The study was declared to not fall within the scope of the Dutch Medical Research Involving Human Subjects Act by the Medical Ethics Committee Leiden-Den Haag-Delft (N19.0878) as the study was non-invasive. Participants signed a digital informed consent form before participating. To ensure privacy, the data set displayed only participant numbers (ie, pseudonymized data), and the combination of participant numbers and email addresses was securely stored in a separate protected file. Participants were not reimbursed for completing the questionnaires.

The current quantitative study is part of a mixed methods study with a prospective pre-post interventional design (Figure 1). The qualitative results (ie, focus group interviews) are reported elsewhere [16]. This study used web-based user data and questionnaires as obtained over 5 measurements over a period of 8 months: T0 (baseline), T1 (60 days), T2 (120 days), T3 (180 days), and T4 (240 days) after baseline, respectively. The study was declared to not fall within the scope of the Dutch Medical Research Involving Human Subjects Act by the Medical Ethics Committee Leiden-Den Haag-Delft.

A total of 2562 participants were recruited on the web and offline through social media posts, newsletters, advertisements in local newspapers, and personal communication by various institutions in the Netherlands. These institutions included health insurance companies, a health care coalition, an academic hospital, the Ministry of Health, Welfare, and Sport, an employers’ organization, local governments, municipality teams, a knowledge and quality institute for oncological and palliative care, as well as a web-based platform for patients with cancer and their caregivers. Interested individuals could submit their email addresses, and they were subsequently invited to participate in the study. The invitation email included a web-based information letter about the study and a link to the research environment, where individuals were screened for eligibility through self-report questions. The inclusion criteria were as follows: (1) being aged 18 years or older; (2) being able to understand, read, and speak the Dutch language; and (3) having access to the internet. If eligible, individuals were asked to provide digital informed consent and to complete the web-based baseline questionnaire (T0). Hereafter, individuals were provided access to the health app platform.

Participants received access to the FitKnip health app platform, which enabled them to purchase preselected, trustworthy apps. Apps were preselected by a health care coalition based on the following criteria: (1) the app must be ready for use and can be used without the involvement of a health care professional; (2) the app should not include in-app purchases; (3) the provider of the app must adhere to national and international laws, such as the General Data Protection Regulation; and (4) the app should enable anonymous purchases, and the provider of the app should not be allowed to publish or sell user data.

A total digital budget of 100 euros was available for each participant. The budget could be used during the entire research period of 8 months. The platform offered a total of 38 apps, which could be categorized into 1 or more of the six dimensions of positive health [15], namely, (1) bodily functions, (2) mental functions and perception, (3) spiritual or existential dimensions, (4) quality of life, (5) social and societal participation, and (6) daily functioning. For example, apps can help users live more healthily, sleep better, process grief, or communicate more clearly.

The majority of these apps were progressive web applications (PWAs). PWAs are web apps that offer mobile app–like experiences while giving the user a faster and more reliable version of the app. These PWAs are, from here on, referred to as apps in this manuscript.

All quantitative analyses were performed in SPSS (version 24.0; IBM Corp). Descriptive analyses, that is, mean (SD), median (IQR), and n (%), were used to describe the sociodemographic and clinical characteristics of the study population, as well as study outcomes. Health outcomes were evaluated for changes over time. Therefore, the health outcomes on T1-T4 were one-by-one statistically compared to T0 using linear mixed model analyses. These models included a random intercept to account for within-subject correlation among repeated measures, as through the –2 restricted likelihood test it was determined that including random slopes led to a better model fit. Time contrasts were created using dummy coding (T0=1; T1=2; T2=3; and T4=3). Age and gender were forced into the models as covariates. Next to this, a stepwise forward selection approach was used to select potential confounders and covariates in the associations between time and health outcomes, more specifically education and work status. However, the addition of both education and work status did not lead to a change of >10% in β coefficient; hence, they were not included in the models. Furthermore, we evaluated whether change over time for health outcomes differed between subgroups (ie, effect modification) based on age categories, gender, and diagnosis. The variable diagnosis was an effect modifier in the SF-12 PCS and MCS scores and the PSS-10 scores. Therefore, the results for these outcomes were reported separately for subgroups of participants with and without a diagnosis.

The sociodemographic and clinical characteristics are shown in Table 1. The average age of the study population was approximately 45 (range 18-81) years. The majority were female (n=1177, 71.3%), highly educated (n=1307, 79.2%), had a full-time (n=722, 43.8%) or part-time (n=450, 27.2%) job, and was living together with a partner (n=545, 33%) or with both a partner and children (n=580, 35.2%). Approximately half of the study population had a healthy BMI, whereas roughly 30% were overweight and 18% were obese.

Approximately one-third (592/1650, 35.9%) of the study population reported having a medical, physical, or psychological diagnosis, and 38.5% (635/1650) of the participants reported health care use in the past month, of whom most went to a general practitioner (310/635, 52.4%) or a medical specialist (276/635, 46.6%).

Most participants scored on or above the middle of the scale on the eHLQ. These results suggest sufficient-to-good eHealth literacy skills (Table S1 in Multimedia Appendix 1).

Platform use data showed that 42% (685/1650) of participants purchased at least 1 app (Table 2). Of those, the majority of participants purchased 1 app (244/685, 35.6%), followed by 2 (175/685, 25.5%) and 3 apps (104/685, 15.2%). Only a minority of participants (18/685, 2.5%) purchased over 10 apps. At T1, the majority of participants purchased 1 app (291/685, 42.5%), whereas at T2, T3, and T4, most participants did not purchase any apps (512/685, 74.7%; 623/685, 90.9%; and 601/685, 87.7%, respectively). Detailed results on the budget spent per end user can be found in Table 2. The top 5 apps purchased by participants were as follows: (1) an activity tracker (173 times), (2) a mindfulness app (171 times), (3) a healthy living app (142 times), (4) a communication app providing insight and exercises in personal communication styles (121 times), and (5) an app targeting self-image (109 times).

The 3 most selected reasons for participants to try out the health app platform were to improve general health (989/1650, 59.9%), be interested in health (913/1650, 55.3%), and improve physical health (696/1650, 42.2%) (Table S2 in Multimedia Appendix 1). The most selected app interest areas among participants were “bodily functions” (1335/1650, 80.9%) and “mental functions and perceptions” (938/1650, 56.8%); the least selected interest areas were “spiritual dimension” (512/1650, 31%) and “social and societal participation” (368/1650, 22.3%).

Table 3 presents the results concerning the acceptability of the health platform. In general, user friendliness, as assessed by the SUS, was rated above average. The acceptability of the health app platform, as assessed by the CSQ-8, was scored as moderate. Furthermore, results showed that on average, participants were “satisfied” to “highly satisfied” with the ease of the payment system, the look and feel of the website, and the provided budget of 100 euros. Regarding the provided budget, approximately 40% (660/1650) of participants rated it a 10 on a scale of 1 to 10 (range T1-T4 38.5%-39.4%). Participants were less satisfied with the amount and type of apps. Around 40% (660/1650) scored these elements as insufficient (≤5), respectively, ranging from 39.8% to 47.2% and from 41.1% to 45.1%.

Concerning health empowerment, in general, participants did not perceive the health app platform as “supporting” (Table 3), especially in terms of the health app platform supporting them in finding appropriate help (range 2-3).

Both the physical and mental subscale scores of the SF-12 were stable across the research period, with scores below the norm of 50 for the general population (Table 4).

The subgroup analysis based on diagnosis showed that the mean MCS score for participants without a diagnosis was significantly higher on T1 (48.7, SD 9.1) compared to T0 (47.0, SD 9.4; β=1.15, SE 0.36; P<.001; Table S3 in Multimedia Appendix 1).

The PCS score on T1 was significantly lower, and thus declined, for participants with a diagnosis (β=–1.60, SE 0.74; P=.03). The PCS score for participants without a diagnosis was higher, and thus improved, on T4 (β=0.98, SE 0.44; P=.03; Table S4 in Multimedia Appendix 1).

Table 4 shows that participants’ perceived stress levels remained quite stable across the research period, indicating low to moderate levels of stress.

The subgroup analyses based on diagnosis showed that PSS-10 scores were significantly lower, and thus improved, for participants without a diagnosis on T1 (mean 12.8, SD 6.0; β=–0.91, SE 0.22; P<.001), T3 (β=–0.60, SE 0.29; P=.04), and T4 (β=–0.82, SE 0.29; P<.001) compared to T0 (mean 14.0, SD 6.1; Table S5 in Multimedia Appendix 1).

This study aimed to evaluate the feasibility and acceptability of a health app platform where individuals were given a budget of 100 euros to purchase preselected, trustworthy health apps (primary aim), and to explore the potential impact on health empowerment and health outcomes (secondary aim). Results showed that of the 1650 participants, only approximately 42% (685/1650) purchased at least 1 app during the 8-month-long experiment. The number of app purchases among these participants was rather low, and the majority of participants had budgets left to purchase additional apps. In general, participants were satisfied with the health app platform and found the platform easy to use; participants were satisfied with the look and feel of the platform, the ease of the payment system, and the provided monetary budget. However, they were less satisfied with the number and diversity of apps offered on the platform. Results further suggested that the health app platform in its current form did not seem to contribute to participants’ health empowerment and outcomes.

Although during the current experiment, the user engagement stayed behind what was expected, the results of the qualitative study about this experiment showed that participants were enthusiastic about the concept of a health app platform to promote health and vitality. Furthermore, a previous study showed the added value of a health app platform called Intellicare with regard to the number of individual health app downloads [30]. Intellicare is a hub app, which is a catalog for individual apps targeting depression and anxiety [30,31]. The results of Lattie et al [30] showed that the number of individual health app downloads among hub app users was higher as compared to nonhub app users. Thus, while user engagement with the FitKnip platform was not as expected, the concept and function of a health app platform have shown potential in other studies.

A health app platform is meant to guide individuals to suitable health apps; however, in current initiatives, this process is not sufficiently supported [32,33]. Facilitating the selection process for individuals might be the key to higher app downloads.

Other factors that increase app downloads and user engagement on health app platforms are the incorporation of personalized reminders and personal contact or coaching [30-32]. Indeed, the results of a follow-up study of Intellicare suggest that personal coaching through a phone call and SMS text messages, aimed at encouraging engagement with the hub app and individual health apps, can indeed be an effective strategy for increasing app downloads [31]. Participants in the qualitative study also suggested incorporating contact with professionals into future health app platforms [32].

Studies of individual apps also offer ideas on how to increase engagement on a health app platform [34-37].

During the experiment, the study dropout increased over time; only approximately 1 in 4 participants completed the 8-month follow-up questionnaire. Such high study dropout rates are common in digital health studies [38-42]. Our results showed that participants with a medical, physical, or psychological diagnosis were less likely to drop out of the study. This is in line with previous literature, where low dropout levels were observed for individuals who perceive their own health to be poor [43], as well as those who want to be involved in their health care and who are interested in self-monitoring their health status [44]. These studies conclude that individuals with poor health, a chronic disease, or a family history of disease are more willing to work on their own health care [43,45,46]. More generally, by understanding different types of users and their motivations for using health apps, developers can create platforms that are better tailored to meet the needs of end users, or they can even create separate platforms for specific target groups and improve engagement accordingly [32,44].

The results indicated that, in general, the current health app platform did not meet participants’ needs in terms of supporting health empowerment, that is, in gaining the knowledge, skills, and confidence needed to take action to improve their health. This is in line with the results of this qualitative study, which showed that some participants believed that a certain level of health empowerment is a prerequisite for the use of a health app platform and that the health platform first needs to be improved to fulfill its potential [32]. Additionally, no major changes were observed in health outcomes over time. The platform should first be improved to better support individuals in their health in terms of the ability to select apps, the app catalog (eg, more physical health–related apps), and personalization in terms of language and type of apps offered [32]. More research is needed to gather more in-depth information about the barriers and facilitators to health empowerment and health outcomes, and how a health app platform could best support users in this process.

To our knowledge, this is the first national, government-initiated experiment that aims to tackle barriers to the uptake of health apps by providing participants access to a platform of preselected, trustworthy health apps and providing them with a monetary budget to purchase these apps. The platform is therefore an innovative way to provide individuals with the tools to work on their own health and vitality. A strength of this study is the substantial study population of 1650 participants, as well as the evaluation of FitKnip in a real-world setting, providing a rather realistic representation of how users engage with the platform in their everyday lives.

Some study limitations should be considered when interpreting the findings, one of which is the generalizability of the study results. The current study population was predominantly female, highly educated, aged 30 years or older, and diagnosed with a medical, physical, or psychological issue. The health app platform might have been more useful and interesting for participants with a diagnosis, as they might have been more willing to work on their own health care, which could have positively biased results on the feasibility and acceptability of the platform. Furthermore, as young individuals (aged between 18 and 30 years) were underrepresented in the study population, the current results may not be generalizable to a younger population.

The current results highlight the need for future health app platforms to place a strong emphasis on maximizing uptake and user engagement with the platform. This could be achieved by prioritizing the needs and perspectives of end users in the design and development of health app platforms and developing such a platform in cocreation to ensure that it is a valuable, empowering, and effective tool in promoting health and well-being. Cocreation is identified as essential for the success of eHealth initiatives [11,47]. Thus, it is crucial to involve end users already in the early stages of the development of an app or digital intervention through cocreation to increase engagement, adherence, and ultimately adoption.

A shift from reactive to proactive care is a crucial step in improving health outcomes and reducing the burden on the health care system. A health app platform can support such proactive care by enabling and empowering individuals to work on their health, vitality, and well-being. Hence, a health app platform presents a promising opportunity to enhance public health. To ensure its success, however, feasibility and acceptability are paramount, as these aspects ensure the platform’s accessibility, user-friendliness, and alignment with the needs and preferences of end users. Addressing these factors is instrumental in boosting uptake and engagement, and ultimately, the platform’s adoption and effectiveness.