The Associations between Multi-Theory Model (MTM) Constructs and Health Behavior Change among College Students: A Systematic Review and Meta-Analysis

College students, as a core group of young adults with high educational potential, play a pivotal role in driving global social progress and sustainable development¹. They are in a critical period of physical and psychological development, and face challenges such as academic pressure, social adaptation, and independent living. Consequently, their health is increasingly influenced by lifestyle habits and campus environments². For example, heavy study loads, irregular routines, a lack of exercise, and poor self-management often lead to both mental and physical health issues³. A growing body of evidence underscores the significant mental health challenges faced by college students. A large-scale meta-analysis revealed pooled prevalence rates of 34% for depression and 32% for anxiety among this demographic⁴. These rates are substantially higher than those in the general population, underscoring the unique pressures of this life stage, a pattern that is also evident in East Asia⁵. This period of transition often disrupts healthy habits, making the cultivation of positive health behaviors during this time crucial for shaping lifelong well-being.

The WHO notes that healthy behaviors increase health awareness, self-management, and quality of life⁶. Health behaviors are actions that are associated with physical and mental well-being while individuals adapt to the environment⁷. They play a vital role in the development of healthy lifestyles. Effective theoretical models are necessary to understand the factors associated with improvements and the potential facilitation of replacing unhealthy habits with healthy habits in both the short and long term.

Among the various health behavior intervention models, the Multi-Theory Model (MTM) of health behavior change, first proposed by Professor Manoj Sharma in 2015⁸, is prominent. Compared with other behavior change theories, such as the Transtheoretical Model (TTM), the Theory of Planned Behavior (TPB), and Social Cognitive Theory (SCT), the MTM has unique advantages, particularly in addressing the specific needs of college students. While the TTM effectively outlines stages of change, it offers limited actionable strategies for sustaining new behaviors, a particular challenge for students navigating fluctuating academic and social demands⁹. The TPB emphasizes behavioral intention as a key predictor but does not explicitly differentiate between the initiation and maintenance of behavior, a distinction that is critical in the dynamic context of student life¹⁰. SCT, despite highlighting self-efficacy and observational learning, provides a less structured framework for designing stage-specific interventions¹¹. In contrast, the MTM not only bifurcates the change process into initiation and maintenance but also introduces specific, measurable constructs for each stage, rendering it particularly suitable for developing targeted and sustainable health behavior interventions in college settings¹². This theory divides health behavior change into two stages: initiation and maintenance, thereby addressing the complex and diverse needs of health behavior interventions¹³.

The development of the MTM has progressed through four stages. Early theories focused on knowledge dissemination and awareness-raising to change behavior through information provision¹⁴. Second-generation theories incorporated skill training and consciousness improvement¹⁵. Third-generation theories emerged in the 1990s, introducing evidence-based techniques and emphasizing empirically grounded frameworks such as Social Cognitive Theory¹¹. In the fourth and latest stage, the MTM integrates core elements from multiple classical theories to comprehensively analyze and predict the initiation and maintenance of health behaviors¹⁶. The MTM combines key components from various theories to better predict and explain how individuals initiate and sustain healthy behaviors.

The MTM categorizes behavior change into two stages: initiation and maintenance. Initiation is influenced by participatory dialog (PD), behavioral confidence (BC), and changes in the physical environment (CPE). Maintenance is affected by emotional transformation (ET), practice for change (PC), and changes in the social environment (CSE)¹⁷. The MTM has been successfully applied in areas such as diet management¹⁸,handwashing behavior¹⁹, smoking cessation²⁰, alcohol abstinence¹², and gambling cessation²¹. However, most existing studies have focused on qualitative descriptions of its applicability or single-factor correlation analysis^17,22,23 and lack a quantitative synthesis of the strength of the association of each MTM construct with health behaviors among college students.

Given the unique challenges faced by college students, exploring the role of the MTM in their health behaviors is necessary. This study begins with the two stages, initiation and maintenance, and focuses on the six key constructs of the model. It aims to quantify the strength of the cross-sectional associations between each MTM construct and health behavior change among college students. By quantitatively integrating the data, this study provides comparable evidence on the strength of the associations of various MTM constructs and their cross-sectional relationships with college students' health behaviors. By synthesizing correlational data, this study aims to generate hypotheses about which constructs are most critical for each stage and thus should be tested in future longitudinal and experimental studies. This study addresses the limitations of previous systematic reviews, which remained largely at the qualitative description level and lacked a quantitative synthesis of evidence regarding the cross-sectional associations between MTM constructs and college students' health behaviors. Additionally, it fills a gap in quantitative research on the MTM, specifically in the college student population, by offering a more precise theoretical basis for health interventions aimed at this group. Furthermore, this study identifies stage-specific intervention priorities through quantitative analysis, providing direct preliminary evidence for the design of phased health behavior intervention strategies for college students.

This systematic review and meta-analysis protocol was registered on PROSPERO with a registration number of CRD 420251012795. This study was designed and conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA)²⁴ guidelines (Supplementary File 1).

1. Literature search strategy

The relevant literature was retrieved by searching PubMed, Embase, Web of Science, Scopus, Wanfang, and Chinese National Knowledge Infrastructure (CNKI) databases for studies published from the inception of each database to October 2024. The Chinese search terms used were "duo li lun mo xing" AND "da xue sheng" AND "jian kang xing wei". The English search terms used were (("Multi-Theory Model") OR ("MTM")) AND (("college student*") OR ("university student*") OR ("undergraduate*") OR ("academician")) AND (("health behavior*") OR ("healthy behavior*") OR ("health-related behavior*")).

2. Literature inclusion and exclusion criteria

The inclusion criteria were as follows: (1) the study subjects were college students; (2) the Multi-Theory Model (MTM) was applied as a predictive framework for health behaviors; (3) sufficient data were provided for calculating association measures (ORs); (4) the study design was a cross-sectional study; and (5) The included articles were limited to English and Chinese. The exclusion criteria were as follows: (1) duplicate publications; (2) studies with incomplete or inaccessible data; (3) studies not utilizing the Multi-Theory Model (MTM); and (4) reviews, case reports, conference abstracts, animal studies, etc.

3. Literature screening and data extraction

The identified records were imported into EndNote X21 software for literature review and duplicate removal. The study selection process was performed independently by two reviewers (X.L.H. and K.X.H.). First, they screened titles and abstracts against the eligibility criteria. Afterward, the full texts of potentially relevant articles were retrieved and assessed. Any discrepancies at each stage were resolved through consensus or, when necessary, by adjudication from a third senior reviewer (X.X.Z.). Data extraction was conducted independently by the same two reviewers (X.L.H. and K.X.H.) using a prepiloted, standardized data extraction form. The extracted data included the following: first author, publication year, country of study, sample size, demographic characteristics (e.g., sex and mean age), specific health behavior investigated, measurement tools for each MTM construct, and effect size data (e.g., odds ratios with 95% confidence intervals) for the association between each construct and behavior change.

4. Appraisal of methodological quality

The methodological quality of the included studies was assessed using a combined evaluation tool specifically designed for cross-sectional studies²⁵. The tool comprises seven criteria: (1) scientifically sound design; (2) reasonable data collection strategy; (3) reporting of sample response rate; (4) good representativeness of the sample relative to the population; (5) appropriate research objectives and methods; (6) reporting of statistical power; and (7) reasonable statistical methods. Responses were categorized as "yes", "no", or "unclear" and assigned scores of 1, 0, and 0.5, respectively. If an indicator was not applicable, it was denoted by "_" and scored as 1.0 points²⁶. The total score of the scale was 7.0 points. A score of 6.0-7.0 indicated Grade A quality, 4.0-5.5 indicated Grade B quality, and a score below 4.0 indicated Grade C quality. The quality assessment was performed independently by two reviewers (X.L.H. and X.Z.). Disagreements in scoring were discussed until a consensus was reached.

5. Statistical analysis

Data were analyzed using Review Manager 5.3 software. Heterogeneity among studies was assessed using P values and I² statistics. If P > 0.1 and I² < 50.0%, indicating acceptable heterogeneity among the study results, a fixed-effects model was used²⁷. If P ≤ 0.1 and I² ≥ 50.0%, indicating substantial heterogeneity, a random-effects model was applied²⁸. The odds ratio (OR) was used as the quantitative estimator throughout the study. Sensitivity analyses or subgroup analyses were conducted to explore the sources of heterogeneity when necessary. Statistical significance was set at P < 0.05. Publication bias was assessed both visually using funnel plots and statistically using Egger's linear regression test for outcome measures that included a sufficient number of studies (n ≥ 10). The primary data analysis, including the generation of forest and funnel plots, was performed using Review Manager (RevMan) version 5.3 software. Additionally, the statistical assessment for publication bias (Egger's test) was conducted using Stata version 18. Publication bias was assessed using funnel plots for outcome measures with a sufficient number of studies (n ≥ 10), with a significance level set at α = 0.05.

Literature search
A total of 90 articles were initially identified. After 43 duplicate records were removed using EndNote software combined with manual screening, 47 articles were retained. By reviewing the titles and abstracts, 19 articles that were clearly irrelevant or did not meet the study's subject and design criteria were excluded. After the full texts were read, 15 additional articles were excluded because of irrelevance to the research topic, incomplete data, or unavailability of full texts. Ultimately, 13 articles12,22,23,29,30,31,32,33,34,35,36,37,38 were included in the meta-analysis, all of which were in English. The literature screening process is presented in Figure 1.

The 13 cross-sectional studies12,22,23,29,30,31,32,33,34,35,36,37,38 included a total of 2,605 college students. The sample sizes varied across the studies. The study characteristics are presented in Table 1.

Methodological quality

Four articles22,23,30,32 were rated as Grade A in quality, and 9 articles12,29,31,33,34,35,36,37,38 were rated as Grade B in quality. No articles were rated as Grade C. The results of the methodological quality appraisal are shown in Table 2.

Cross-sectional associations between MTM constructs and health behavior change
The results of the meta-analysis indicated that participatory dialogue (PD), behavioral confidence (BC), changes in the physical environment (CPE), emotional transformation (ET), practice for change (PC), and changes in the social environment (CSE) were all positively cross-sectionally associated with healthy behavior change among college students (P < 0.05).

Cross-sectional associations of MTM constructs in the initiation stage
In the initiation stage, participatory dialogue (PD, OR = 1.03; 95% CI: 1.02, 1.05; P = 0.001), behavioral confidence (BC, OR = 1.12; 95% CI: 1.10, 1.14; P < 0.001), and changes in the physical environment (CPE, OR = 1.09; 95% CI: 1.07, 1.10; P < 0.001) demonstrated significant cross-sectional associations with health behavior change. Among them, behavioral confidence (BC) showed the highest level of statistical significance (Z = 11.75), and its odds ratio (OR = 1.12) indicated the strongest association with health behavior change. In practical terms, for every one-unit increase in behavioral confidence scores, the odds of initiating health behavior change increase by 12%, underscoring its potential as a primary target for interventions aimed at overcoming initial barriers to action30. Although heterogeneity was observed for BC (I² = 66.0%), significant cross-sectional associations were consistent across subgroups such as dietary behavior (P < 0.001) and physical activity (P < 0.001), indicating that behavioral confidence is closely associated with initial behavior change. The heterogeneity of changes in the physical environment (CPE) was low (I² = 41.0%), suggesting a relatively universal cross-sectional association (Table 3).

Participatory dialogue (PD)
Data from five studies29,32,33,35,38 were pooled. Owing to significant heterogeneity among the studies (I²=62.0%, P=0.030), a random-effects model was used. The analysis indicated that PD was associated with health behavior change (OR = 1.03; 95% CI: 1.02, 1.05; P < 0.001). The pooled estimates are displayed in Figure 2.

Subgroup analysis by behavior category revealed statistically significant cross-sectional associations for dietary behavior (OR=1.03; 95% CI: 1.01, 1.05; P = 0.003), physical activity (OR=1.04; 95% CI: 1.01, 1.08; P = 0.02), and preventive and protective measures (OR=1.03; 95% CI: 1.01, 1.06; P = 0.01), with no significant difference between the subgroups (P=0.83) (Figure 3).

Behavioral confidence (BC)
Data from all 13 studies12,22,23,29,30,31,32,33,34,35,36,37,38 were pooled. Given the significant heterogeneity among the studies (I²= 66.0%, P = 0.0004), a random-effects model was used. BC was significantly associated with health behavior change (OR = 1.12; 95% CI: 1.10, 1.14; P < 0.001) (Figure 4).

Subgroup analyses based on the dependent variable category revealed that studies on dietary behavior had moderate heterogeneity (I² = 45.0%, P = 0.14) and statistically significant differences (OR = 1.09; 95% CI: 1.07, 1.12; P < 0.001). Studies on physical activity had low heterogeneity (I² = 17.0%, P = 0.30) and statistically significant differences (OR = 1.10; 95% CI: 1.07, 1.12; P < 0.001). Studies on preventive and protective measures showed no heterogeneity (I² = 0.0%, P = 0.86) and had statistically significant differences (OR = 1.15; 95% CI: 1.13, 1.18; P < 0.001). Studies in other categories also showed no heterogeneity (I² = 0.0%, P = 0.46) and had statistically significant differences (OR = 1.16; 95% CI: 1.13, 1.19; P < 0.001). Comparisons between subgroups were statistically significant (P < 0.001) (Figure 5).

Changes in the physical environment (CPE)
The results of 10 studies12,22,29,30,31,32,34,35,37,38were pooled for analysis. Given the moderate heterogeneity among the studies (I²= 41.0%, P = 0.08), a fixed-effects model was employed for the meta-analysis. The results demonstrated that changes in the physical environment (CPE) according to the Multi-Theory Model (MTM) could be associated with health behavior change among college students (OR = 1.09; 95% CI: 1.07, 1.10; P < 0.001) (Figure 6).

Subgroup analyses based on the dependent variable categories revealed the following findings: studies on dietary behavior exhibited high heterogeneity (I²= 74.0%, P=0.02) and statistically significant differences (OR=1.11; 95% CI: 1.08, 1.14, P < 0.001); studies on physical activity had moderate heterogeneity (I²= 49.0%, P=0.16) and statistically significant differences (OR=1.08; 95% CI: 1.04, 1.12, P=0.0001); studies on preventive and protective measures demonstrated low heterogeneity (I²= 28.0%, P=0.25) and had statistically significant differences (OR=1.08; 95% CI: 1.05, 1.10, P< 0.001); and studies in other categories displayed no heterogeneity (I²= 0%, P=0.63) and had statistically significant differences (OR=1.09; 95% CI: 1.06, 1.12, P<0.001). The comparison between subgroups yielded a P value of 0.46 (Figure 7).

Cross-sectional associations of MTM constructs in the maintenance stage
In the maintenance stage of health behaviors, emotional transformation (ET, OR = 1.13; 95% CI: 1.09, 1.17; P < 0.001), practice for change (PC, OR = 1.15; 95% CI: 1.09, 1.21; P < 0.001), and changes in the social environment (CSE, OR = 1.06; 95% CI: 1.04, 1.08; P < 0.001) had crucial effects on long-term behavior adherence. Practice for change (PC) demonstrated the strongest association (OR = 1.15, Z = 5.21). This translates to a 15% increase in the odds of maintaining a health behavior for every one-unit increase in PC scores, suggesting that interventions fostering consistent practice and goal-setting could substantially improve long-term adherence12. The association was particularly pronounced for physical activity (OR = 1.22), indicating a 22% increase in odds, which highlights the critical role of structured practice plans in this domain23. Although the strength of the association for changes in the social environment (CSE) was more modest (OR = 1.06, implying a 6% increase in odds per unit score increase), its high statistical significance (P < 0.001) and negligible heterogeneity (I² = 0%) indicate a stable and consistent cross-sectional association with behavior maintenance across the studied populations22. This stability enhances its value as a reliable, albeit smaller, component of maintenance strategies (Table 4).

Emotional transformation (ET)
The results of ten studies22,23,29,30,31,32,33,36,37,38 were pooled. Owing to significant heterogeneity among the studies (I²= 75.0%, P<0.001), a random-effects model was employed for the meta-analysis. The results indicated that emotional transformation (ET) in the Multi-Theory Model (MTM) is associated with health behavior change among college students (OR = 1.13; 95% CI: 1.09, 1.17; P < 0.001) (Figure 8).

Subgroup analyses based on the dependent variable category revealed high heterogeneity in studies on dietary behavior (I²=87.0%, P=0.0004), with statistically significant differences (OR=1.15; 95% CI: 1.05, 1.25; P=0.002). Studies on physical activity exhibited low heterogeneity (I²=24.0%, P=0.25) and had statistically significant differences (OR=1.11; 95% CI: 1.06, 1.17; P<0.001). Studies on preventive and protective measures had low heterogeneity (I²=30.0%, P=0.23) and statistically significant differences (OR = 1.17; 95% CI: 1.13, 1.21; P<0.001). Studies in other categories showed no heterogeneity (I² = 0%, P = 0.60) but had statistically significant differences (OR=1.07; 95% CI: 1.04, 1.11; P = 0.0001). Comparisons across subgroups revealed a P value of 0.0001 (Figure 9).

Practice for Change (PC)
The results of eight studies12,22,29,30,31,34,37,38 were pooled. Owing to significant heterogeneity among the studies (I²=84.0%, P<0.001), a random-effects model was employed for the meta-analysis. The results demonstrated that practice for change (PC) in the Multi-Theory Model (MTM) was associated with health behavior change among college students (OR = 1.15; 95% CI 1.09, 1.21; P < 0.001) (Figure 10).

Subgroup analyses based on the dependent variable category revealed the following findings: studies on dietary behavior showed no heterogeneity (I²=0, P=0.63) and had statistically significant differences (OR=1.14; 95% CI 1.10, 1.18; P<0.001); studies on physical activity exhibited high heterogeneity (I²=80.0%, P=0.02) and had statistically significant differences (OR=1.22; 95% CI 1.09, 1.36; P=0.0005); studies on preventive and protective measures showed a statistically significant difference (OR=1.04; 95% CI 1.01, 1.07; P=0.01); and studies in other categories demonstrated no heterogeneity (I²=0, P=0.40) and had statistically significant differences (OR=1.15; 95% CI 1.10, 1.20; P<0.001). The subgroup comparison yielded a P value <0.001 (Figure 11).

Changes in the social environment (CSE)
The results of six studies22,23,29,31,33,38 were pooled. Since homogeneity was observed among the studies (I²=0, P=0.52), a random-effects model was used for the meta-analysis. The results indicated that changes in the social environment (CSE) in the Multi-Theory Model (MTM) were associated with health behavior change among college students (OR=1.06; 95% CI 1.04, 1.08; P<0.001), as shown in Figure 12.

The independent variables from the included literature were grouped according to their application domains, with subgroup analyses of behavioral confidence, emotional transformation, and practice for change showing statistically significant results (P < 0.05) (Table 5).

Sensitivity analysis
Through a step-by-step exclusion of the literature, a sensitivity analysis was conducted on studies with I²>50.0%; no significant changes in heterogeneity were found for behavioral confidence, emotional transformation, or practice for change. Five studies29,32,33,35,38 reported the effects of participatory dialogue on college students' health behaviors, with substantial heterogeneity among the studies (I² = 62.0%, P = 0.03). After excluding the study by Robert E. Davis et al.38, homogeneity was observed (I²=0, P=0.75), indicating that this study may be the source of heterogeneity in participatory dialogue. Ten studies12,22,29,30,31,32,34,35,37,38 reported the effects of changes in the physical environment on college students' health behaviors, with moderate heterogeneity among the studies (I²=41.0%, P=0.09). After excluding the study by Sharma et al.30, homogeneity was observed (I²=0, P=0.66), suggesting that this study may be the source of heterogeneity in changes in the physical environment (Table 6).

Publication bias analysis
Funnel plots were constructed for the independent variables with the greatest number of included studies (behavioral confidence, changes in the physical environment, and emotional transformation). With respect to behavioral confidence, most studies were located above the inverted funnel, with fewer studies at the base, and the distribution was roughly symmetrical on both sides, suggesting no obvious publication bias. Moreover, Egger's test revealed no significant publication bias (P=0.23) (Figure 13). With respect to changes in the physical environment, most studies were above the inverted funnel with an asymmetrical left-right distribution, indicating possible publication bias. Egger's test further confirmed the absence of significant publication bias (P=0.51) (Figure 14). With respect to emotional transformation, most studies were above the inverted funnel, with fewer studies at the base, and the distribution was roughly symmetrical on both sides, suggesting no obvious publication bias. Egger's test revealed no significant publication bias (P=0.23) (Figure 15). For the remaining outcome indicators, owing to the small number of studies (n<10), funnel plot analyses for publication bias were not conducted.

DATA AVAILABILITY:
This study is a systematic review and meta-analysis. All the data underlying the findings presented in this manuscript have been extracted from the previously published studies cited in the reference list. The complete dataset generated during the extraction process is provided in Supplementary File 2.

Figure 1: Flowchart of the study selection process. Please click here to view a larger version of this figure.

Figure 2: Forest plot of the effect of participatory dialogue on health behavior change. Please click here to view a larger version of this figure.

Figure 3: Subgroup analysis forest plot of participatory dialogue interventions. Please click here to view a larger version of this figure.

Figure 4: Forest plot of the association between behavioral confidence and health behavior change. Please click here to view a larger version of this figure.

Figure 5: Forest plot of the subgroup analysis for behavioral confidence. Please click here to view a larger version of this figure.

Figure 6: Forest plot of the effect of changes in the physical environment on health behavior change. Please click here to view a larger version of this figure.

Figure 7: Forest plot of the subgroup analysis for changes in the physical environment. Please click here to view a larger version of this figure.

Figure 8: Forest plot of the effect of emotional transformation on health behavior change. Please click here to view a larger version of this figure.

Figure 9: Forest plot of the subgroup analysis for emotional transformation effects. Please click here to view a larger version of this figure.

Figure 10: Forest plot of the effect of practice for change on health behavior change. Please click here to view a larger version of this figure.

Figure 11: Forest plot of the subgroup analysis for practice for change. Please click here to view a larger version of this figure.

Figure 12: Forest plot of the effect of changes in the social environment on health behavior change. Please click here to view a larger version of this figure.

Figure 13: Funnel plot for behavioral confidence studies. Please click here to view a larger version of this figure.

Figure 14: Funnel plot for studies on changes in the physical environment. Please click here to view a larger version of this figure.

Figure 15: Funnel plot for emotional transformation studies. Please click here to view a larger version of this figure.

Table 1: Basic characteristics and quality assessment of the included studies.

Table 2: Heterogeneity test and meta-analysis results in the initiation and maintenance phase.

Table 3: A Subgroup analysis of the associations of the Multi-Theory Model (MTM) on health behavior change research.

Table 4: Results of the sensitivity analysis.

Supplementary File 1: PRISMA guidelines. Please click here to download of this file.

Supplementary File 2: Complete dataset generated during the extraction process. Please click here to download this file.

This meta-analysis quantified the cross-sectional associations between MTM constructs and health behavior change among college students. A novel finding is the identification of key stage-specific and behavior-specific constructs: behavioral confidence was paramount for initiation, whereas practice for change (PC), particularly for physical activity (OR=1.22), had the strongest association for maintenance. Furthermore, changes in the social environment (CSE) demonstrated high stability across studies (I²=0%), highlighting its consistent role. The results reveal distinct patterns of association for the initiation and maintenance stages.

In the initiation phase, three constructs are associated with the changes in college students' healthy behavior. Participatory dialogue (PD) was significantly associated with initial behavior change, consistent with Vinayak K. Nahar et al.²⁹, who identified it as a primary structural factor for initiating physical activity. Through participatory dialogue, students gain not only scientific health knowledge but also actionable strategies and step-by-step plans from professionals and peers¹⁷. For example, in campus health programs, this could involve collaboratively setting specific, measurable goals (e.g., 'incorporate two servings of vegetables into lunch daily'), identifying potential barriers (e.g., limited time between classes), and cocreating solutions (e.g., preparing precut vegetables in advance). This approach moves beyond simply clarifying the importance of healthy eating and exercise to provide a concrete roadmap for initiation²⁹. Its association strength is shaped by interaction quality, with face-to-face formats (facilitated by nonverbal cues) often being more impactful¹⁷. The strong association observed for behavioral confidence (BC) is consistent with the theoretical proposition that self-efficacy is a key enabling factor for overcoming early barriers to behavior change³⁹. Sharma et al.³⁰ demonstrated that BC, as a core MTM construct, strongly predicts plain water consumption (over that of sugar-sweetened beverages) among college students. This confidence is associated with persistence in challenges such as resistance to dietary temptations during weight-loss efforts³⁹.Building on this evidence, digital health tools represent a promising way to foster BC through tailored goal-setting and virtual coaching, potentially increasing engagement in the initiation phase²⁰. Changes in the physical environment (CPE) are associated with initiation, potentially co-occurring with improved access to health resources. Sharma et al.¹² linked CPE to intentions for responsible drinking, while enhanced campus sports facilities increase physical activity engagement by strengthening perceived behavioral control⁴⁰, which aligns with the Ecological Model's focus on environmental influences⁴¹.

In the maintenance phase, three constructs sustain long-term adherence. Emotional transformation (ET) is associated with sustained commitment, and this relationship may be explained by concurrent positive emotional experiences, as shown in research on mask-wearing maintenance²². Engaging in activity programs with positive feedback enhances emotional experiences and increases persistence in physical activity⁴², which is consistent with Fredrickson's Broaden-and-Build Theory⁴³ that positive emotions expand cognitive capacity to support behavior maintenance. Practice for change (PC) consolidates behaviors by establishing clear goals and skills. Research¹² highlights the role of PC in sustaining gambling abstinence, aligning with the Transtheoretical Model's emphasis on translating intention into action⁴⁴. To further support this process, future research should explore the integration of the MTM with digital health platforms. AI-driven tools and mobile apps, for example, could be designed to operationalize PC by delivering adaptive reminders, facilitating progress tracking, and providing personalized feedback, thereby offering scalable support for long-term habit maintenance²⁰. Changes in the social environment (CSE) are associated with maintenance. This association may be related to concurrent levels of social support, norms, and interactions. Community studies⁴⁵ have shown that socioenvironmental constructs influence. However, most existing studies have focused on qualitative descriptions of its applicability or single-factor correlation analysis^15,20,21and lack a quantitative synthesis of the strength of the association of each MTM construct with health behaviors among college students.ng college students.physical activity levels, with support networks increasing the frequency of exercise. For students, multilevel support (from peers, families, and campus communities) is critical for sustaining healthy behaviors³³.

This review provides the first quantitative evidence that practice for change is disproportionately important for maintaining physical activity, while the stable, low-heterogeneity effect of changes in the social environment (CSE) offers a reliable target for broad-based maintenance interventions²². These findings help prioritize the constructs for testing in future experimental research, thereby advancing the scientific field by guiding the development of more theory-driven, efficient interventions. Specifically, the interventions should focus on Behavioral Confidence (BC) for initiation and Practice for Change (PC) for maintenance, rather than using scattershot approaches. This addresses a key gap identified in MTM systematic reviews¹⁷. Additionally, the use of quantitative research reduces subjective bias, thereby providing more targeted, objective, and comprehensive theoretical support and practical guidance for health behavior interventions among college students.

The current findings should be interpreted considering the limited demographic and cultural diversity of the included studies, which focused primarily on college students from the United States and India. The cross-cultural applicability of MTM constructs is supported by their successful use in diverse populations, although nuances exist. For example, research among rural populations³³ and low-income communities³⁰has demonstrated the critical role of changes in the social environment (CSE) and participatory dialogue (PD), highlighting how resource constraints and collective social structures can influence behavior change pathways. Future MTM research should intentionally include students from low-income, rural, and varied cultural backgrounds to validate and refine the model's constructs across settings.

This study has several limitations. First, the causal relationships between MTM constructs and health behaviors cannot be confirmed because of the cross-sectional designs of all the included studies²⁵. To address this, future longitudinal cohort studies are necessary to track how changes in MTM constructs over time predict the subsequent initiation and maintenance of health behaviors. Such designs would provide stronger evidence for causality¹⁷. Second, the relatively small number of included studies (n=13) raises the possibility of publication bias and may limit the statistical power of some subgroup analyses. Therefore, an important next step is to update this meta-analysis as more primary studies are published, particularly those focusing on underrepresented health behaviors and diverse cultural contexts. This will enhance the robustness and generalizability of the pooled estimates. Third, this review could not examine how sociodemographic factors, such as gender, socioeconomic status, or preexisting health conditions, moderate the associations between MTM constructs and behavior change. The primary studies lacked the necessary disaggregated data for such analyses. Understanding these potential moderators is crucial for identifying whether the model's effectiveness is universal or varies across key subgroups, which is vital for designing equitable interventions. Furthermore, future research should move beyond reporting overall associations. Primary studies should be designed and powered to conduct subgroup analyses that explore how gender, socioeconomic background, and cultural context might influence the relative importance of MTM constructs. Fourth, the geographical concentration of the included studies in the United States and India may limit the generalizability of the findings. Health behaviors are influenced by cultural, economic, and institutional factors that vary across regions. Future studies in more diverse settings are necessary to assess the cross-cultural applicability of the model. Fifth, this systematic review was restricted to studies published in the English and Chinese languages. This raises the likelihood of publication bias, thereby potentially limiting the comprehensiveness and cross-cultural generalizability of the findings. Future systematic reviews on this topic should strive to incorporate multilingual literature, possibly utilizing professional translation services, to capture a more global representation of the MTM's application across diverse cultural contexts.

Conclusions
This study consists of a systematic review and meta-analysis of 13 cross-sectional studies (2605 college students) that were conducted to quantify the cross-sectional associations of six Multi-Theory Model (MTM) constructs on changes in college students’ health behavior. The results confirmed that MTM constructs show significant cross-sectional associations with such changes. In the initiation stage, behavioral confidence (BC) had the strongest cross-sectional associations with health behavior change. In the maintenance stage, practice for change (PC) performed best, with changes in the social environment (CSE) showing high stability. Future research should expand sample sizes, include diverse populations, and, most importantly, adopt longitudinal cohort and experimental designs. Randomized controlled trials specifically testing MTM-based interventions are needed to establish causality and measure their effect on actual, sustained behavior change rather than on behavioral intentions alone. Future interventions should prioritize construct-specific and stage-matched strategies. Initiation efforts should target behavioral confidence (BC) through skill-building workshops (e.g., healthy cooking classes for dietary change) and guided practice sessions (e.g., introductory gym orientations for physical activity). For long-term interventions, such as activity planners and self-monitoring apps, this involves actively engineering changes in the social environment (CSE) through peer support networks, group challenges, and the integration of health behaviors into campus culture to provide ongoing reinforcement.