In this study, a total of 2998 records were retrieved from 10 major English-language databases. After deduplication and initial title and abstract screening, 273 articles were selected for full-text review. Based on the predefined exclusion criteria, exclusions were made for the following reasons: nonuniversity samples (n=68); interventions not primarily digital-based (n=20); wearable devices only, without active intervention components (n=5); absence of behavioral interventions (n=83); lack of relevant behavioral outcomes (n=35); and protocol or abstract only (n=19). Additionally, 3 more articles were identified through manual reference tracing of relevant review papers. Ultimately, 46 publications met the inclusion criteria and were included in the final analysis, as depicted in Figure 1.

Following 2 rounds of quality assessment, the first-round MMAT evaluation indicated that the 46 included studies demonstrated an overall high level of methodological quality. Specifically, 28 (61%) studies were rated as high quality, 14 (30%) as moderate quality, and 4 (9%) as low quality (see Figure 2; also see [10,11,13-15,17,20,26,29-66]). Major methodological concerns identified during the assessment were primarily concentrated in MMAT items C4 and C5. Item C4 was primarily related to the implementation of blinding procedures, the adequacy of outcome interpretation, and the control of risk of bias, whereas item C5 reflected issues such as insufficient intervention adherence and the lack of rigorous statistical analyses. In the second round of assessment, the RoB 2 tool was applied to evaluate 30 QRCTs, indicating that the primary sources of bias were related to outcome measurement, deviations from intended interventions, and the handling of missing outcome data (see Figure 3; see also [13,14,17,29-31,33-35,37,38,42-48,50-52,54,56,58,59,61-65]). Concurrently, the JBI critical appraisal of the remaining 16 studies indicated that key factors influencing study quality primarily included sample representativeness, intervention adherence, and the objectivity of outcome measurement (see Figure 4; see also [10,11,15,20,26,32,36, 39-41,49,53,55,57,60,66]).

The emergence of these methodological issues can be primarily attributed to 2 factors. On the one hand, the behavioral nature of DHIs makes the implementation of blinding inherently challenging, and several key behavioral outcomes rely on participant self-report measures. On the other hand, relatively high dropout rates associated with DHIs contribute to issues such as low intervention adherence and elevated loss-to-follow-up rates in some studies. When combined with insufficiently rigorous statistical analyses, these challenges may result in suboptimal handling of missing data or deviations from intended interventions. Although studies rated as moderate to low quality constitute a notable proportion of the included literature, it is important to recognize that many of their methodological limitations are closely related to the inherent characteristics of DHIs. Moreover, many of these studies primarily aimed to explore the feasibility and applicability of DHIs rather than to provide definitive evidence of intervention efficacy. Therefore, these studies retain substantial value for informing future research and intervention development. Given these considerations, no studies were excluded from this review solely based on methodological quality. Instead, all eligible studies were included, and findings from risk-of-bias assessments were systematically incorporated into the narrative synthesis. This approach allows for a comprehensive presentation of the current evidence landscape while explicitly identifying both its strengths and limitations.

This review included a total of 46 studies. The basic characteristics of the included studies are summarized in Table 1, with detailed data extraction results provided in Multimedia Appendix 3. Given the substantial heterogeneity among the included studies with respect to study design, target behaviors, intervention formats, core functions, and primary outcome measures, as well as variations in methodological quality, a meta-analysis was not conducted. Instead, a comprehensive analysis was performed using descriptive synthesis and comparative approaches. By systematically organizing and describing key characteristics of DHIs—including intervention targets, participant characteristics, sample sizes, formats, functions, durations, outcomes, and effects—this review delineates the overall patterns and heterogeneity within the field. Specific details are elaborated in the subsequent sections and illustrated through relevant tables, charts, and figures.

In terms of annual distribution (see Figure 5), the number of studies during the early period (2014-2015) was low, with only 1 publication per year. Since 2016, the number of publications increased markedly, reaching a first minor peak in 2016 (n=8), possibly associated with the rapid adoption of smartphones and mobile apps among college students. From 2017 to 2020, the number of studies fluctuated between 1 and 5 annually, maintaining an overall moderate level. The number increased again and stabilized in 2021-2022, declined slightly in 2023, reached a second peak in 2024 (n=8), and remained high in 2025 (n=4). Publications from the last 5 years accounted for more than half of all studies identified.

The regional and country distribution of the included studies demonstrates a clear geographical concentration. At the regional level, most studies were conducted in North America (n=18, 39%), followed by Oceania (n=10, 22%) and Europe (n=9, 20%). Asia accounted for 6 (13%) studies, while Africa contributed the smallest share with 3 (7%) studies. At the country level, the United States recorded the highest number of publications (n=15, 33%), followed by Australia (n=9, 20%). The United Kingdom, Germany, Canada, South Korea, and India each contributed 2 studies. The remaining countries were represented by a single study, indicating a relatively dispersed distribution beyond the leading contributors.

The distribution of study design types among the included studies exhibited a clear structural pattern. The largest proportion comprised QRCTs (n=30, 65%). This was followed by QNRSs (n=10) and QDSs (n=4), which were primarily used for exploratory analyses and descriptive accounts of phenomena. By contrast, QR and MMS were represented by only 1 article each, accounting for less than 2% of the total. Overall, DHI studies addressing college students’ lifestyle behaviors are predominantly quantitative, with a marked preference for QRCTs.

With respect to ethical compliance, all included studies adhered to relevant ethical guidelines, with all 46 (100%) explicitly reporting informed consent procedures and ethics committee approval or review status. Regarding privacy protection and data security, 24 (52%) studies explicitly reported the implementation of protective measures, including secure server storage compliant with data safety standards, encrypted data transmission, data deidentification, and strict access control mechanisms. With respect to adverse events and intervention-related risks, no serious adverse events were reported across the included studies. Only a small number of studies reported minor negative issues related to technology use, such as fluctuations in intervention engagement, higher dropout rates, or reduced compliance attributable to participants’ competing academic or personal commitments. No health risks were identified that were directly attributable to the DHIs.

Among the intervention objectives examined in the included studies, 30 addressed physical activity, 26 addressed diet, 10 targeted sedentary behavior, and 6 targeted sleep. Single-behavior interventions accounted for a large proportion of the studies; however, multibehavior crossover interventions were also substantial, with combined physical activity and diet interventions being the most common (n=18). Notably, physical activity was both the most frequent single-behavior intervention target and the primary entry point for multibehavior combined interventions, whereas sleep was relatively underemphasized in intervention design.

Based on the PROGRESS-Plus (Place of Residence, Race/Ethnicity, Occupation, Gender/Sex, Religion, Education, Socioeconomic Status, Plus Other Relevant Factors) framework, a synthesis of sociodemographic characteristics from 46 DHI studies identified 10 primary participant categories (see Multimedia Appendix 3), including health status (n=46), age (n=45), gender/sex (n=45), education (n=41), occupation (n=39), place of residence (n=36), race/ethnicity (n=28), socioeconomic status (n=14), social capital (n=8), and religion (n=1). The analysis revealed the following: (1) All participants were college students, predominantly aged 18-30 years, which is consistent with typical college student demographics and showed no substantial deviation across studies. (2) Most interventions targeted students with generally healthy status, whereas 14 out of 46 (30%) focused on subpopulations with specific health risks or special needs, such as overweight or obesity, sleep disorders, psychological stress, or disordered eating behaviors. (3) Gender/sex distribution was relatively balanced across studies, whereas education and occupation exhibited limited variability owing to the homogeneity of the study population. (4) By contrast, PROGRESS-Plus dimensions such as race/ethnicity, socioeconomic status, social capital, and religion received notably limited attention, with a lack of systematic analysis from a health equity perspective.

The sample sizes of the included studies varied considerably. Histograms indicated that most studies had sample sizes concentrated below 200 participants, with a median of approximately 95, whereas a few studies had small (<50) or extremely large (>400) samples. As shown in Figure 6, box-and-whisker plots further revealed an uneven distribution with long-tailed characteristics. Variations in sample size were closely associated with study design. Rigorous QRCTs typically require larger samples to ensure statistical power and therefore tend to employ medium- to large-scale sample sizes. By contrast, QDSs and QR are more inclined toward small-sample explorations, sometimes recruiting only a few dozen participants, and are more susceptible to selection bias.

The intervention formats in the included studies fell into 3 main categories. The first category, single, referred to interventions employing only 1 digital health technology (n=29), such as mobile apps. The second category, multiple, involved combining multiple digital health technologies within the same intervention (n=10). For example, the TXT2BFiT program integrated phone calls, websites, apps, and SMS text messaging simultaneously to achieve intervention goals. The third category, combined (n=7), compared the effectiveness of different combinations of digital health technologies, such as a “web-based nutrition intervention only” versus a “web-based intervention combined with daily SMS text message reminders.” Regarding the types of intervention technologies, these could be categorized into 7 groups: (1) mobile apps, used 21 times; (2) web-based platforms, including websites (13 times), online forums (3 times), and digital learning or eHealth tools (4 times); (3) mobile communications, including SMS text messages (11 times), emails (5 times), and phone calls (3 times); (4) social media (7 times); (5) wearable devices (4 times); (6) gamification and multimedia, including gamification and exergames (5 times), videos (2 times), and virtual reality (1 time); and (7) intelligent technologies, represented only by artificial intelligence (1 time). Overall, mobile apps and web-based platforms were the most frequently used technologies.

The technological functions of the DHIs included in this review exhibited distinct patterns of emphasis. Educational and guidance-related functions predominated across most interventions, followed by monitoring and prompting functions; by contrast, feedback and interactive functions were used less frequently, while immersive, screening, and engagement-related functions were rarely incorporated. Coding these interventions using the Behavior Change Technique Taxonomy version 1 (BCTTv1) indicated that the most frequently employed techniques were “4.1 Instruction on how to perform the behavior” and “5.1 Information about health consequences,” suggesting that current DHIs primarily emphasize foundational behavioral support functions. Further frequency analysis of BCT coding among effective intervention studies (see Table 2) showed that BCTTv1 codes 4.1 (16/87, 18%), 5.1 (14/87, 16%), and 2.3 (13/87, 15%) constituted the core set of techniques, collectively accounting for nearly half of all techniques used in effective interventions.

The duration of interventions varied considerably across the included studies (see Figure 7; see also [10,11,13-15,17,20,26,29-66]), with the majority concentrated in the short- to medium-term range (1-16 weeks). Studies involving long-term interventions (>16 weeks) were relatively scarce, with only 4 studies identified. Among these studies, most incorporated follow-up periods, and medium- to long-term interventions were typically associated with more systematic follow-up protocols. With respect to study design, randomized controlled trials predominantly employed interventions of medium duration (8-16 weeks). Among the QDSs (n=4) and MMS (n=1) analyzed, some studies employed longer intervention durations to observe behavioral maintenance; however, these accounted for a relatively small proportion of the evidence base. Subgroup analysis demonstrated a progressive increase in the proportion of studies classified as “effective” with increasing intervention duration (see Table 3): 2 out of 4 (50.0%) for ultra-short-term (<1 week), 10 out of 16 (63%) for short-term (>1 and <8 weeks), 12 out of 18 (67%) for medium-term (8-16 weeks), and 3 out of 4 (75%) for long-term (>16 weeks). Notably, medium-duration interventions (8-16 weeks) not only represented the largest proportion of the existing evidence but also demonstrated both a relatively high “effective” rate (12/18, 67%) and a low “ineffective” rate (1/18, 6%). These findings indicate that current DHI research remains skewed toward short- and medium-term interventions, with the 8-16-week category standing out in terms of evidence volume and the apparent stability of intervention effects.

As a result of substantial heterogeneity among the included studies with respect to outcome measurement instruments, outcome definitions, and assessment time points, it was not feasible to define a unified primary outcome or to conduct a statistically valid meta-analysis. Accordingly, this review adopted a descriptive synthesis framework to summarize and integrate the relevant outcomes. The outcome metrics in the included studies were classified into 2 main categories. The primary outcomes focused on lifestyle behaviors, including physical activity (eg, activity level, step count, and activity intensity), sedentary behaviors (eg, total sedentary time and frequency of breaks from sitting or resting), diet (eg, dietary quality; intake of fruits, vegetables, and sugar-sweetened beverages; energy intake; and nutritional knowledge), and sleep (eg, sleep quality, duration, efficiency, and severity of insomnia). These indicators directly reflect changes in core health behaviors resulting from the intervention and serve as a key basis for evaluating its effectiveness. Secondary outcomes, serving as supplementary indicators, were more diverse and encompassed physical health status and psychosocial dimensions, such as weight and body composition (eg, weight, BMI, waist circumference, and body fat percentage), physical fitness indicators (eg, flexibility, muscle strength, and cardiorespiratory fitness), cardiometabolic indicators (eg, blood pressure, blood glucose, and blood lipid profiles), and psychological and self-perception measures (eg, self-efficacy, body image, and life satisfaction). Overall, current studies remain primarily focused on primary outcomes, while secondary outcomes have expanded but continue to exhibit limited coverage.

Based on the reported effect measure types, effect estimates, confidence levels (%), and CIs across the included studies, together with a comprehensive assessment of the authors’ conclusions (see Multimedia Appendix 3), the results indicated that 31 (67%) studies demonstrated evidence of intervention effectiveness, suggesting that DHIs are generally associated with positive outcomes in improving lifestyle behaviors among college students. Four studies reported no statistically significant effects, with limitations primarily attributed to small sample sizes or short intervention durations. The remaining 11 studies demonstrated limited effectiveness, with improvements observed only in selected secondary outcomes or during short-term follow-up periods.

Based on a comprehensive assessment of each behavioral domain using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) framework, the certainty of evidence for the physical activity and diet domains was rated as “moderate,” whereas the evidence for the sedentary behavior and sleep domains was rated as “low.” With respect to evidence credibility, this review indicates a moderate level of confidence in the overall estimate that DHIs are effective in improving lifestyle behaviors among college students. The certainty of evidence in some domains was downgraded due to methodological limitations in the existing primary studies, including small sample sizes, challenges in implementing blinding, and inconsistencies in outcome assessment tools. Nevertheless, these GRADE assessments provide an accurate reflection of the current state of the evidence and its overall strength for DHIs among college students, thereby offering valuable guidance for interpretation and future research.

In terms of temporal trends, research on DHIs targeting college students’ lifestyle behaviors has gradually emerged since 2014, expanded rapidly after 2016, and reached a peak in the past 5 years [6]. This trend has been driven primarily by 4 categories of factors. First, technological advances have laid a solid foundation for DHIs, with the proliferation of smartphones, wearable devices, and app ecosystems significantly enhancing their accessibility and operability [67]. Second, conceptual advancements have accelerated theoretical and methodological innovations in DHIs, underscoring their distinctive advantages in facilitating behavioral improvement [68]. Third, demand has increased substantially, particularly during the COVID-19 pandemic, with DHIs gaining broad recognition as viable alternatives when traditional approaches were constrained [14]. Fourth, resource investment has continued to expand, with funding, supportive policy frameworks, and interdisciplinary collaboration creating a favorable environment for research. Overall, future research is expected to shift from assessing short-term feasibility to evaluating long-term effectiveness, scalability, and the capacity to accommodate personalization [29,30].

In terms of spatial distribution, research on DHIs is predominantly concentrated in high-income countries, particularly in the United States and Australia. This concentration is primarily driven by a combination of technological infrastructure, research resources, and supportive policy environments. On the one hand, North America and Oceania initiated mHealth development relatively early, benefiting from substantial technological and financial advantages [31]. On the other hand, colleges in these regions generally possess mature health promotion systems and well-established ethical review mechanisms, facilitating the implementation of intervention trials. In addition, higher levels of health awareness and greater digital acceptance in Western cultures further contribute to this pattern. However, the generalizability of these findings may be limited when extrapolated to low- and middle-income countries. For example, resource-constrained settings may encounter infrastructural and hardware-related barriers, such as uneven network coverage and low rates of digital device ownership. Furthermore, substantial cross-cultural variations exist in perceptions of privacy, the role of family involvement, and prevailing health communication practices. In the future, cross-cultural validation and localized adaptation of DHIs should be strengthened [69,70], particularly in resource-constrained settings. Moreover, the development of low-cost, low-threshold DHI models should be explored to advance global health equity [11,71].

In terms of population structure, current research on DHIs has predominantly focused on generally healthy college students, a focus attributable to this group’s modifiable health behaviors and susceptibility to environmental influences. However, some studies have extended to special populations, including college students with overweight or obesity, individuals at risk for eating disorders [15,32], and students experiencing sleep disorders or psychological stress [33,34]. This differentiation strategy is partly motivated by the fact that special populations face higher health risks, thereby increasing the potential benefits and clinical value of interventions [72,73]. It also aligns with the need for precision interventions and stratified management. However, existing research has not yet sufficiently examined variations in engagement levels among students from diverse sociodemographic backgrounds. Limited attention to factors such as socioeconomic status and access to digital devices may result in disproportionate benefits for students with greater financial or digital resources, while those experiencing economic constraints or limited device access may be marginalized in the intervention process. Accordingly, future research is likely to advance along 2 complementary directions: first, continuing large-scale studies targeting general undergraduate populations to assess the generalizability of interventions; and second, strengthening targeted interventions for high-risk groups while prioritizing the reduction of participation barriers among students from diverse backgrounds [74]. Such efforts may drive the development of DHIs toward greater refinement, equity, and personalization.

In terms of intervention objectives, physical activity and diet are the 2 lifestyle behavior categories receiving the most research attention [35], whereas sedentary behavior and sleep are relatively underrepresented. Both single-behavior and multibehavior combined interventions coexist. This pattern is primarily influenced by several factors. First, physical activity and diet are directly associated with weight management, energy balance, and metabolic health—core variables that affect college students’ physical fitness and chronic disease risk [36]. Related measures (eg, step count, energy intake) are more easily quantifiable and standardized, making them more likely targets for intervention. By contrast, sedentary behavior and sleep, despite their recognized importance [37], pose technical and operational challenges for DHIs, including measurement complexity and delayed feedback on intervention effects [38], contributing to a relative paucity of research [39,40]. Current DHIs demonstrate limited effectiveness in reducing sedentary time among college students [75], whereas sleep interventions, although promising, remain understudied and predominantly focus on insomnia relief [18,76]. Second, intervention strategies reflect researchers’ assessment of behavioral variability: physical activity and diet exhibit a wide window for controllability and modification, whereas sedentary behavior typically occurs in academic or leisure contexts, complicating immediate adjustment via a single technique. The prevalence of single-behavior interventions is attributed to their suitability for early exploratory phases, allowing easier control of variables and validation of intervention effects. Conversely, the increase in multibehavior interventions reflects the aggregation of lifestyle risks among college students, which complicates achieving sufficient health benefits through changes in a single behavior. Notably, the combination of physical activity and diet is the most frequent, reflecting the necessity for integrated interventions targeting weight management and energy metabolism [41]. Overall, future research is expected to increasingly adopt multibehavior approaches, integrating behavioral science theories and technological tools to develop synergistic interventions that address the complexity of lifestyle risks.

In terms of intervention modalities, an evolutionary trend is evident, progressing from single to multiple formats and from low to high levels of interaction, driven by the combined forces of technological advancement, user demand, and intervention science. In early studies, SMS text messages and emails were the predominant forms of DHIs [42], owing to their low technological threshold, ease of deployment, and minimal cost, which made them suitable for rapid implementation in resource-limited contexts [43,44]. However, these approaches primarily involved 1-way information delivery, lacked personalization and real-time interaction, and were insufficient in maintaining user engagement. With the widespread adoption of smartphones and the maturation of the app ecosystem, mobile apps have gradually become the mainstream form of DHIs. These apps are highly integrated and interactive, capable of incorporating multiple functions such as goal-setting, feedback, reminders, and data tracking [13], aligning with college students’ high-frequency mobile usage habits and significantly enhancing the intervention experience and engagement [45]. Web platforms retain advantages in scalability but are somewhat less user-friendly and less effective in delivering push notifications compared with apps [46,47]. In addition, the integration of social media and wearable devices increases the interactivity and contextual adaptability of DHIs [48], further enhancing behavioral monitoring and the provision of immediate feedback [49-51]. Future trends are expected to emphasize technological convergence and intelligent development. On the one hand, combinations of multiple formats (eg, apps, social platforms, and gamification) will become increasingly prevalent to address the multidimensional needs of behavioral interventions [52]. On the other hand, personalized interventions leveraging artificial intelligence, virtual coaching, and immersive experiences (eg, augmented reality/virtual reality) are anticipated to emerge as key research directions [17,53], shifting DHIs from being information-driven to experience-driven and ultimately facilitating sustained behavior change.

In terms of intervention functions, current DHIs are predominantly characterized by education, guidance, monitoring, and prompting components [54], indicating that these interventions primarily emphasize information delivery and basic behavior management. This design approach is partly driven by the substantial demand for health knowledge and skills among college students, with education and guidance functions facilitating improvements in cognition and self-efficacy. Concurrently, monitoring and prompting functions leverage technology to enable data recording and behavioral reinforcement, thereby promoting the initiation of target behaviors in the short term [55]. However, high-engagement features such as feedback, interaction, and gamification-based incentives remain underutilized [56,57], suggesting that DHIs often lack deep personalization and social support components [58], which may be a critical factor limiting long-term user engagement and intervention effectiveness. In terms of future trends, the convergence of behavior change theories (eg, Capability-Opportunity-Motivation-Behavior [COM-B], behavioral economics) with intelligent algorithmic applications is expected to drive the evolution of DHI functionality toward greater personalization, interactivity, and emotional engagement [59]. For example, artificial intelligence–driven real-time feedback could enhance intervention adaptability, virtual communities could strengthen social support, and gamification mechanisms coupled with reward systems could foster intrinsic motivation. Such advancements are likely to not only increase intervention engagement but also substantially improve behavioral maintenance, fostering a gradual shift from information delivery–oriented DHIs to approaches that place greater emphasis on user experience and social interaction.

Overall, 31 of 46 (67%) studies reported effective outcomes (yes), indicating the high feasibility and considerable potential of DHIs in improving the lifestyle behaviors of college students [60-62,77]. However, a subset of studies yielded insignificant (no) or limited (limited) effects, which can be examined from several analytical dimensions. First, insufficient refinement and lack of theoretical underpinning in intervention design represent key factors constraining effectiveness. In several cases, interventions lacked explicit theoretical frameworks for behavior change, relying predominantly on information delivery. Such approaches often failed to sufficiently stimulate participant motivation or reinforce behavior maintenance, leading to short-term gains that were difficult to sustain [26]. Second, existing intervention studies generally lack robust validation of long-term effects. Most studies are limited to durations of 8-16 weeks and include insufficient follow-up, which constrains the ability to verify the sustainability and stability of behavioral changes [63]. As a result, the long-term value and durability of DHIs remain difficult to assess adequately. Third, intervention effectiveness appears to be strongly influenced by participant adherence. Analyses of engagement-related metrics indicate that higher levels of user engagement, compliance, and intervention consistency are generally associated with more favorable behavioral and clinical outcomes. By contrast, studies characterized by high dropout rates often rely predominantly on 1-way information delivery, with limited opportunities for feedback and interaction. Fourth, the type of target behavior and associated measurement challenges also contribute to these outcomes. Compared with physical activity, the intervention effects on diet, sedentary behavior, and sleep were more vulnerable to external environmental influences (eg, academic workload, dietary contexts), and measurement tools relied predominantly on self-reporting, thereby increasing bias and uncertainty. Taken together, variations in intervention design, technological application, and behavioral characteristics collectively contribute to the substantial heterogeneity observed in intervention outcomes [64].

To gain a deeper understanding of variations in intervention effectiveness, the COM-B framework can be applied as a systematic analytical tool [78]. (1) Within the “Capability” dimension, most interventions primarily enhanced college students’ health-related knowledge through educational content and guidance materials. Examples included the provision of diet guidelines, exercise plans, and sleep regulation strategies designed to increase participants’ awareness of the importance of healthy behaviors. However, these improvements often remained at the cognitive level, with limited emphasis on the development of practical behavioral skills. Specific components, such as diet substitution options, situational coping strategies, or flexible exercise planning, were frequently absent. In addition, some studies did not provide adequate support for data interpretation, which limited participants’ ability to translate behavioral monitoring data into actionable steps [79]. (2) Within the “Opportunity” dimension, DHIs generally rely on virtual platforms to create enabling behavioral conditions, such as goal tracking, reminder functions, and online resource sharing, which may theoretically reduce psychological barriers to behavior enactment. However, the structuring of opportunities within real-world contexts remains insufficiently optimized. Some interventions do not adequately account for the distinctive time pressures and contextual constraints experienced by college students on campus. For example, strategies aimed at reducing sedentary behavior often remain limited to generic standing reminders, without adaptation to classroom environments or common study spaces, thereby constraining opportunities for sustained behavior change. Furthermore, although some interventions attempt to incorporate social support mechanisms (eg, community interactions or peer challenges), the depth and quality of participant engagement are generally limited. These interactions frequently involve 1-way information transmission, with limited capacity to foster emotional connection or effective behavioral modeling. (3) Within the “Motivation” dimension, existing interventions primarily emphasize the stimulation of extrinsic motivation through short-term incentives, such as point-based rewards and task completion reminders. While such strategies may promote initial engagement, they generally lack mechanisms for the sustained cultivation of intrinsic motivation. Specifically, many interventions have not effectively supported college students in developing a sense of self-worth derived from continued engagement in healthy behaviors. In addition, strategies aimed at enhancing positive emotional experiences are rarely incorporated. For example, gamification designs often remain confined to superficial point-based systems, with limited capacity to stimulate participants’ sense of exploration, mastery, or accomplishment. Additionally, insufficient personalization of feedback appears to substantially constrain the maintenance of motivation over time. Participants often receive generic informational messages rather than timely, individualized feedback closely aligned with their actual behavioral performance.

In summary, current DHIs predominantly adopt a “technology-driven” or “utility-oriented” design logic, with a primary emphasis on functional implementation and surface-level engagement metrics. Because of the limited integration of behavior change theory, such interventions tend to exhibit constrained effectiveness in sustaining long-term outcomes. By contrast, theory-driven interventions—such as those grounded in the COM-B framework—extend beyond short-term behavior initiation, emphasizing the synergistic development and dynamic support of capability, opportunity, and motivation. Through structured and phased behavioral support strategies, such interventions may facilitate the establishment of enduring foundations for sustained change across cognitive, skill-based, environmental, and emotional dimensions [10]. As a result, long-term behavior maintenance may become more attainable [65]. Future research should further position behavior change theory as a central guiding principle in intervention design, moving beyond the view of technology as a standalone tool and instead embedding it organically within support systems centered on behavior change mechanisms.

To fully realize the potential of DHIs while ensuring the sustainability and broad accessibility of intervention effects, systematic improvements in policy design and implementation pathways are required. First, college students should be explicitly incorporated into national and regional digital health strategies to facilitate a shift from traditional health education toward integrated digital platforms, and higher-education institutions should be encouraged to develop or adopt scientifically grounded, standardized tools with clearly articulated mechanisms of action. Second, localized development of intervention content and functionality should be supported, with attention to adaptability across behavioral domains, cultural contexts, and student needs, thereby advancing refined, human-centered design with respect to technological thresholds, data security, and personalized recommendations. Third, intervention practice should strengthen students’ active engagement and establish feedback-driven, behavior-reinforcing, and peer-support mechanisms to enhance sustained use and intrinsic motivation. In parallel, cross-departmental cooperation mechanisms should be established at the college level, and health interventions should be embedded within curricula, psychological support systems, and campus service resources to form a synergistic support network. Finally, at the policy level, ethical oversight and effectiveness evaluation of DHIs programs should be strengthened, and an evidence-based evaluation framework for DHIs should be established to ensure fairer, more adaptable, and more effective interventions for college students.

This study indicates that current research on DHIs for college students remains constrained by unrepresentative samples, single-focus intervention content, and unclear technological mechanisms; future work should be refined and deepened in the following respects. First, geographical and cultural diversity should be expanded, prioritizing studies from low- and middle-income countries, varied higher-education institution types, and diverse social groups to enhance the external validity of the findings. Second, the design and evaluation of multibehavior-integrated interventions should be strengthened by moving beyond single-behavior paradigms and examining behavioral synergies and optimal combinations of intervention components. Third, higher-quality study designs—such as QRCTs, MMSs, and long-term follow-up—should be employed to strengthen causal inference and the sustainability of intervention effects. Fourth, theoretical development and empirical testing of intervention mechanisms should be strengthened by grounding analyses in behavior change theory to clarify how technology enhances Capability, Opportunity, and Motivation, and to advance DHIs from merely providing technical functions to creating a supportive ecosystem conducive to sustained behavior change. Finally, future studies should emphasize the assessment of intervention equity, systematically account for potential moderators such as gender, socioeconomic status, and psychological status, and identify subgroups with limited responsiveness, thereby providing a robust evidence base for constructing a more inclusive and adaptive DHI model for college students.