strategies and highlight the potential of instrumented analysis to uncover early or otherwise hidden shifts in performance.

The regression analyses examined associations between KOKU usage, baseline performance, and changes in functional mobility outcomes. In the absence of a control group, causal interpretations are not possible, and the findings should be interpreted as exploratory associations.

For the iTUG, the multiple regression model explained a modest but significant proportion of the variance in change scores. Higher KOKU training frequency was associated with greater improvement in iTUG performance. Additionally, participants with poorer baseline performance tended to show larger improvements, which may reflect greater potential for change in this group.

In contrast, the iCRT model showed stronger predictive capacity for changes in inactive time, a key indicator of movement efficiency. Here, only baseline performance predicted change: participants with more inactive time at baseline tended to show greater improvements. KOKU usage variables (ie, training frequency and level reached), however, showed no significant associations with iCRT outcomes.

In the future, the variety of KOKU exercises should be expanded, including stratified progression and adaptive difficulty, allowing users with different levels of balance and strength to train at an appropriate level and achieve further progress.

In this study, all instrumented measurements were carried out by assessors. In the real world, however, digital interventions that can be used unsupervised and independently need to be combined with digital self-assessments to provide a comprehensive digital prevention strategy. An iTUG self-assessment could serve as an orientation tool to guide exercise selection rather than as a strict eligibility criterion. For example, users with slower iTUG performance (>12  s) could be guided to exercises targeting measurable improvement, while fitter users (<12  s) might benefit primarily from maintaining mobility and functional capacity. Adaptive difficulty would enable these users to remain challenged without expecting large gains in performance, supporting long-term function and engagement.

The iTUG self-assessment could be repeated on a regular basis, for example, every 3 months, to enable self-monitoring and evaluation of training progress, as well as to adjust exercise selection accordingly. The mTUG medical device system used in this study, however, is designed for supervised measurements in clinical or laboratory settings and not recommended for self-assessment. A recent study demonstrated that a smartphone-based iTUG self-test (Up&Go app) yields comparable results to the validated mTUG system with excellent concurrent validity and test-retest reliability [18]. Five test rounds are performed within this app, allowing users to familiarize themselves with the test procedure and minimizing potential variability in test performance due to initial uncertainty in self-assessment [18]. The Up&Go app is freely available for download on all Android and iOS smartphones in German-speaking countries. It was co-designed with older adults and includes comprehensive information and video instructions for independent test preparation and execution [43]. From our point of view, the integration of a self-assessment into KOKU usage would add considerable value to increase the target group-intervention fit and improve users’ self-management. Similar types of self-assessment would also be conceivable for the iCRT.

Several limitations should be acknowledged. The effect sizes for most performance metrics were small; they did not reach the level of minimal important differences. By definition, this process evaluation is not a causal interpretation regarding the effects of the KOKU training. The study did not investigate the underlying reasons for declining engagement, which could be influenced by personal motivation, external life factors, or the perceived difficulty of the exercises. In addition, activities outside of KOKU—such as training in a sports club—were not recorded. Gaining qualitative insights into user experiences and exercise habits beyond KOKU usage could offer a deeper understanding of the factors that facilitate or hinder engagement. As part of a single-blinded randomized controlled trial, this was not feasible in this study. While instrumented assessments offer numerous advantages, they are also prone to technical errors. Incorrectly recorded data may only be recognized retrospectively and cannot always be reconstructed. This issue occurred with the iCRT and the real-world measurements, resulting in a reduced sample size.

This study demonstrated that KOKU is a feasible and safe digital training intervention for older adults, offering an accessible way to empower them to independently engage in balance and strength exercises at home. While initial engagement was strong, usage declined after 3 weeks. This might require a phone call or teleconsultation to improve mid-term adherence.

The study sample consisted mainly of individuals with good physical capacity and function. With this in mind, the small changes in physical capacity and real-world walking highlight the importance of better tailoring of such interventions to individual levels of physical capacity, with potential improvements through more challenging exercises. These findings may support a stratified training approach with adaptive difficulty levels. Instrumented assessments provided valuable insights. Moreover, combining digital interventions with self-assessments, such as iTUG self-testing, could enhance user autonomy and optimize the fit of the intervention to individual capacities. Overall, digital training interventions, such as KOKU, have the potential to complement traditional care models, offering older adults an effective means to maintain their mobility, independence, and quality of life.