Meta-Analysis of Implementation Intentions Interventions in Promoting Physical Activity among University Students

Physical activity (PA) is essential for preventing chronic disease and for physical and psychological well-being, yet global inactivity remains high among adolescents and young adults, with participation declining sharply in late adolescence and early adulthood, including university students. Prior reviews of PA interventions in university populations report promising but inconsistent effects across total PA and moderate PA. Behavior change techniques (BCTs) embedded in e-health (e.g., goal setting, monitoring, feedback) show potential. Implementation intentions (IIs) are a key BCT defined by if-then planning, manifested as action planning (AP: when, where, how) and coping planning (CP: responses to obstacles). IIs can mediate or moderate the intention–behavior relationship, addressing the well-documented intention–behavior gap for complex behaviors like regular PA. While meta-analyses show II effectiveness for various health behaviors and for PA in adults and general populations, evidence specific to university students is limited and potentially heterogeneous. This study asks: What is the effect of implementation intentions on promoting PA among university students, and to what extent does efficacy differ across moderators (publication year, participant activity level, intervention strategy, reinforcement, and duration)?

Previous syntheses of PA interventions among university students yield mixed findings: small effects for moderate PA but not total PA, and medium effects from diverse interventions (educational materials, incentives, social media, wearables) without statistical significance. E-health interventions show potential to improve PA in young adults, with marginally medium effects in university students, and individualized, determinant-focused approaches are recommended. IIs, as if-then plans expressed via AP and CP, have strong evidence as determinants of behavior change and as mediators/moderators of the intention–behavior link. Meta-analyses report small-to-medium II effects on PA in adults and general populations, and small pooled effects in chronic disease groups. However, combining across wide age spans may increase heterogeneity. There remains a gap in meta-analytic evidence focused specifically on university students, including moderator analyses within this population.

Design followed the Cochrane Handbook and PRISMA. Protocol registered on PROSPERO (CRD42023424579). Search strategy: comprehensive database searches (PubMed, Embase, Cochrane Library, Web of Science, PsycINFO, PsycArticle) using MeSH and free terms for participants (university students), intervention (implementation intentions, action planning, coping planning, planning), outcomes (physical activity, exercise), and design (randomized controlled trials). Recursive manual searches of references, conference studies, and grey literature (thesis databases) were conducted. Eligibility: RCTs (including pilot and cluster RCTs) of II interventions to promote any form of PA in university students; outcomes could include steps, energy expenditure, exercise time, number/frequency; included any PA form (steps, moderate, vigorous, total PA). Exclusion: university students with disabilities or mentally incompetent; unspecified II measures; multi-component interventions without extractable separate II group; no control group. Data extraction: study characteristics (authors, year, sample size/distribution, region); participant info (age, female proportion, baseline PA level); intervention details (model, content, reinforcement, duration); outcomes (measurement instruments and PA outcome formats). Quality assessment: Cochrane Risk of Bias 2 tool across seven domains; funnel plots and Egger tests assessed publication bias; assessments performed independently by two authors with adjudication by a third. Statistical analysis: continuous outcomes standardized to SMDs (95% CIs); means and SDs extracted or transformed; random-effects inverse variance model (DerSimonian–Laird). Heterogeneity assessed via Cochran’s Q (p < 0.1 significant) and I² (25/50/75% as low/medium/high). Pre-specified subgroup analyses: publication year (≥2013 vs <2013), participants (general vs inactive), intervention strategy (AP vs AP+CP), reinforcement (yes vs no), duration (≥6 weeks vs <6 weeks). Sensitivity analyses via stepwise elimination. Analyses conducted in STATA16.

• Included 12 RCTs (n=1916 participants). Most conducted in developed countries; interventions predominantly AP, delivered face-to-face, with durations from 2 weeks to 11 months; outcomes mainly self-reported.
• Overall effect: II interventions significantly increased PA vs controls (SMD = 0.31, 95% CI: 0.12 to 0.50, p < 0.001). Heterogeneity moderate-to-high (I² = 70.1%, p < 0.001). Individual study SMDs ranged 0 to 1.
• Publication bias: Funnel plot showed asymmetry, but Egger test p > 0.1 indicated no significant small-study bias.
• Subgroup analyses (SMD [95% CI], p-value): • Publication year ≥2013: 0.60 (0.02, 1.17), p = 0.042; <2013: 0.22 (0.04, 0.39), p = 0.015; between-group p = 0.216. • Participant type: General: 0.27 (0.04, 0.50), p = 0.023; Inactive: 0.40 (0.05, 0.74), p = 0.025; between-group p = 0.550. • Intervention strategy: AP: 0.36 (0.14, 0.57), p = 0.001; AP + CP: 0.15 (−0.22, 0.52), p = 0.418; between-group p = 0.349. • Reinforcement: Yes: 0.45 (0.08, 0.83), p = 0.019; No: 0.22 (0.01, 0.42), p = 0.036; between-group p = 0.285. • Duration: ≥6 weeks: 0.05 (−0.07, 0.18), p = 0.407; <6 weeks: 0.42 (0.17, 0.67), p = 0.001; between-group p = 0.010.
• Sensitivity analysis: Stepwise exclusion showed minimal variation in pooled effect sizes, indicating robustness.

The meta-analysis demonstrates that implementation intentions produce small-to-medium increases in physical activity among university students, aligning with findings in adults, general populations, and chronic disease groups. Compared with broader health behavior meta-analyses reporting larger effects, PA may be more complex and influenced by covariates (e.g., multiple goals, affect, habit), possibly attenuating II effects when plans lack specificity or fail to address barriers. Subgroup results suggest stronger effects in inactive students, potentially via enhanced self-efficacy through simulated planning and coping with obstacles. AP-only interventions were significant, whereas combined AP+CP were not, possibly due to dilution or less precise focus; optimizing how CP integrates with AP warrants further investigation. Shorter follow-up periods yielded larger effects, and reinforcement improved outcomes, indicating relapse over time and the importance of continued prompts or booster sessions. More recent publications showed larger effects, hinting at maturing intervention designs. Integrating IIs with e-health delivery (text, email, web, social media) is promising given university students’ digital nativity, enabling scalable, cost-effective, and sustainable campus interventions.

Implementation intention interventions significantly promote physical activity in university students, with particularly favorable effects among inactive students, AP-only strategies, reinforced interventions, and follow-up periods under six weeks. These findings offer practical guidance for campus health promotion, highlighting II interventions as cost-effective and feasible. Future work should integrate IIs with digital technologies to refine delivery, explore personalized and multi-component models, and optimize reinforcement frequency and modes to sustain behavior change.

• Limited number of included studies constrained moderator analyses to exploratory subgroups; meta-regression was not performed.
• PA outcomes were primarily self-reported, introducing measurement variability; objective tools (pedometers/accelerometers) are recommended in future trials.
• Delivery mode effects could not be assessed (only two online interventions vs ten face-to-face), limiting insights on e-health vs in-person implementation.
• Combining diverse PA outcome types may challenge interpretation despite focusing on a single population; standardized PA measures would improve comparability and evidence stability.