A secondary analysis of INSPIRERS observational studies involving 32 secondary care centers (allergy, pulmonology, and pediatrics departments) and 17 primary care centers in Portugal was performed (Figure 1), as part of the INSPIRERS project. The design of the INSPIRERS observational studies was disseminated through email contacts, social networks, and oral communications at national meetings/conferences, and physicians/centers interested in being part of the study contacted the research team. A convenience sample of adolescents and adults with persistent asthma was recruited for the INSPIRERS studies between November 2017 and August 2020. Depending on the study, each center was asked to recruit a minimum of 2-10 patients. The 3 INSPIRERS observational studies address the topic of adherence to asthma inhalers among adolescents and adults with persistent asthma (Figure 1). Further details on the project setting and methods have been previously published [20,21]. This study is reported in accordance with the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines [22].

Data were collected using a questionnaire during a face-to-face visit. The questionnaire had a section to be completed by the physician addressing patients’ asthma characteristics and a set of questions to be self-completed by the patient, as detailed below. Patients were included if they had a previous medical diagnosis of persistent asthma, were at least 13 years old, and had an active prescription for an inhaled controller medication for asthma. Patients were excluded if they had a diagnosis of a chronic lung disease other than asthma or a diagnosis of another significant chronic condition with possible interference with the study aims.

Users and nonusers were defined as individuals who answered “yes” or “no,” respectively, to the question “Have you ever downloaded and used a health and fitness app?” Health and fitness apps were defined as a range of apps related to personal fitness, workout tracking, diet and nutritional tips, health and safety, etc. In accordance with the conceptual model proposed by Andersen et al [19], variables collected included predisposing factors, enabling factors, and need (Figure 2). Predisposing factors included demographic data (ie, age, gender, marital status, parish, and postcode), and enabling factors included education level, use of smart devices, and digital literacy, both collected from patients. Digital literacy was defined as the median of 5 items of the Media and Technology Usage and Attitudes Scale (MTUAS, ie, use of the GPS, browsing the web, taking pictures, gaming, and checking social networks) rated by frequency of use in a 10-point Likert scale (1=never to 10=all the time) [23]. Additionally, socioeconomic level was explored as an enabling factor, which was defined as the Portuguese ecological deprivation index, extracted from the patient residence information (civil parish/postcode), and categorized into 5 quintiles (Q1=least deprived to Q5=most deprived) [24]. Need variables included smoking status, patients’ perceived overall health status (from EQ-5D Visual Analog Scale [VAS], ranging from 0 [worst imaginable health state] to 100 [best imaginable health state]) [25], the presence of anxiety or depression (cut-off≥8 in the Hospital Anxiety and Depression Subscales) [26], and physicians’ input, including asthma control level (uncontrolled, partially controlled, or well controlled according to the classification of the Global Initiative for Asthma [27]), number of exacerbations (episodes of progressive increase in shortness of breath, cough, wheezing, or chest tightness, requiring a change in maintenance therapy) in the past year [28], and the number of unplanned appointments in the past year.

Age was categorized into age bands (13-18,18-30, 30-40, 40-50, 50-65, and ≥65 years). Other continuous variables (socioeconomic level, median digital literacy, and overall health status) were categorized into quartiles.

Counts and proportions were calculated for each variable. Multivariate logistic regression with the Enter method was used to ascertain the determinants of being a user of a health and fitness mobile app (dependent variable) in accordance with Andersen et al’s [19] conceptual model. Categorical variables, such as gender, age, marital status, educational level, socioeconomic level, use of smart devices, digital literacy, overall health status, smoking status, presence of anxiety or depression, and asthma status were explored as independent variables, irrespective of significance in preliminary univariate logistic regressions. We assessed model fit using the pseudo-R2 Nagelkerke method and tested for evidence for poor model fit using the Hosmer–Lemeshow test. In addition, to assess the increasing contribution of each covariate to the model, we adopted a stepwise approach to build models starting from 1 covariate until including all covariates in the full models. For each model, we computed the described goodness of fit statistics. The quality of the final model was also assessed using the Aikake Information Criterion. Adjusted odds ratios and 95% CIs were calculated. Statistical analyses were conducted using R and the “glm” package. The map of Portugal was created using Paintmaps [29].

The studies were approved by the ethics committees of all participating centers. The studies were conducted in accordance with the ethical standards established in the Declaration of Helsinki. Eligible patients were approached by physicians during medical visits and invited to participate. Written informed consent was obtained before enrollment. Adult patients signed a consent form; adolescents signed an assent form and a parental consent form was also obtained.

Use of health and fitness mobile apps was positively associated with a single status, >10 years of education, and higher digital literacy scores, and negatively associated with depressive symptoms. No significant associations were found with other variables, including gender, age, socioeconomic level, general health status, smoking status, anxiety, and asthma control.

According to our results, single participants are more likely to use health and fitness apps. In a recent mixed methods study, Zhou et al [18] explored the barriers to and facilitators of the use of mobile health apps and found that single users had less strong concerns about information security and privacy and less desire to have stringent security protection, which could contribute to higher usage levels in this group.

Consistent with our findings, previous studies have described educational attainment as an important predictor of use of mobile devices and apps [14-16]. However, the relationship may be more complex than initially predicted, as in another study by Carroll et al [14] where both patients with a degree and those with less than high school education were significantly associated with a reduced likelihood of using health apps. The reasons for the educational differences are not fully understood but may reflect the effect of digital skills and confidence, and social norms related to the perceived value of using health and fitness apps [16].

A significant association between digital health literacy and use of health and fitness apps was also found. Digital health literacy has been previously shown to affect the use of health apps [17]. However, comparisons between different studies are limited by the heterogeneity of tools used to evaluate patients’ digital health literacy; therefore, standardization of the assessment methods used is recommended in the future. It is also important to note that although individuals with low general health literacy tend to use less health information technology [31], previous evidence has shown that tailored approaches including apps programmed with computer-animated characters, text, and graphics to provide health communication and education could be a widely accepted option for these patients [32].

Interestingly, our study found a negative association between the use of health and fitness apps and the presence of depressive symptoms. Despite the breadth of research exploring the determinants of use of mental health apps [33-35], there was a lack of evidence specifically exploring the impact of the presence of mental health symptoms on usage rates. However, several studies show that patients with depressive symptoms often exhibit poor engagement with health services, and health care avoidance [36,37], and such behaviors may also contribute to a lower interest and usage of health and fitness apps by these groups of patients. This finding opens a new research avenue, and future work should explore the drivers for this effect and should involve patients with mental health problems in the co-design of digital solutions and interventions that promote both early and sustained use. The other needed variables, such as the ones related to asthma control, were not significant in explaining mHealth behaviors, although they showed a similar trend to the one observed with the variable depression. This is in line with previous asthma mHealth studies showing that patients with worse asthma control engage less with the tested apps [15,38]. Regarding age, odds ratios were higher for patients aged 18-30 years (1.93, 95%CI 0.97-3.89), but no significant differences were detected, despite the P value being close to the significance threshold (P=.06). Increasing the sample size in future studies could help further explore this effect and to confirm or exclude a potential type II error [39]. Previous studies found that younger adults were more likely to engage with health apps [14-16] and suggested that the effect of age likely reflects both social norms and cohort effects, such as the increased exposure to these devices and apps at younger ages [14]. Previous evidence also suggests that younger adults seem to have higher digital health literacy levels, which can contribute to increased use of digital solutions [40]. While it is recognized that the use of digital health solutions by older persons could improve patient engagement and reduce both financial burden and pressure on health systems, usage rates among this group remain low. According to a mixed methods study conducted by Fox et al, this digital health divide is deepening owing to older adults’ perceived inability and unwillingness to use digital technologies, stemming from mistrust, high-risk perceptions, and a strong desire for privacy [41].

Finally, no significant associations were found with gender. Previous literature shows mixed evidence on this subject: while some studies had found a higher use among male subjects [15], others reported the opposite [14,16]. The reasons for gender differences in some samples are unclear but may reflect sample-specific differences in health-seeking behavior, and interest and participation in healthy lifestyle interventions in general.

This study has several strengths. It is the first national-level study performed in Portugal, evaluating the use of health and fitness apps among patients with asthma, covering the majority of the geographic regions of the country. A comprehensive set of individual-level characteristics was collected and analyzed, which allowed us to explore the impact of a range of sociodemographic factors, health literacy, and cofactors such as general health status and asthma status. Although no power calculation was performed (which was associated with the secondary analysis nature of this study), the overall large sample size contributes to the robustness of these findings.

Other limitations also need to be acknowledged. Intrinsic to the study design using convenience sampling, a potential selection bias cannot be excluded. This can possibly explain the low number of patients above 65 years of age. Nevertheless, this risk was mitigated by sampling patients from different health care settings, centers, and geographic regions. Health and fitness app use was patient-reported; therefore, a potential information bias cannot be excluded either. As an alternative, future studies could use patient log-in as a measure of app use. Furthermore, digital literacy was assessed using selected items of MTUAS scale, not the complete instrument. This choice emerged as a mitigation measure to reduce the data collection burden and allow us to efficiently collect data on an aspect seldom reported in the literature. Future studies should explore the possibility of including complete validated tools, such as MTUAS [23] or the eHealth Literacy Scale [42]. We also need to consider that although the questionnaires were independently answered by the study participants, they were provided to them by their family physicians provided during an in-person visit, which may have influenced them to give desirable responses, namely those related to their general health status and asthma control.

Finally, most subjects included in this study were relatively young, had at least 7 years of education, were in the 3 higher quintiles of the socioeconomic deprivation index, had an overall good health status, and had good asthma control. Consequently, attempts to generalize these findings to other populations that do not share the same characteristics need to be cautious. Future studies should consider involving patients with specific needs (ie, lower education attainment, lower sociodemographic levels, poorer overall health status, or poorer asthma status control) and evaluate the replicability of these findings.

Our results show a negative association with low literacy and the presence of depressive symptoms, highlighting the need for future research to explore the effect of disparities on app use, particularly among those with lower digital health literacy or mental health issues. Importantly, such studies should also assess participants’ general health-seeking behavior (ie, interest and participation in healthy lifestyle interventions in general) and the key constructs of their acceptance the unified theory of acceptance and use of technology) [43,44]. This theory was developed through a consolidation of the constructs of 8 previous models (theory of reasoned action, technology acceptance model, motivational model, theory of planned behavior, a combined theory of planned behavior/technology acceptance model, model of personal computer use, diffusion of innovations theory, and social cognitive theory), and includes four constructs: performance expectancy, effort expectancy, social influence, and facilitating conditions. Gender, age, experience, and voluntariness of use have been suggested to moderate the impact of the 4 key constructs on usage intention and behavior [43,44].

A better understanding of which patients with asthma are (and are not) using general health and fitness apps is key to design tailored mHealth interventions to improve sustained use by this specific group of patients. Additionally, this knowledge can also inform high-level delivery strategies to ensure that these solutions reach out comprehensive groups of patients with asthma, thus improving rather than entrenching health inequities within this population.