According to the latest Global Burden of Disease Study, the drivers of increased Disability-adjusted life years (DALYs) from 1990 to 2019 include six health problems that mostly affect adults older than 50 years (such as ischemic heart disease and diabetes) and four that mostly affect individuals from 10 to 49 years (such as depressive disorders) (1). In 2019, eight of the top 10 risk factors for death and disability were behavioral or metabolic problems [such as high systolic blood pressure, smoking, high fasting plasma glucose, high body-mass index, high low-density lipoprotein (LDL) cholesterol, and alcohol use] (2). Such figures, in addition to the extensive literature on the matter (3–10), emphasize the importance of addressing unhealthy behaviors to prevent morbidity and mortality.

Lifestyle Medicine (LM) profits from evidence-based strategies to prevent and treat the progression of chronic diseases and improve quality of life (11). As per the American College of Lifestyle Medicine (12), the LM foundation is established over six main domains: diet, physical activity (PA), avoiding substance use (SU), sleep, social relationships, and stress management. Thus, LM aims toward a comprehensive change in unhealthy behaviors. Unhealthy behaviors tend to cluster and present an additive effect on decreased survival time without disability and earlier mortality (5, 8, 10, 13). However, most epidemiological studies do not include all the lifestyle domains considered in LM when evaluating health outcomes. Instead, unhealthy behaviors have been considered as independent risk factors. The multidimensional evaluation of lifestyle is becoming more frequent, either by using lifestyle indexes (5, 10) or including multiple lifestyle domains and testing their interaction effects (9). In this sense, our research group recently developed and validated a scale for the multidimensional evaluation of lifestyle (14–17).

Digital health interventions (DHIs) may be understood as interventions delivered via digital technologies (such as apps, digital platforms, and wearables) to improve the health of individuals (18–21). DHIs have increasingly been used to promote behavior change and a healthier lifestyle. For instance, a study conducted in Australia (June 2018–July 2018) searched for digital apps, using the words “health” and “well-being” (in the Australian iTunes and Google Play), and retrieved 2,12,352 related apps (22). In addition, DHIs have been advocated as a way to increase access to health, such as mental health (23), due to their relatively low cost and ease of scalability. However, the debate on how DHIs must be evaluated to ensure their quality and efficacy is far from over (24, 25). Due to the velocity and dynamic changes of DHIs, some argue that the gold standard of medical evaluation—randomized clinical trials (RCT)—would not be appropriate to DHIs assessment. Such velocity may also jeopardize the broader evaluation of effectiveness—which is usually made through systematic reviews and meta-analyses. Whenever there are multiple systematic reviews on the same topic it may be necessary to conduct a meta-review (i.e., the systematic review of systematic reviews) (26). Moreover, if the field is highly heterogeneous and/or wide, scoping meta-reviews may be necessary to summarize concepts, theories, sources, and knowledge gaps.

An important step in the advancement of DHI evaluation was the publication of the Evidence Standards Framework for Digital Health Technologies (27), in the United Kingdom (UK). Although developed to “demonstrate (DHIs) value in the UK health and social care system,” it may be useful for guiding other countries, especially those presenting universal healthcare systems. Regarding digital interventions for preventing unhealthy lifestyles, the guideline recommends a minimum of high-quality observational or quasi-experimental studies demonstrating relevant outcomes. The determination of relevant outcomes in LM, however, is not simple. In 2017, a meta-review on DHIs for cardiometabolic events pointed to the heterogeneity in study measures, such as DHI modalities, populations, and outcomes. Such heterogeneity precluded the formation of strong conclusions regarding the effectiveness of the evaluated interventions (28). In a meta-review evaluating the effect of DHIs on mental health, studies focusing on symptoms of anxiety or depression presented positive small/medium effect sizes (29). However, mental health symptoms are also heterogeneous and may not always represent a clinical psychiatric diagnosis, which could jeopardize future attempts to reproduce the findings.

Besides the outcomes, control groups also may represent a challenge. For instance, a meta-review evaluating DHIs for weight loss concluded that the interventions were more effective than minimal treatment but less effective than face-to-face interventions (30). Also, it is possible that some lifestyle domains, such as PA and diet, are of greater interest for addressing and evaluating DHIs due to the easily measurable outcomes associated with these domains. This hypothesis may make it even harder to evaluate lifestyle as a multidimensional construct, since easily measurable outcomes may lead to an unbalance in the field, with higher representation of lifestyle domains from which there is more available evidence.

Given the huge heterogeneity of DHIs and the specific challenges related to lifestyle research, in this article, we aimed to (1) identify the most frequent lifestyle domains evaluated (diet, PA, SU, sleep, social relationships, and stress management—as proposed by the American College of Lifestyle Medicine (12), and most common combinations of them); (2) describe the most frequent outcomes used to measure lifestyle changes in each domain or in multidimensional evaluations; and (3) to identify the most frequent delivery methods. Answering these questions and revealing research gaps is a crucial step for designing, implementing, and evaluating DHI in order for these interventions to have clinical and public health relevance.

The methodology followed Arksey and O'Malley's framework for scoping reviews (31) and the Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviews (PRISMA-ScR) (32). A scoping review is defined as a type of research synthesis that aims to map the literature on a particular topic or research area to identify key concepts, research gaps, and sources of evidence to inform practice, policymaking, and research.

The review included the following key phases: (a) formulating a research question, (b) Population, Intervention, Comparator, Outcome, Study Design (PICOS) acronym and eligibility criteria definition (c) identifying relevant studies, (d) study selection, (e) data extraction, and (f) collating, summarizing, and reporting the results.

Overall, 953 records were identified by searching the databases, and 1 record was additionally found through the manual search. Title and abstract screenings resulted in 152 included systematic reviews or meta-analyses reporting on the effectiveness of DHIs focusing on one or more of the lifestyle domains evaluated (i.e., diet, SU, PA, social relationships, sleep, and stress management) among adults. A total of 72 studies were included after full-text assessment, of which 35 conducted meta-analyses (Figure 1). All of them were published in English.

The systematic reviews included data ranging from a total of 2 (35) to 117 (36) RCTs. Overall, the sample size of the included RCTs ranged between 6 (37) and 69,219 (38) participants. Unfortunately, some systematic reviews did not report the size of the included RCTs. Most DHIs were compared to more than one alternative strategies. Table 2 shows the summary of systematic reviews included.

Of the 72 included reviews, only one evaluated all six lifestyle domains (i.e., diet, PA, SU, social relationships, sleep, and stress management) (100). In this study, the authors evaluated 14 web-based lifestyle programs designed to improve brain health among healthy individuals. A meta-analysis was performed including three RCTs. It showed a small-to-medium positive effect but high heterogeneity.

One systematic review evaluated five lifestyle domains (diet, PA, SU, social relationships, and sleep) (93). The effects of eHealth cardiac rehabilitation on health outcomes were evaluated in 15 trials, and meta-analyses were performed with data from 11 studies. Results indicated that interventions were effective in engaging patients into an active lifestyle, improving quality of life (QoL), and decreasing re-hospitalization. The authors also highlighted that empowerment components and tele-monitoring were crucial to the success of the interventions.

Five systematic reviews evaluated 4 lifestyle domains (59, 62, 74–76): all of them evaluated PA and stress management, 4 evaluated diet, 3 evaluated SU, 2 evaluated social relationships, and 1 evaluated restorative sleep. Most of the studies (n = 4) were related to cardiovascular disease (CVD), and 3 of them performed meta-analyses (59, 62, 76). All meta-analyses showed that DHIs are promising methods for managing CVDs.

Fourteen systematic reviews evaluated 3 lifestyle domains and all of them included diet and PA. Nine evaluated SU, 3 evaluated stress management, and 1 evaluated sleep. Only five conducted meta-analyses. These meta-analyses concluded that there is insufficient evidence on the effectiveness of DHI for reducing overall CVD (78). DHI significantly improved physical activity, HbA1c levels, body weight, empowerment, QoL. DHIs were effective, particularly in the short-term, to decrease the number of drinks consumed/week, DHIs may improve healthy behaviors (i.e., PA and healthy diet) but did not appear to reduce unhealthy behaviors (i.e., smoking, alcohol, and unhealthy diet) (39).

The remaining 52 systematic reviews evaluated only one or two domains. Fifty-one evaluated either diet, PA, or their combination (24 presented meta-analyses) and one evaluated DHI for smoking cessation.

To our knowledge, this is the first scoping review mapping the most frequent lifestyle domains and outcomes addressed by DHIs. Up to September 2021, we observed that only one systematic review simultaneously evaluated the six lifestyle domains proposed by the American College of Lifestyle Medicine (12). The majority of the reviews targeted diet and physical activity. Objective measures, such as weight and BMI, were the most frequently observed outcomes, followed by self-reported questionnaires. DHIs were delivered using different technologies, with smartphone apps and websites being the most common.

Lifestyle is a developing research theme, which has exponentially attracted interest during the COVID-19 pandemic. Researchers around the world are evaluating the delayed effect of unhealthy behaviors during the pandemic over the population health (i.e., increased incidence of chronic diseases and mental health problems after the changes in lifestyle). Overall, studies evaluated only one or a couple of behaviors that may have changed during this period (106, 107). Such finding is convergent with our results, showing that most reviews evaluated a single behavior/domain or a combination of diet and PA. This approach is valuable for estimating prevalence and specific targets for public policies, as well as increasing our understanding of each specific domain. However, at the individual level, it disregards that a healthy lifestyle represents a complex balance among multiple behaviors and that the development of chronic diseases results from the combination of different risk factors (i.e., genetic, lifestyle, and environmental) (108). LM, relying on six main domains, may be a step forward to address those behaviors, their complex interactions, and interconnections—especially if a multidimensional approach is applied. From a research perspective, the main priority should be to clarify the concept of lifestyle, and how to operationalize it. In this sense, guidelines and reporting guidelines are still missing and could be of great value to inform further studies.

From the 72 systematic reviews included in our study, 58 addressed diet/nutrition, and 60 evaluated PA. Diet and PA are fundamental features of a healthy lifestyle and well-known risk factors for chronic diseases (1, 11, 109). Given this importance, it is not a surprise that the most frequent domains and outcomes evaluated in the literature were related to them. Additional factors that may explain the frequency of DHIs addressing diet/PA include their importance for clinical populations, and the availability of objective outcomes to be measured (such as weight, BMI, and biochemical measures). Still, the high heterogeneity of those outcomes, delivery methods, and control groups precluded most of the systematic reviews to perform meta-analyses on the effectiveness of the DHI. Beyond the consensus on those aspects, an upcoming challenge is if and how hard outcomes (mortality or stroke for example) will ever be potential targets for DHIs. Hard outcomes take a longer follow-up time to be observed, while DHIs are fast-changing and volatile. It is possible we will need “adaptive” interventions, with real-time evaluation and modifications guided by short-term outcomes. Although methodological and ethical issues will be raised, creating new short-lived interventions will hardly be cost-effective and/or sustainable in the long run.

Regarding SU, LM literature concentrates on the use of tobacco and alcohol which, at the population level, present the highest impact on health. Overall, 7.7 million deaths were attributable to smoking in 2019, with smoking being the major cause of death among men in the world (110). Since 2015, it is recommended that clinicians screen all adults for cigarette smoking and prescribe some behavioral and pharmacotherapy therapy for smokers (unless there is some medical contraindication) (111). In 2016, there were 2.4 billion current drinkers globally, and 2.8 million deaths were attributed to alcohol (112). The burden of disease related to alcohol use may be seen as a result of binge drinking and alcohol use disorders, which usually require different treatment/prevention approaches. Given such figures, it is remarkable that we found only sixteen reviews addressing DHI to decrease SU (mostly to reduce smoking). It is possible that researchers do not recognize the importance of these substances or do not have sufficient training to provide evidence-based interventions. In fact, alcohol use disorders are among the least diagnosed medical conditions in clinical practice (113), and alcohol remains the blind spot of global health (114). In addition, as in the case of diet and PA, most reviews evaluating SU were conducted among clinical or at-risk populations, which may represent the most severe and least responsive individuals. Nevertheless, a recent review of systematic reviews found that DHIs have a small but positive effect to decrease alcohol and tobacco use among the general population (115). Such results are encouraging in the sense that DHI, together with public policies, may represent a step forward in the primary prevention of more than 200 diseases caused by smoking and/or alcohol use. One important aspect to evolve in this endeavor is to rigorously evaluate the severity of the SU disorder to determine who is going to benefit and which components of the DHIs are effective.

Stress management and social relationships were among the least evaluated lifestyle domains in our study. Compared to diet and PA, social relationships and stress management present less evidence of the effect on hard outcomes such as mortality. Furthermore, it is difficult to find reliable objective measures/outcomes to be used in DHI. Innovative digital methods to address this gap are necessary and these areas could benefit from passive sensing data obtained via smartphones and wearable devices, in the same way, step count is used to evaluating PA. Digital phenotyping through these tools allows quantifying the biological stress response in real time (116). In experimental conditions, algorithms developed based on physiological proxies of the autonomic nervous system activity (e.g., heart rate, body temperature, skin temperature, and conductance) have demonstrated a high accuracy to detect stress (117). Moreover, smartphone-based episodic audio recordings allow analyses of voice and speech features that may be potential vocal markers of stress (118, 119). Phone usage and the number of calls, texts, and interactions in online social media can be used as proxies of social contact or interaction (116, 120–122), while global positioning system (GPS)-based mobility and conversations captured by smartphone microphones were used to estimate social activity and loneliness in individuals with schizophrenia (123). Social support, however, is more complex than social contacts. To also account for the quality of social relationships, calls to close contacts with strong ties (such as family members and close friends) have been evaluated (124). Moreover, preliminary evidence suggests that passive audio collection on smartphones combined with machine learning techniques helped to identify auditory stimuli from the social environment of adolescents and young mothers that can signal social support (125).

In the same manner, sleep was evaluated in only three reviews. Sleep problems are a growing concern in global public health due to the severe consequences of poor sleep on cognition, emotion, the risk for serious medical conditions, and mortality (126). Particularly during the COVID-19 pandemic, sleep problems have been highly prevalent with ~40% of the general population reporting poor sleep quality (127). Sleep comprises of different areas, such as sleep onset latency, sleep restfulness, sleep disturbances, sleep schedule, efficiency, and daytime napping, among others (126, 128). Traditional, non-digital interventions to improve sleep quality are well-established and have been used to target specific problems, such as cognitive-behavioral therapy (CBT), relaxation and mindfulness therapy, stimulus control therapy, sleep hygiene education, and similar cognitive-behavioral interventions aiming to improve these areas (129, 130). DHIs for sleep improvements, such as digital CBT and digital sleep restriction therapy, have demonstrated potential as effective e-health interventions, however, further refinement and investigations of these delivery methods are required (131, 132).

Digital Health Interventions have emerged as promising methods to help patients and people at risk by providing immediate access to suggestions on improvement of nutritional and PA habits, education about their conditions, and medication adherence. The most frequent delivery method we found was the use of a smartphone, which has increased over the last decade following the widespread of this technology. The progressive advancement of efficient, accessible, and reliable technology has made it possible for various populations to have quick access to information and health. This progress has led to the exponential increase of the number of apps addressing lifestyle (22, 133), and the COVID-19 pandemic will likely lead to a further increase in those numbers. Beyond issues regarding costs (a considerable proportion of apps is paid, although it is hard to estimate the exact proportion) and accessibility (in populations with low cell phone coverage), adherence, usability, and long-term engagement remain to be improved. Co-design and user experience design are important ways to overcome these barriers, as well as behavioral sciences and health economics (134). As mentioned before, the proper scientific evaluation of DHI effectiveness is still to be determined but important steps, such as the publication of the NICE framework (27) and the WHO recommendations (21) have advanced the field. Importantly, while RCTs are the gold standard to evaluate DHIs, the CONSORT EHEALTH extension should be widely adopted to improve their quality, transparency, and reproducibility (135). In the same direction, systematic reviews and meta-analyses should also evaluate how closely are those guidelines being followed by authors. Another aspect that has not been adequately assessed in the systematic reviews is the possibility of harm, including confidentiality breaches. Although most people may believe DHIs present low-level harm, it is still necessary to report on that, and even bring to light harms that were not considered beforehand.

Our study has some limitations, such as the absence of data (e.g., the sample size of included RCTs) for some systematic reviews, and the complexity of study characteristics: different populations (general population and people at risk or living with specific diseases), interventions, delivery methods, and comparators. To properly answer our research question, the inclusion criteria were broad, including as many systematic reviews as possible, which limited synthesis and comparisons between studies. However, scoping reviews usually aim to provide an overview/map of the evidence instead of the result/answer to a particular question (136). Nevertheless, our review highlighted the following major gaps to be addressed regarding DHIs to improve lifestyle: (1) the concept of lifestyle is unclear and fragmented, including mostly one or two unhealthy behaviors; (2) lifestyle domains for which there are few objective outcomes to be measured (such as SU, social relationships, and stress management) are understudied; (3) there is a lack of assessment of hard/long-term outcomes; (4) although DHIs are valued by their high accessibility, most studies are not designed to include the general population; and (5) there is high heterogeneity on reporting the methods, such as outcomes, description of the interventions, and control groups.

The COVID-19 pandemic has shown the importance of connections and real-time data monitoring in health (137) and has consolidated the use of technology (such as tele-medicine) across the globe. DHIs will be increasingly important to improve lifestyle at primary, secondary, and tertiary prevention. However, to be cost-effective and sustainable, such interventions will need to be constantly monitored, adapted, and fully integrated into health systems. Integrating technologies and data are crucial to inform precision medicine, allowing algorithms to suggest the best interventions for each individual. LM and Public Health may profit from those advances to prevent chronic diseases.

The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author/s.

RC and RD conceived the study. RC and CR searched the literature. RC, HP, CO, CR, MS, and RD reviewed the literature and extracted data. MR-A analyzed data. RC, MS, and RD wrote the first draft. VB-M and FK reviewed for important intellectual content. All authors contributed to the article and approved the submitted version.

RD would like to thank the Conselho Nacional de Desenvolvimento Científico e Tecnologico (CNPq #312543/2020-4) and Fundação de Amparo ‘a Pesquisa do Estado do Rio de Janeiro (FAPERJ E-26/203.154/2017).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.