A comprehensive systematic review and meta-analysis examining the deployment of smartphone applications in occupational settings has revealed that digital health tools can encourage positive behavioural adjustments among employed adults, although the statistical strength of these improvements depends heavily on experimental design. Published in BMC Digital Health, the investigation evaluated research published between 2008 and 2026 to assess how mobile phone software influences daily habits such as physical activity, sleep patterns, dietary routines, and psychological well-being in working populations. While mobile devices have become ubiquitous instruments for communication and personal management across diverse corporate environments, existing empirical evaluations have historically yielded mixed conclusions regarding their genuine effectiveness. By collating data from twenty peer-reviewed investigations that encompassed more than 39,000 participating workers across multiple continents, the research team sought to clarify whether occupational app interventions produce measurable, reproducible health benefits compared to traditional workplace guidance or unassisted baseline routines.
The methodological framework for the synthesis was preregistered on the PROSPERO international prospective register under registration identifier CRD42025646349 and executed in compliance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. The researchers conducted exhaustive literature queries across four electronic databases, specifically PubMed, SCOPUS, PsycINFO, and Web of Science, covering records from January 2008 to coincide with the initial launch of major commercial app marketplaces through an updated search conducted in January 2026. The initial and updated database searches retrieved a combined total of 818 citations, which yielded 565 non-duplicate records for preliminary screening. Following the elimination of reports that lacked occupational cohorts, omitted empirical behavioural indicators, or utilised non-app modalities, thirty-nine full-text articles underwent rigorous quality appraisal. Methodological quality and potential risk of bias were independently evaluated using the Standard Quality Assessment Criteria for Evaluating Primary Research Papers from a Variety of Fields, commonly designated as the QualSyst tool, applying an inclusion threshold of 0.5 that ultimately retained twenty qualifying empirical studies.
The demographic composition of the aggregated cohort was characterised by an overall mean participant age of 36.94 years with a standard deviation of 8.81 years, reflecting an active adult workforce across an array of professional sectors. Among the twenty studies, eleven were randomised controlled trials, including cluster-randomised, pilot, and mixed-methods trials, whereas the remaining nine studies incorporated non-randomised formats such as pre-post feasibility trials, proof-of-concept projects, and open-label single-arm interventions. The employment contexts represented in the sample spanned office-based administrative environments, healthcare institutions, manufacturing facilities, government agencies, universities, long-haul commercial trucking, shift-work operations, and military personnel. The collective gender distribution exhibited a pronounced imbalance, comprising 29,153 male participants representing 73.8 percent and 10,345 female participants representing 26.2 percent, alongside 67 individuals who were recorded as non-binary, preferred not to answer, or had missing demographic documentation. However, the authors noted that this aggregate gender skew was heavily influenced by a single large-scale commercial step challenge study that supplied over 30,000 predominantly male participants, whereas the remaining nineteen individual studies frequently featured balanced or female-majority cohorts reflecting specific sector demographics.
Quantitative synthesis focused primarily on physical activity, which served as the most frequently evaluated primary outcome across the literature base. The investigators conducted two distinct meta-analyses using Comprehensive Meta-Analysis Version 4 software, standardising differing outcome metrics such as weekly step counts, daily step totals, and moderate-to-vigorous physical activity into uniform daily activity durations using established conversion formulas. The first meta-analysis pooled five randomised controlled trials comprising 640 total participants, split evenly between 321 intervention subjects and 319 control subjects, to determine between-group differences following app usage. Under a random-effects model, the overall effect size across these controlled trials was small and fell just short of conventional statistical significance, yielding a standardised mean difference of 0.205 with a 95 percent confidence interval spanning from negative 0.002 to 0.412, a test statistic Z of 1.94, and a p-value of 0.052. Heterogeneity across these five controlled trials remained moderate and statistically non-significant, with an I-squared value of 32 percent and a Cochran’s Q statistic of 5.872 across four degrees of freedom with a p-value of 0.209.
Because the between-study heterogeneity in the controlled trial group did not reach statistical significance, the investigators additionally evaluated the five randomised studies under a fixed-effects statistical assumption, which presupposes that all examined interventions estimate a common underlying population effect size. Under this fixed-effects model, the pooled between-group effect size reached statistical significance, producing a small-to-medium standardised mean difference of 0.259 with a 95 percent confidence interval from 0.102 to 0.415, a Z statistic of 3.245, and a p-value of 0.001. In contrast, the second meta-analysis synthesised seven studies that utilised within-participant, pre-post experimental designs without separate control arms, tracking physical activity shifts within the same cohort from baseline to post-intervention follow-up. This within-group analysis encompassed 36,775 total workers and produced a statistically significant pooled effect size of 1.506 under a random-effects model, with a 95 percent confidence interval ranging between 0.764 and 2.248, a Z score of 3.979, and a p-value below 0.001. However, this within-group comparison displayed extreme and statistically significant heterogeneity, marked by an I-squared value of 97 percent, a Q statistic of 201.235 across six degrees of freedom, and a p-value below 0.001.
The authors stressed that the apparent magnitude of the within-group effect must be interpreted with substantial methodological caution rather than taken as unreserved proof of digital app superiority. Crucially, a single multi-country employee wellness program contributed 36,562 individuals, constituting approximately 99.4 percent of the entire within-participant meta-analytic sample and thus disproportionately driving the pooled summary estimate. Uncontrolled pre-post designs inherently fail to adjust for extraneous confounding influences such as temporal seasonal changes, statistical regression toward the mean, or the Hawthorne effect, wherein participants alter their daily routines simply because their activity is being actively recorded. When mobile phone applications were compared directly against concurrent control groups receiving standard educational leaflets, delayed access, or unguided usual routines in formal randomised trials, between-group differences were substantially smaller and did not achieve statistical significance under conservative random-effects modeling. The researchers concluded that the available body of randomised controlled trials remains underpowered to definitively confirm the modest positive between-group trend indicated by the data.
Beyond general step counts and daily movement metrics, the systematic review synthesised narrative evidence concerning specific secondary health behaviours, including sedentary workplace patterns, sleep habits, dietary modifications, and psychological indicators. Five included studies evaluated sedentary sitting time, yielding divergent conclusions across workplace settings; while certain trials documented marked drops in occupational sitting and leisure computer usage at six-month and twelve-month milestones, other controlled trials identified reductions in sitting time that failed to achieve statistical significance when compared directly with active or education-only control cohorts. Interventions directed toward sleep hygiene, insomnia severity, and sleep consistency demonstrated more reliable advantages across several trials, particularly when the underlying applications integrated structured clinical frameworks such as cognitive behavioural therapy for insomnia. Conversely, dietary outcomes represented the least examined domain, investigated by only two studies that observed transient increases in daily vegetable and fruit servings or overall dietary quality indices that subsequently attenuated or returned to baseline thresholds during unassisted post-intervention follow-up assessments.
An analysis of intervention architecture revealed that eleven studies evaluated stand-alone smartphone applications, whereas nine implemented multi-component strategies that paired software tools with external elements such as wearable activity sensors, automated text messaging, printed guides, financial incentives, or instructional workplace seminars. Nine of the twenty applications were commercially available off-the-shelf platforms, including commercial step trackers and fitness utilities, while eleven were custom research-specific applications developed expressly for trial execution. Explicit grounding in behavioural change theory was reported in only seven of the twenty evaluated programs, incorporating paradigms such as Self-Determination Theory, Social Cognitive Theory, the Transtheoretical Model, the Habit Formation Model, and the Model of Arousal Dynamics. Across all platforms, the most frequently integrated Behaviour Change Techniques consisted of self-monitoring of activity, goal specification, and automated feedback on individual performance, which were near-universally deployed across both stand-alone and multi-component formats.
The review identified notable challenges regarding the long-term durability of employee behavioural improvements once active trial periods concluded or external motivators were discontinued. Multiple trials observed that initial spikes in physical activity or daily step counts diminished significantly toward the conclusion of extended observation windows or dissipated entirely following the withdrawal of material incentive programs and redeemable points schemes. For instance, interventions relying on gamification and external reward tokens experienced participant activity declines once extrinsic reinforcers were removed, underscoring the difficulty of cultivating self-sustaining, intrinsically motivated health habits through digital prompts alone. Furthermore, participant retention rates varied dramatically across the twenty studies, ranging from high completion rates exceeding 90 percent in four closely monitored investigations down to attrition rates exceeding 50 percent in three trials, culminating in one large remote study where only 20.4 percent of enrolled workers completed follow-up assessments at the second evaluation point.
In evaluating the broader strengths and limitations of the evidence base, the authors highlighted the rigor of their preregistered search across multiple indexing databases and the execution of parallel between-group and within-group syntheses, while acknowledging that structural constraints within the published literature restrict definitive clinical conclusions. The existing evidence is constrained by a scarcity of adequately powered randomised trials, extensive reliance on unvalidated subjective self-report questionnaires, missing demographic variables regarding employment status, and substantial heterogeneity across intervention durations, which spanned from brief four-week pilot tests to fifteen-month longitudinal evaluations. The review concludes that while mobile health applications represent accessible, cost-effective instruments capable of supporting workplace wellness initiatives and encouraging daily movement, their role in driving long-term lifestyle transformation remains promising but not yet definitively established through rigorous comparative trial evidence. Establishing robust proof will require future investigators to conduct large-scale, methodologically consistent randomised controlled trials that combine objective sensor-based tracking with extended follow-up windows and theoretically sound behavioural techniques.
Subject of Research: Systematic review and meta-analysis of workplace mobile health applications for employee behavioural change
Article Title: Mobile apps to promote behavioural change in occupational settings: a systematic review and meta-analysis
Article References: Baldwin, T., Cadar, D., & Sigala, N. (2026). Mobile apps to promote behavioural change in occupational settings: a systematic review and meta-analysis. BMC Digital Health, 4(1), Article 29. https://doi.org/10.1186/s44247-026-00268-7
Image Credits: AI Generated
DOI: 10.1186/s44247-026-00268-7
Keywords: mobile health applications, workplace wellness, occupational health, physical activity, behavioural change techniques, systematic review, meta-analysis, randomised controlled trials, sedentary behaviour, employee health promotion, self-monitoring, goal setting, insomnia interventions, sleep quality, habit formation, digital health interventions, occupational medicine