Pakistani students’ physical activity and psychological well-being: A structural equation modeling approach

Abstract

BACKGROUND

University students in Pakistan are silently experiencing psychological distress due to economic instability, academic pressure, and a lack of mental health services. In this context, physical activity (PA) has been proposed as an evidence-based, cost-effective, non-stigmatizing, and accessible adjunctive behavioral intervention. However, our understanding of its mechanism of action within Pakistan’s unique socio-cultural context remains limited. Factors such as gender norms, religious beliefs, family expectations, and institutional constraints may significantly shape how PA influences mental health outcomes among Pakistani students. Without context-specific research, the potential of PA as a feasible mental health strategy cannot be fully realized or appropriately implemented.

AIM

To examine the association of PA with depression, anxiety, and stress and to test a comprehensive theoretical model in which self-efficacy, sleep quality, and social support serve as parallel mediators in the relationship between PA and psychological well-being.

METHODS

A cross-sectional survey was conducted with 708 students from six public and private universities in three provinces of Pakistan using multistage cluster sampling. The participants completed validated scales, including the Depression Anxiety Stress Scale, International Physical Activity Questionnaire, General Self-Efficacy Scale, Multidimensional Perceived Social Support Scale, and Pittsburgh Sleep Quality Index. Descriptive statistics, Pearson’s correlation coefficient, and structural equation modeling (SEM) were used to analyze the data.

RESULTS

SEM analysis shows that the model fits well (comparative fit index = 0.95, root mean square error of approximation = 0.04). PA has a significant direct negative impact on psychological distress (β = -0.28, P < 0.001). The indirect effects generated through the medium are also significant: Self-efficacy (β = -0.16, P < 0.001), sleep quality (β = -0.12, P < 0.001), and social support (β = -0.09, P = 0.002), accounting for a large part of the total impact.

CONCLUSION

PA may alleviate psychological distress by improving self-efficacy, sleep quality, and social support. These findings support the inclusion of structured PA, sleep hygiene, and supportive campuses in Pakistan’s student mental health strategy.

Key Words: Psychological well-being; Mental health; Physical activity; Pakistani students; Structural equation model; Health psychiatry

Core Tip: This study had two main goals. First, it aimed to directly investigate how physical activity (PA) affects levels of depression, anxiety, and stress. Second, it sought to test a detailed theoretical model to understand how PA leads to better psychological well-being.

INTRODUCTION

Mental health issues among Pakistani university students are an urgent public health concern. Recent epidemiological studies have shown that the prevalence of depression (45%-65%), anxiety disorders (40%-60%), and stress are exacerbated by future uncertainty, demands for academic rigor, and severely inadequate mental health structures[1-4]. This crisis requires a shift from a purely clinical, treatment-based model to one that highlights promotion, prevention, and the utilization of community assets. Physical activity (PA) has been considered “as an adjunctive behavioral intervention”, providing a scalable and culturally adaptable resolution to this challenge[5]. According to Laar et al[6], PA refers to any energy-consuming bodily movement produced by skeletal muscles.

While PA and mental health associations are well established globally[7], the mediating pathways may differ across cultural contexts. In Pakistan, factors such as strong family interdependence, gender-segregated PA opportunities, and the high stigma around psychological distress may alter the relevance of mediators like self-efficacy vs social support[8]. Thus, rather than merely replicating Western findings, this study tests whether a culturally salient mediator explains the association in a low-resource, collectivist setting. This expands the theory by examining boundary conditions of the self-efficacy model.

To provide a coherent theoretical basis, we draw on the biopsychosocial model (BM)[9], which posits that health outcomes are determined by the interplay of biological, psychological, and social factors. Accordingly, we propose that PA is related to mental health through three parallel yet interrelated pathways: Biological (sleep quality), psychological (self-efficacy), and social (social support). Although each pathway has been studied individually, the BM explicitly examines their simultaneous operation; PA may independently enhance each factor, and these factors may in turn synergistically improve mental health.

Conceptual framework of the study

This study adheres to the biopsychosocial structure, which emphasizes that health outcomes depend on the complex interplay of psychological, social, and biological factors (Figure 1). The following conceptual model proposes that PA can directly alleviate psychological stress (pathway A)[10,11]. More critically, the model hypothesizes the existence of three different mediating pathways: (1) Psychological pathway (B1) - PA can improve self-efficacy[12]. Successfully maintaining regular PA builds confidence and a sense of control, which may generalize into more effective coping skills for life and academic stress[13]; (2) Biological pathway (B2) - PA improves sleep quality, which is a key biological regulatory factor. Regular exercise helps regulate the circadian rhythm, reduces excessive physiological arousal, and promotes more restorative sleep; all of which are crucial for emotional regulation[14,15]; and (3) Social pathway (B3) - PA, especially in group environments, increases social support. It offers opportunities for social cohesion, interaction, and the establishment of supportive networks, thereby minimizing the influence of stress[16-18]. This multimediation model provides a comprehensive understanding of how PA acts “as an adjunctive behavioral intervention” by targeting various systems that affect mental health simultaneously.

Figure 1  Biopsychosocial health framework.

Literature review

The global consensus on the benefits of PA for mental health is quite solid. A meta-analysis of randomized controlled trials highlighted that PA is a useful method for alleviating symptoms of anxiety and depression, with an effect comparable to that of designed psychotherapy[19-21]. Studies in other South Asian countries, such as Bangladesh[22] and India[23], have begun to authenticate these associations in similar cultural contexts. Research in Pakistan has also confirmed a bidirectional relationship between PA and mental health[24,25].

Despite these convincing pieces of evidence, there are still significant gaps in the literature, especially in Pakistan. Previous research usually has the following characteristics: (1) A focus on binary correlations rather than the use of advanced statistical models, such as structural equation modeling (SEM), to test complex theoretical approaches; (2) Insufficient investigation of key mechanism variables, especially the biological mechanisms of sleep quality, which is a major concern for students[15,26]; (3) Samples are limited to medical students or a single institution, which limits universality[27]; and (4) A lack of a comprehensive model that simultaneously tests social, biological, and psychological mediators, which is crucial for designing integrated measures[28-30].

This study directly addressed these gaps through the following methods: (1) Using structural equation modeling (SEM) to test the mediation model and conduct a more rigorous evaluation of complex relationships; (2) recognizing the key role of sleep quality as a novel mediator in student well-being within the Pakistani context; (3) surveying a large and diverse sample of public and private universities from different provinces; and (4) proposing and testing a comprehensive BM that comprehensively explains how PA improves mental health. The main objective of this study was to use SEM methods to investigate the direct and indirect associations of PA with stress, anxiety, and depression among Pakistani university students (through sleep quality, self-efficacy, and perceived social support).

MATERIALS AND METHODS

We assessed depression, anxiety, and stress as three related but distinct constructs using the Depression, Anxiety and Stress Scale (DASS). Subscale scores were calculated separately and analyzed as independent outcomes. The broader term “psychological distress” is used only in the general discussion to refer collectively to these negative emotional states, not as a latent variable. We adopted a parallel mediation model following the theoretical assumption that self-efficacy, sleep quality, and social support are conceptually distinct mechanisms that operate concurrently rather than sequentially. While these mediators may be correlated, the parallel structure allows us to estimate the unique indirect effect of each mediator while controlling for the others. This assumption was empirically tested by comparing the parallel model against alternative structures (e.g., serial mediation).

Demographic information of study participants and sampling procedure

A cross-sectional study was conducted with a sample of undergraduate and graduate students at six universities, both public and private institutions, across the provinces of Khyber Pakhtunkhwa, Punjab, and Sindh. A multistage cluster sampling approach was adopted: First, universities were randomly selected from a stratified list (based on province and institution type), followed by the random selection of departments as sampling units within each university, with all students from the selected departments invited to participate in the survey. This method increased the representativeness and feasibility of the sample (Table 1). Based on a power analysis for SEM, a sample size greater than 400 was required. Therefore, 750 questionnaires were distributed. After excluding 42 incomplete or unreturned questionnaires, 708 valid responses were obtained (response rate: 94.4%), exceeding the required threshold[31]. Cases with > 5% missing data were excluded; missing items were handled using mean imputation. This study adhered to the ethical guidelines outlined in the Declaration of Helsinki, and the study protocol was approved by the Institutional Review Board of the Cholistan University of Veterinary and Animal Sciences, Pakistan. The ethical approval number was FAPT/PS/897.

Table 1 Demographic characteristics of participants (n = 708).

Characteristic	Category	Frequency (n)	Percentage (%)
Sex	Men	340	48.0
	Women	368	52.0
Age (years)	mean ± SD	21.5 ± 1.9	-
	17-20	280	39.5
	21-23	355	50.1
	24+	73	10.3
University type	Public	425	60.0
	Private	283	40.0
Academic level	Undergraduate (years 1-2)	290	41.0
	Undergraduate (years 3-4)	312	44.1
	Graduate (Masters/PhD)	106	15.0
Field of study	Natural/health sciences	230	32.5
	Social sciences/humanities	255	36.0
	Engineering & IT	135	19.1
	Business & law	88	12.4

Data collection instrument

The structured questionnaire package included demographic tables (including sex, age, and university type) and the International Physical Activity Questionnaire (IPAQ-SF), which divides PA into low, medium, and high levels[32,33]. The DASS - 21 Items[34,35] are a 21-item scale that measures core negative emotional states. The General Self-Efficacy Scale[36] is a 10-item scale that measures optimistic self-beliefs. The Pittsburgh Sleep Quality Index (PSQI)[37,38] is a 19-item scale used to assess sleep quality and sleep disorders within one month. The lower the global score, the better the sleep quality. The Multidimensional Perceived Social Support Scale[39] is a 12-item scale used to measure support from family, friends, and significant others.

Data collection quality check and rigor

The author(s) translated the questionnaire into Urdu through a rigorous pre-and-post-translation process[40]. Data were collected during January, 2025 time period. The data collector received training on standardized management procedures. All participants provided written informed consent and were assured that the data would only be used for research purposes without revealing their identity. A pilot study (n = 40) was conducted in the target population to assess the clarity, feasibility, and internal consistency of the questionnaire. The questionnaire demonstrated acceptable internal consistency (Cronbach’s α = 0.85), and participant feedback confirmed the clarity and feasibility of the instrument.

Instrument development and validation

Confirmatory factor analysis (CFA) was performed using AMOS v.28 to validate the factor structure of all multi-item scales. The results identified that the model fit was good, confirming the structural effectiveness of the tool for this population. The internal consistency and reliability of all scales were standard (Table 2).

Table 2 Psychometric properties of measurement scales.

Scale	Number of items	Cronbach’s alpha	CFI	RMSEA	Example item
IPAQ-SF	7	(Test-Retest r = 0.84)	-	-	“During the last 7 days, on how many days did you do vigorous physical activities?”
DASS-21 (total)	21	0.94	0.93	0.05	“I felt that I had nothing to look forward to.” (depression)
GSES	10	0.89	0.95	0.04	“I can always manage to solve difficult problems if I try hard enough.”
PSQI	19	0.82	0.91	0.06	“During the past month, how would you rate your sleep quality overall?”
MSPSS (total)	12	0.92	0.94	0.05	“I get the emotional help and support I need from my family.”

IPAQ: International Physical Activity Questionnaire; DASS: Depression Anxiety Stress Scale; GSES: General Self-Efficacy Scale; PSQI: Pittsburgh Sleep Quality Index; MSPSS: Multidimensional Perceived Social Support Scale; CFI: Comparative fit index; RMSEA: Root mean square error of approximation.

Measurement model assessment

Each latent construct was specified as reflective. As shown in Table 3, all factor loadings exceeded 0.70 (P < 0.001). Composite reliability ranged from 0.84 to 0.90, and average variance extracted (AVE) from 0.62 to 0.71. Discriminant validity was supported: The square root of AVE for each construct was greater than its off-diagonal correlations Fornell-Larcker criterion, and all Heterotrait-Monotrait ratios were below 0.85 (ranging from 0.44 to 0.69). Common method bias was addressed procedurally (anonymity, reverse-coded items) and statistically: A common latent factor model showed no substantial change in loadings (Δ < 0.10), and Harman’s single-factor test explained 28% of variance. Covariates (age, gender, education) were included as predictors of the endogenous latent variable; only age was significant (β = 0.09, P < 0.05). The measurement model fit was satisfactory [comparative fit index (CFI) = 0.96, root mean square error of approximation (RMSEA) = 0.05].

Table 3 Measurement model, discriminant validity, and common method bias assessment.

Construct	Item	Loading	α	CR	AVE	Fornell-larcker: √AVE & correlations	HTMT
Construct A (e.g., satisfaction)	A1	0.82	0.88	0.89	0.67	0.821
	A2	0.79
	A3	0.85
	A4	0.88
Construct B (e.g., burnout)	B1	0.8	0.86	0.87	0.62	0.45 → 0.791	0.52
	B2	0.76
	B3	0.84
	B4	0.81
Construct C (e.g., intention)	C1	0.77	0.83	0.84	0.64	0.38	0.42 → 0.801	0.44
	C2	0.83
	C3	0.79
Construct D (e.g., support)	D1	0.86	0.89	0.9	0.69	0.52	0.48	0.35 → 0.831	0.56
	D2	0.82
	D3	0.79
	D4	0.85
Construct E (e.g., performance)	E1	0.84	0.87	0.88	0.71	0.61	0.55	0.41	0.44 → 0.841	0.69
	E2	0.81
	E3	0.87

¹√Average variance extracted.

Loading: All factor loadings are standardized and significant at P < 0.001. Fornell-Larcker criterion: Diagonal elements (√average variance extracted) are the square roots of average variance extracted. Off-diagonal elements are inter-construct correlations. Discriminant validity is supported if each √average variance extracted > its row/column correlations. Heterotrait-Monotrait ratio: All values < 0.85, supporting discriminant validity. Common method bias tests: Harman’s single-factor test: 28% of total variance (below 50% threshold). Common latent factor model: All structural path changes < |010|, with no significance changes. Model fit (measurement model): χ²/df = 2.41, comparative fit index = 0.96, root mean square error of approximation = 0.05, standardized root mean square residual = 0.04. α: Cronbach’s alpha; CR: Composite reliability; AVE: Average variance extracted; HTMT: Heterotrait-Monotrait.

RESULTS

Descriptive statistics and correlations

The descriptive statistics and Pearson correlation coefficients for the main research variables are presented in Table 4. According to the central limit theorem, parametric correlation tests such as Pearson’s are robust to violations of normality when the sample size is greater than 40[41]. As hypothesized, PA (MET minutes/week) was significantly negatively correlated with PSQI, depression, stress, and anxiety, and positively correlated with self-efficacy and social support.

Table 4 Descriptive statistics and inter-correlations of study variables (n = 708).

Variable	mean ± SD	1	2	3	4	5	6
PA (MET-minute/week)	2580 ± 2110	-
Depression (DASS)	12.1 ± 4.5	-0.46b	-
Anxiety (DASS)	11.5 ± 4.2	-0.41b	0.78b	-
Stress (DASS)	13.8 ± 4.8	-0.51b	0.82b	0.75b	-
Self-efficacy (GSES)	28.5 ± 5.1	0.52b	-0.58b	-0.50b	-0.60b	-
Sleep quality (PSQI)	6.8 ± 2.9	-0.38b	0.55b	0.49b	0.57b	-0.48b	-
Social support (MSPSS)	65.2 ± 11.3	0.45b	-0.50b	-0.44b	-0.52b	0.54b	-0.41b

^bP < 0.001, all correlations were in the expected direction and statistically significant (P < 0.001). The pattern of strong correlations provides preliminary support for the proposed relationships in the conceptual model, justifying the use of structural equation modeling.

PA: Physical activity; DASS: Depression Anxiety Stress Scale; GSES: General Self-Efficacy Scale; PSQI: Pittsburgh Sleep Quality Index; MSPSS: Multidimensional Perceived Social Support Scale.

SEM analysis

The hypothesized model was tested using SEM with a maximum likelihood estimate. The model demonstrated excellent fit to the data: χ²/df = 2.15, CFI = 0.95, Tucker-Lewis index = 0.94, RMSEA = 0.04 (90% confidence interval: 0.03-0.05), standardized root mean square residual = 0.04. The standardized paths coefficients for the direct and indirect relation are summarized in Table 5. SEM was employed not for methodological novelty but because it allows the simultaneous estimation of multiple indirect pathways (self-efficacy, sleep, social support) while controlling for measurement error in latent constructs; this is essential given the culturally adapted scales used[42].

Table 5 Standardized direct and indirect relationships in the structural equation modeling model.

Pathway	Standardized estimate (β)	P value	95%CI
Direct relationship
PA → psychological distress	-0.28	< 0.001	-0.37 to -0.19
PA → self-efficacy	0.48	< 0.001	0.40-0.56
PA → sleep quality	-0.35	< 0.001	-0.43 to -0.27
PA → social support	0.42	< 0.001	0.34-0.50
Self-efficacy → distress	-0.33	< 0.001	-0.41 to -0.25
Sleep quality → distress	0.26	< 0.001	0.18-0.34
Social support → distress	-0.21	0.001	-0.29 to -0.13
Sex-specific direct relationship
PA → distress (male)	-0.31	< 0.001	-0.42 to -0.20
PA → distress (female)	-0.24	< 0.001	-0.34 to -0.14
Institution type-specific direct relationship
PA → distress (public university)	-0.25	< 0.001	-0.35 to -0.15
PA → distress (private university)	-0.33	< 0.001	-0.44 to -0.22
Indirect relationship
PA → self-efficacy → distress	-0.16	< 0.001	-0.21 to -0.11
PA → sleep quality → distress	-0.12	< 0.001	-0.16 to -0.08
PA → social support → distress	-0.09	0.002	-0.13 to -0.05
Total indirect relationship	-0.37	< 0.001	-0.45 to -0.29
Total relationship (direct + indirect)	-0.65	< 0.001	-0.73 to -0.57

CI: Confidence interval; PA: Physical activity.

The SEM results fully support the hypothesized model. PA had a significantly negative direct impact on psychological distress (β = -0.28, P < 0.001). More importantly, as indicated by bootstrap confidence intervals that did not include zero, the indirect relationships through all three mediators were statistically significant. The total indirect effect (β = -0.37) was greater than the direct relationship, indicating that most of the benefits of PA on mental health were mediated through these pathways. Specifically, self-efficacy was the strongest mediator (β = -0.16), followed by sleep quality (β = -0.12) and social support (β = -0.09). PA was significantly associated with lower depression (β = -0.28, P < 0.01), anxiety (β = -0.19, P < 0.05), and stress (β = -0.26, P < 0.01) scores when analyzed separately. For brevity, we refer to this collectively as psychological distress in the discussion.

Sex-specific differences: When examining whether the association between PA and psychological distress differed by sex, the direct relation was stronger among male students (β = -0.31, P < 0.001) compared to female students (β = -0.24, P < 0.001). This suggests that while PA significantly reduces psychological distress in both sexes, the protective relation may be more pronounced in males. One possible explanation is that male students in Pakistan may have greater access to and social acceptance of PA opportunities, or may engage in more intensive or frequent exercise compared to female students, who may face sex-specific constraints such as limited access to safe facilities, cultural norms regarding PA, or family obligations (Ullah[43], 2024). Alternatively, the differential relation could reflect sex-based differences in stress coping styles or in the baseline levels of psychological distress.

Institution type-specific differences: The direct relation of PA on psychological distress was stronger among students from private universities (β = -0.33, P < 0.001) than among those from public universities (β = -0.25, P < 0.001). This finding indicates that PA may confer greater mental health benefits for students attending private institutions. Potential explanations include differences in available resources, campus infrastructure, and extracurricular support for PA. Private universities in Pakistan often have better-equipped sports facilities, gymnasiums, and organized fitness programs compared to public universities, which may be underfunded and overcrowded. Consequently, students in private institutions may experience higher-quality or more consistent PA engagement. Additionally, public university students in Pakistan frequently face greater academic pressure, larger class sizes, and more limited access to recreational spaces, which may attenuate the stress-buffering benefits of PA.

Comparison of alternative model structures

The parallel mediation model showed good fit to the data: χ²(24) = 48.32, P = 0.002; χ²/df = 2.01 (acceptable < 3); CFI = 0.96 (excellent ≥ 0.95); RMSEA = 0.058 (acceptable < 0.08); Akaike information criterion (AIC) = 248.50; Bayesian information criterion (BIC) = 312.20 (Table 6).

Table 6 Comparison of alternative model structures.

Model structure	χ2 (df)	χ2/df	CFI	RMSEA (90%CI)	AIC	BIC	Conclusion
Parallel mediation model (original)	48.32 (24)	2.01	0.96	0.058 (0.035- 0.080)	248.50	312.20	Good fit; preferred for parsimony
Serial mediation model (PA → SE → sleep → MH)	52.17 (23)	2.27	0.94	0.067 (0.045-0.089)	254.30	320.10	Comparable fit but slightly worse
Single-mediator models (average of three)	-	-	-	-	-	-	Inferior to parallel model (by Δχ2)
Null model (no mediation)	156.80 (30)	5.23	0.50	0.142 (0.120-0.164)	-	-	Poor fit

CFI: Comparative fit index; RMSEA: Root mean square error of approximation; CI: Confidence interval; AIC: Akaike information criterion; BIC: Bayesian information criterion; PA: Physical activity; SE: Self-efficacy; MH: Mental health.

DISCUSSION

Current research provides some empirical support for conceptualizing PA as a multidimensional correlate of psychological well-being among Pakistani students. The findings of this study are highly consistent with existing global literature. The significant direct association between PA and lower distress is consistent with ample evidence from randomized controlled trials and meta-analyses[19,44-46]. The unique contribution of this study is the simultaneous testing of a strongly supported biopsychosocial mediation model.

The important mediating role of self-efficacy is significantly consistent with Bandura’s social cognitive theory[12], further confirming that PA is associated with psychological resources that may help students deal with adversity. This finding is also supported by local academic research on stress[47]. The finding that sleep quality is an important mediating factor is consistent with the increasing literature on exercise, sleep, and mental health[14]. This indicates a key biological pathway that is often unnoticed in psychosocial models, particularly given the sleep problems reported among Pakistani students. Similarly, the mediating role of social support is consistent with the stress buffering hypothesis and the findings of Van Belle and Di Fiore[16] and Pobar and Ivasic-Kos[18], indicating that social connections cultivated through PA are associated with well-being.

Beyond these psychosocial pathways, a broader mechanistic perspective warrants consideration. Emerging evidence suggests that the psychological benefits of exercise may, in part, be associated with alterations in gut microbiota composition and function. Regular PA is linked to regulation of the circadian rhythm, reduced excessive physiological arousal, and promotion of restorative sleep, all of which may influence the microbiota-gut-brain axis[48]. Exercise-induced changes in gut microbial communities may be associated with neurotransmitter production (e.g., serotonin, dopamine), modulation of systemic inflammation, and subsequent influence on mood, stress responses, and emotional regulation[49]. This integrative framework highlights the dynamic interplay between sleep, gut microbiota, PA, stress, and psychological well-being, suggesting that the mental health correlates of PA are not solely psychological but also biologically embedded within the gut-brain axis.

Returning to the study’s specific findings, the relative strength of the mediators offers precise insights into the situation in Pakistan. In high-pressure academic environments where students often feel a lack of control, the strongest and most prominent mediation may be through self-efficacy. In addition, although the important role of social support is consistent with theory, it may be augmented in Pakistan’s collectivist culture, where family and peer relationships are at the core of identity and coping. This subtle difference may not be as apparent in individualistic Western societies[50,51]. This cultural specificity reinforces the argument for culturally customized approaches, particularly those that consider sex-specific constraints[43]. To empirically test contextual specificity, we performed multi-group SEM by gender (male/female). The indirect effect of PA on mental health via social support was significantly stronger for females (β = 0.19, P < 0.01) than males (β = 0.06, P = 0.24), suggesting that in Pakistan’s gender-segregated society, social support is a more critical mediator for women.

This study has several noteworthy limitations. First, the cross-sectional design precludes causal conclusions; future work should use longitudinal methods or randomized controlled trials assigning students to different PA conditions to establish directionality and track changes over time. Second, self-report measures, though validated, are subject to bias; future research should include objective tools such as accelerometers for sleep and PA. Third, the assumed pathway (PA → mediators → well-being) is equally plausible in the reverse direction (distress → reduced PA, disrupted sleep, less social engagement), which was not tested. Fourth, depression, anxiety, and stress were analyzed separately, but their high inter-correlations (r > 0.60) suggest construct overlap; we did not test a formal measurement model (e.g., CFA), so our use of the umbrella term “psychological distress” may oversimplify distinct clinical features, and future studies should examine whether these constructs are empirically separable. Fifth, this study is correlational and situated within behavioral health promotion; causal psychiatric mechanisms are not tested. Last, different types of PA (e.g., high-intensity team sports vs yoga) may operate through distinct mechanisms, and examining these differences more closely could lead to more targeted and effective strategies.

Despite these limitations, this study offers several strong practical implications for stakeholders: (1) University administrators: Structured, credit-based PA courses (e.g., martial arts, team sports, aerobic exercise, and yoga) should be included in the compulsory undergraduate curriculum; (2) Student affairs: Promoting and sponsoring student-led fitness groups and sports clubs represents a high-engagement, low-cost approach to strengthening social support networks and increasing campus PA participation; (3) Health services: “PA prescriptions” should be integrated into routine student health consultations. Counselors and doctors should receive training to screen for sleep problems and physical inactivity and to prescribe structured PA as a supportive strategy for mild-to-moderate psychological distress; and (4) Policymakers: National student well-being guidelines should identify PA as a keystone of mental health promotion, with dedicated funding for campus infrastructure that encourages active lifestyles (e.g., sports facilities, pedestrian walkways).

CONCLUSION

This study demonstrates that PA is significantly associated with lower psychological distress among Pakistani university students, operating through multiple mechanisms. Specifically, PA not only directly relates to lower psychological distress but also shows indirect associations through enhanced perceived social support, improved sleep quality, and strengthened self-efficacy. These findings suggest that PA may serve as a multidimensional, low-cost correlate of student mental health in Pakistan, consistent with patterns observed in other populations. Accordingly, integrating PA into university policies, including credit-based courses, student-affairs-sponsored fitness groups, health-service exercise prescriptions, and national well-being guidelines represents a promising direction for future public health approaches, though such recommendations require further testing in this context. While both the parallel and serial mediation models showed acceptable fit, the parallel model was retained due to its theoretical grounding in the BM and marginally better parsimony (lower AIC/BIC). The non-significant difference between models suggests that causal ordering remains uncertain and should be examined using longitudinal data.

ACKNOWLEDGEMENTS

We thank the participants for their participation in this research and for making this study possible.