Mental disorders are among the leading causes of global disease burden [1] and a significant risk factor for premature mortality [2]. Recent estimates suggest that between 5% and 19% of global disability-adjusted life-years can be attributed to mental disorders [1,3], causing annual economic losses of approximately US $4.7 trillion [4]. These mental health difficulties cause enormous psychological impact, have widespread deleterious effects on health behaviors [5], and vastly increase the odds of experiencing physical health morbidities [6].

There is increasing recognition that poor lifestyle behaviors, such as diet, physical activity, and sleep, contribute to both poor physical and mental health. Concurrently, there is increasing evidence for the role of lifestyle interventions in preventing, managing, and treating mental illness [7-9]. Numerous clinical guidelines and health policy documents now call for lifestyle behaviors to be addressed alongside pharmacotherapies and psychotherapy, as part of mental health care [10].

Digital technology offers a promising avenue for revolutionizing mental health care delivery on a global scale. Digital interventions can also reduce the stigma associated with seeking help [11], while data-driven insights allow for personalized and efficient interventions tailored to individual needs.

In recent years, an increasing number of studies have demonstrated the effectiveness of digital interventions in delivering mental health treatments [12], including compelling outcomes for smartphone-delivered psychological interventions in reducing depression and anxiety symptoms [13,14]. Concurrently, research related to digital lifestyle interventions is increasing rapidly [15]. A growing body of evidence shows that these interventions can be effective for improving health behaviors such as physical activity [16], sleep quality [17], weight loss [18], and healthy diet [19]. However, there is little evidence specifically focused on the effect of digital lifestyle interventions on mental health and well-being outcomes in the general population. The few reviews in this area have focused on singular lifestyle behaviors such as physical activity [20], specific clinical populations [21], or a single delivery method [22] or have reported lifestyle behavior changes but not changes in mental health [19,23]. Moreover, only 1 review has synthesized data from >10 randomized controlled trials (RCTs) [19]. This fragmented approach limits our ability to rigorously discern the effectiveness of the various intervention components and combinations on mental health across the general population.

This systematic review and meta-analysis takes a broad approach, aiming to provide a thorough synthesis of the evidence on digitally delivered lifestyle interventions for depression, anxiety, stress, and well-being outcomes in adults. Recognizing that mental health is more than the absence of disorders and distress [24], we also evaluated the effectiveness of digital lifestyle interventions on psychological well-being (ie, well-being), as operationalized by van Agteren et al [25]. Here, well-being refers to positive aspects of mental health, including positive affect and life satisfaction (ie, subjective well-being), as well as meaning, purpose, and related concepts (ie, psychological well-being).

Given the breadth of the review, the secondary aim was to explore the technology-specific intervention features that are associated with greater mental health outcomes. We further considered the impact of various methodological (ie, study quality), intervention-specific (ie, type, delivery method, and delivery features), and population-specific (ie, general population, those with a mental illness, and those with a physical illness) characteristics. Finally, this review aimed to shed light on the overall quality of the evidence provided in the meta-analyses and discuss the implications of the evidence for future research and lifestyle intervention delivery.

The following electronic databases were searched from January 2013 to January 10, 2023: MEDLINE (Ovid), CINAHL (EBSCOhost), Embase (Ovid), Emcare (Ovid), PsycINFO (Ovid), and Scopus. Subject heading, keyword, and Medical Subject Headings term searches relating to technology, modifiable lifestyle behaviors (ie, physical activity, diet, and sleep), mental health, and study design were used. Full search queries are detailed in Multimedia Appendix 2. Additional searches were conducted using Google Scholar and by scanning the reference lists of included papers and review articles to identify any additional studies. Studies were exported to Covidence (Veritas Health Innovation Ltd), and duplicates were removed. Studies were screened in duplicate with conflicts resolved between the 2 reviewers or a third independent reviewer.

To be included in this systematic review and meta-analysis, studies needed to meet the following criteria: (1) involve adults (aged ≥18 years) of any health status and (2) include a digitally delivered lifestyle intervention targeting physical activity, diet, sleep, or any combination thereof. We defined digital lifestyle interventions as interventions delivered via digital technologies (such as apps, websites, and wearable devices) that were self-guided and did not require real-time clinician delivery (eg, behavior change and education) and aimed to improve lifestyle behaviors [27]. Considering the nature of the research, we included interventions that were supplemented with contact for follow-up or adherence and technology support (eg, orientation session or phone call to create a user profile or a peer-driven social forum moderated by a clinician). Lifestyle interventions were eligible if at least 50% of the intervention was focused on physical activity, diet, and sleep. Physical activity was considered in the broadest sense, including overall physical activity levels, structured exercise interventions, and studies examining reducing physical inactivity (ie, sedentary time). Diet was considered as any intervention targeting food intake (ie, amount or type). Studies focusing on supplementation or specific nutrient treatments were excluded. Sleep interventions were considered as anything targeting sleep; therefore, psychological therapies targeting sleep (eg, cognitive behavioral therapy for insomnia) were eligible for inclusion; (3) studies that had an adequate control condition of no treatment, treatment as usual, waitlist control, or attention (ie, sham) control were included, while control groups that received an alternative intervention comprising behavior change techniques or an in-person version of the digital intervention were excluded, as they would limit our ability to evaluate the effectiveness of the intervention; (4) studies that reported changes in symptoms of depression, anxiety, psychological stress, or well-being using validated tools were included, and due to the broad and varied definitions of well-being [28], we considered well-being measures listed as per the review by van Agteren et al [25]. Studies that focused on momentary affective states were considered outside the scope of this review and were excluded; (5) RCTs were included. No restriction was placed on the setting or context of the included studies.

In total, 2 reviewers independently extracted data using a predefined data extraction sheet, cross-checked the data, assessed study quality, and resolved disagreements by discussion or referral to a third reviewer (JB or CM). The authors of the included studies were contacted to provide additional data for inclusion in the meta-analysis if required.

Intervention effect sizes (ie, differences between intervention and control groups) for outcome data were calculated across all studies and standardized to Hedges g [29] along with the SE. Studies were grouped by outcomes (ie, depression, anxiety, stress, and well-being) for analyses. Random-effects meta-analyses were conducted due to expected heterogeneity. All analyses used an inverse variance method with restricted maximum-likelihood estimator for τ² and Hartung-Knapp adjustment for the random effects model. Standardized mean differences (SMDs) were used as the effect measure for meta-analyses to allow comparison of data from different scales. If means and SDs were not reported in a study, authors were contacted twice before means and SDs were calculated based on available data using recommended formulas (eg, using sample size, median, and range) [30]. Meta-analyses were conducted only when at least 5 studies were included. Effect sizes were categorized as small (0.2), medium (0.5), or large (≥0.8) [31], with a significance level set at .05 [32]. Cochrane Q test was used to assess statistical heterogeneity, and the I² statistic was used to quantify the proportion of the overall outcome effect attributed to heterogeneity. The following cutoff values for the I² statistic were used: 0%-29%=no heterogeneity, 30%-49%=moderate heterogeneity, 50%-74%=substantial heterogeneity, and 75%-100%=considerable heterogeneity [33]. Publication bias was assessed with visual inspection of funnel plots and tested using the Begg-Mazumdar [34] and Egger regression method [35], with a P value <.05 suggesting the presence of bias. Where significant bias was detected, a Duval and Tweedie [36] trim-and-fill analysis was applied. Heterogeneity was explored with subgroup and sensitivity analyses. Sensitivity analyses were conducted by removing outliers, studies with samples <100, studies of poor to fair quality, and those with attrition >25% to investigate changes in effect for each outcome. All analyses were conducted in R software (version 4.2.0, “Mountain Hydrangea”; R Foundation for Statistical Computing).

The risk of bias of each study was assessed using the Physiotherapy Evidence Database Scale (PEDro) [37]. The PEDro scale comprises 10 questions, each answered as yes or no. A study was deemed to be high quality with a score of 6 to 10, fair quality with a score of 4 to 5, and low quality with a score <4.

The overall level of evidence was graded using the Oxford Centre for Evidence-Based Medicine 2011 Levels of Evidence [38] as follows: grade A—consistent level 1 studies (ie, individual RCTs with narrow CIs), grade B—consistent level 2 (ie, individual cohort studies or low-quality RCTs) or level 3 studies (ie, individual case-control studies) or extrapolations from level 1 studies, grade C—level 4 studies or extrapolations from level 2 or 3 studies, or grade D—level 5 (ie, expert opinion without explicit critical appraisal) evidence or inconsistent or inconclusive studies of any level.

The central aim of this review was to examine the effect of various methodological, intervention-specific, and population-specific factors on the effectiveness of digital lifestyle interventions for mental health outcomes [39]. Intervention-specific factors included the lifestyle behavior; the intervention targeted (ie, physical activity alone, diet alone, sleep alone, herein referred to as the respective singular behavior, or any combination of these behaviors, which is referred to herein as multibehavioral); delivery method (ie, web based, app, email or text, and wearable device); features of delivery format (ie, use of a chatbot or gamification); frequency and duration of the intervention; whether the intervention was co-designed, individualized, or publicly available; and whether it contained a theoretical underpinning. The frequency of intervention delivery was split into three categories: (1) intended daily use; (2) 1 to 4 times per week; and (3) “other,” which included fortnightly, monthly, varied, once-off, and self-paced engagement. Methodological factors were control group types and study quality. Study quality was based on the risk of bias, where studies were categorized into poor, fair, good, or excellent as per the PEDro scoring. Control groups were divided into no treatment, including waitlist; treatment as usual; or attention control. To determine which populations might benefit most from digital lifestyle interventions, we conducted subgroup analyses for each outcome for apparently healthy adults; those with a physical health condition; those diagnosed with a mental health condition; and, due to the number of sleep-specific studies, people who met the criteria for insomnia.

We originally planned to include interventions targeting alcohol, smoking, and substance use as behaviors, considering their association with poor mental health outcomes. However, due to the unfeasible volume of search results and their distinct focus on addiction, these lifestyle behaviors were excluded from this review.

Searching of databases yielded 14,357 results. Following the removal of duplicates, 9727 potentially eligible studies remained for which abstracts were screened. At full-text stage, 230 studies were reviewed, and 169 were removed because they failed to meet the inclusion criteria (Multimedia Appendix 3). The remaining 61 studies were included in the review, and 59 studies were included in the quantitative synthesis (Figure 1).

In total, 22,483 participants across 61 studies were included in the review. Sample sizes ranged between 20 and 3755 participants, and the mean age ranged between 19 and 68 years. Interventions lasted between 2 and 52 weeks. Studies originated from Europe, America, Canada, Australia, New Zealand, the United Kingdom, Iran, and Asia. Full details of each study are outlined in Table 1.

A total of 55 studies assessed depressive symptoms [40-42,44-58,60-83,85-90,92-95,97,98,100], 35 assessed anxiety symptoms [45,46,50-52,57,58,60,61,63,65-68,72-74, 76-83,86,88,89,92,93,95,97-100], 11 assessed stress [44,46,57,59,65,78,80,82,83,93,96], and 6 assessed well-being [43,46,61,84,91,96]. Overall, 24 studies included apparently healthy adults [44,46,48,50,51,53,55,57,59,64,70, 76,80,82-85,89-93,99]; 23 included adults with a physical health condition [43,45,47,52,54,56,63,65,67-69,71,74,77-79,86, 88,94,95,98,100]; 6 included adults with a mental health condition, all of which were depressive disorders [40,41,62,73,81,97]; and a further 8 included adults with insomnia [49,58,60,61,66,72,75,87]. Interventions were digitally delivered via the web (38/61, 62%) [40,44,45,48-50,52-56,58-64,67,68,73-79,82-84,86,88-90,92-94,98], app (17/61, 28%) [42,43,46,47,57,60,65,66,69,72,80,81, 85,91,95,97,100], email, or SMS text messages (15/61, 25%) [44,46,51,58,61,63,70,77,79-81,86,89,99,100], and 16% (10/61) of the interventions included a wearable activity tracker [41,42,57,63,71,80,91,94,96,98]. A total of 15 interventions used >1 delivery method [42,44,57,58,61,63,77,79-81,86,89,91,94,98]. Of the 61 interventions, 26 (43%) targeted physical activity [41,42,45,48,56,57,59,63,67,68,70,71,73,74,81,82,84,85, 88,91-94,96,100], 17 (29%) targeted sleep [49,50,55,58,60-62,64,66,72,75-77,86,87,89,90], 2 (3%) targeted diet [51,69], and 16 (27%) targeted multiple lifestyle behaviors [42,44,57,58,61,63,79-81,86,89,91,94,98]. Of the 17 studies targeting multiple behaviors, 4 (24%) targeted physical activity, diet, and sleep [40,47,97,98]; 11 (65%) targeted physical activity and diet [44,46,52-54,65,78,79,83,95,99]; and 2 (12%) targeted physical activity and sleep [43,80].

Mental health (ie, depression, anxiety, stress, and well-being) outcomes were measured as a primary outcome in 14 studies [41,46-51,62,83,85,89,93,97,98]. The risk of bias of the included studies is detailed in Multimedia Appendix 4.

For 2 of the included studies, the outcome data of interest was not presented in the manuscript and was unavailable after contacting the corresponding author [70,87].

Pooled effects from 53 studies (ie, 12,385 participants) showed a small-to-medium significant effect of digital lifestyle interventions for reducing depressive symptoms in comparison to all control conditions (SMD=−0.37; 95% CI −0.46 to −0.27; P<.001; Figure 2).

Figure 3 displays the pooled effect size from 35 studies (9383/22,483, 41.73%), showing a small-to-medium positive effect of digital lifestyle interventions compared to all control conditions on symptoms of anxiety (SMD=−0.29; 95% CI −0.36 to −0.21; P<.001).

Pooled effects from 11 studies (n=1608) showed a small positive effect of digital lifestyle interventions compared to all control conditions on stress (SMD=−0.17; 95% CI −0.33 to −0.01; P=.04; Figure 4).

Pooled effects from 6 studies (4204/22,483, 18.7%) suggest no significant effect of digital lifestyle interventions compared to all control conditions on well-being (SMD=0.14; 95% CI −0.07 to 0.37; P=.15; Figure 5), although the data show a nonsignificant (P=.15) trend in the direction that favors digital lifestyle interventions.

Sensitivity analyses for all meta-analyses are detailed in Multimedia Appendix 7. For depression, when small studies were omitted from the analysis, the effect size shrunk but remained significant (SMD=−0.28; 95% CI −0.38 to −0.19; P<.001). For anxiety, the effect size was slightly larger when studies with >25% attrition were removed (SMD=−0.31; 95% CI −0.43 to −0.18; P<.001). For stress, omitting small studies or studies with >25% attrition no longer demonstrated a significant effect (P=.13 and P=.22, respectively). The overall effect on well-being remained nonsignificant when fair to poor quality studies, small studies, and studies with >25% attrition were removed (all P values >.10).

Overall, the level of evidence for digital lifestyle interventions for improving symptoms of depression, anxiety, and stress is grade A: consistent level 1 studies. The grade of recommendation for digital lifestyle interventions for improving symptoms of well-being was grade D: level 5 evidence.

This systematic review aimed to comprehensively examine the effectiveness of digital lifestyle interventions for improving symptoms of depression, anxiety, stress, and well-being in adults. We identified 61 RCTs involving 22,483 participants. The findings suggest that digital lifestyle interventions had small-to-medium favorable effects on depression, anxiety, and stress symptoms across a broad range of populations and via a broad range of digital delivery methods. These effects are similar in magnitude to those observed for established mental health treatments, such as antidepressant medication [101] and psychotherapy [102]. No significant effect was observed for well-being, although in comparison to mental ill health, fewer studies (n=6) of lower average quality measured well-being outcomes. These studies operationalized well-being broadly using either global measures that span various affective and functional dimensions [43,61,84,91] or scales of overall life satisfaction [46,96]. These are broad constructs that tend to be relatively stable [103]. As a result, the available evidence offers limited precision for elucidating the effects of digital lifestyle interventions on specific well-being domains. Two-thirds of the included studies (41/61, 67%) were published within the last 5 years, reflecting the increased interest in using digital technology for health behavior change for mental health, with greater reductions in depressive and anxiety symptoms observed for more recently published interventions. This likely reflects the volume of studies published recently but may also be driven by improved quality of digital interventions as technology, particularly artificial intelligence, is better used within health promotion [104].

To better understand what factors drive the effects of digital lifestyle interventions on depressive and anxiety symptoms, we conducted a range of comparative subgroup analyses on specific features relating to the intervention, methodology, technology, and population. These analyses found that interventions demonstrated comparable effectiveness for reducing symptoms of depression and anxiety in apparently healthy adults or adults with a mental disorder, a physical health condition, or insomnia. Equivalent effectiveness was found regardless of the lifestyle behaviors targeted, method of digital delivery (eg, app and website), technological features (ie, chatbot and gamification), intervention duration, or intended frequency of engagement. Overall, our subgroup analyses indicated that there are no specific features that drive the effectiveness of digital lifestyle interventions for improving symptoms of depression or anxiety, except for publication year.

Although differences among population groups were not significant, it appears that digital lifestyle interventions are effective for improving depressive symptoms in people with a depressive disorder but are potentially less effective at improving symptoms of anxiety in people with a depressive disorder. Interestingly, depressive disorders were the only diagnosed mental disorder in the included studies, highlighting a clear gap. Therefore, the effectiveness of these interventions as a self-management approach for other mental disorders is unclear and should be explored in future studies.

We also observed that interventions using chatbots trended toward being more effective than those without chatbots, though these results were not statistically significant. These findings have been supported with recent research,. Emerging evidence suggests chatbots can help improve lifestyle behaviors [105], although it appears their effectiveness for delivering psychological interventions (ie, therapy) is less conclusive [106]. Co-designed interventions showed a trend for being less effective. This is likely due to the small number of studies included in the subgroup analysis and the high variation of co-design approaches used in the design of health interventions, which are often poorly described and rarely evaluated [107].

A key challenge of digital health interventions is maintaining user engagement over time [13]. It is possible that the lack of significant intervention effects at 4 months was due to drop off in engagement. Alternatively, it is possible that symptoms may recur in some individuals even if they remain engaged with the intervention, reflecting the recurring nature of mental health symptoms. Digital lifestyle interventions may offer an appropriate short-term strategy to engage people in mental health–promoting behaviors and initiate positive shifts in mental health, whereas more sustainable mental health benefits may require other mental health treatment approaches. These findings reinforce the existing literature that the clinical implications of lifestyle interventions, even when delivered digitally, are compelling as a self-management tool for targeting both the risk and protective factors for mental health and should be given consideration as part of standard mental health care. This review did not directly compare digital to face-to-face delivery of these interventions; therefore, the impact that delivery method has on effectiveness should be examined directly in future research. Given the role of patient-practitioner rapport and other psychosocial benefits of in-person care, it is pivotal to understand how digital interventions may best fit and integrate within the broader health care setting, including alongside gold standard treatments, to optimize patient outcomes.

In line with previous research [108], we found interventions targeting multiple or single behaviors to have comparable effectiveness. Interventions that aim to change multiple health behaviors concurrently seem logical, given the cooccurrence of and bidirectional relationship between unhealthy lifestyle behaviors and poor mental health. In particular, some health behaviors may have synergistic impacts (eg, people who improve their sleep may have more energy for physical activity and, in turn, people who are more physically active may sleep better); therefore, future research should explore how behavior change techniques can be implemented most effectively to address multiple behaviors and aim to identify moderators of effective interventions and whether delivering lifestyle behavior change within or alongside traditional psychological interventions has synergistic effects on outcomes or improves adherence.

This review also focused on adults, although adolescence is a period when mental health issues commonly present and offers an opportune time for preventive intervention. Digital interventions would seem to be appealing and accessible for this population, given their familiarity with technology. A recent review by Raeside et al [109] included 11 interventions and found small nonsignificant effects for digital lifestyle interventions on mental health outcomes in adolescents. Considering the shared protective and risk factors for mental health and chronic diseases, future research efforts should focus on if and how digital tools can be best harnessed to address the interplay between lifestyle behaviors, mental ill health, and well-being and engage young people in the design and development of interventions through participatory research methods, such as co-design.

This is the largest review to evaluate the effects of digital interventions targeting lifestyle behaviors on depression, anxiety, stress, and well-being in adults. We adhered to rigorous methodological approaches for the conduct and reporting of systematic reviews and meta-analyses and conducted comprehensive database searches. In total, 87% (53/61) of the included RCTs received good or excellent quality ratings, increasing confidence in the findings of this review. Given the very large number of studies included in the meta-analyses, the dataset was sufficient to support numerous subgroup analyses, facilitating closer interrogation of the factors influencing intervention effectiveness. Our review included digital lifestyle interventions that have been evaluated using an RCT design and published in a peer-reviewed academic journal. Many commercial digital products claim to support lifestyle changes and mental health but often lack rigorous scientific assessment. The conclusions from this study should not be generalized to such unevaluated products.

Most of the study’s limitations arose from the limitations of the included studies. First, only 2 studies reported on diet-only interventions, and few studies measured stress and well-being, resulting in uncertain and underpowered results for these meta-analyses and precluding subgroup analyses with stress and well-being as outcomes. Second, all but 2 studies [41,99] were conducted in high-income countries, limiting the generalizability of our results. Third, it is important to acknowledge that our subgroup analyses, such as the comparison between single- and multibehavioral interventions, were conducted using available data from studies not specifically designed to test these differences. Consequently, these findings should be interpreted cautiously and should not be used as definitive evidence for or against any particular intervention approach. Finally, we acknowledge significant heterogeneity as a limitation, and we conducted numerous subgroup and sensitivity analyses to mitigate its impact and assess the robustness of our findings, which remained consistent with the main analyses.

This review highlights some key areas for future research. First, there were limited studies focused on addressing diet as a stand-alone intervention; however, a nonsignificant effect in the direction of lifestyle interventions for reducing depressive symptoms highlights that further research is needed to better determine if, and how, effective these interventions may be in addressing mental health symptoms. Second, digital interventions may be less effective over longer periods, emphasizing a need to address longer-term strategies or the potential to integrate other mental health treatment approaches, which may demonstrate greater effects over time. Research also needs to address how best to integrate these evidence-based interventions into service delivery to support health care capacity. Third, for a more complete understanding of how digital lifestyle interventions influence mental health, future research should ensure that positive aspects of mental well-being are measured alongside symptoms of mental ill health. A growing body of literature suggests that these are distinct continua, with both overlapping and unique antecedents, that should be measured in tandem [110]. Finally, as the field continues to mature and research elucidates optimal intervention designs, other factors that are imperative to consider are the influence of both health literacy and digital literacy as well as the impact of health status and socioeconomic status [111]. Furthermore, future research should consider the digital determinants of health and the direct and indirect impacts on health equity when designing health promotion interventions [112].

Overall, the small-to-medium effects indicate that digital lifestyle interventions may provide an effective short-term self-management strategy for mental health support by encouraging positive shifts in lifestyle behaviors. All digital delivery platforms and features demonstrated comparable effectiveness for depression and anxiety outcomes. The effectiveness on well-being is inconclusive, with few studies capturing this positive dimension of mental health. Digitally delivered lifestyle interventions support self-management and target both risk and protective factors for mental health and should be given consideration as part of mental health care and support. This is especially relevant in situations where access to in-person treatment is limited, such as due to long waitlists, financial or geographical barriers, or stigma associated with traditional treatment options.

Future research should explore how best to implement these findings and integrate such interventions with existing health care services, with concerted efforts to improve health equity by considering the digital determinants of health and impacts of digital and health literacy in the design and implementation of these interventions.