Introduction
Background/Rationale
Infectious diseases continue to pose major global public health threats, with emerging and re-emerging diseases such as human immunodeficiency virus/acquired immunodeficiency syndrome, severe acute respiratory syndrome, H1N1 influenza, and coronavirus disease 2019 (COVID-19) causing substantial morbidity and mortality [
1,
2]. The rise of antimicrobial resistance further compounds these threats by complicating treatment and control efforts [
2]. Without early detection, infectious diseases can rapidly escalate from localized outbreaks to epidemics or pandemics [
3]. Because these diseases often originate within communities, community members are typically the first to recognize changes in local health patterns. Timely reporting of such changes to health authorities enables prompt intervention and may help contain outbreaks before they become unmanageable.
Community-based surveillance (CBS) has emerged as a critical strategy for the early detection and reporting of health events [
4]. The World Health Organization defines CBS as “the systematic detection and reporting of events of public health significance within a community by the community members” [
5]. Successful CBS systems depend on community willingness to participate and on participatory engagement principles, including acceptability, collaboration, communication, local ownership, and trust [
6]. Key informants—individuals with extensive community connections, such as schoolteachers, village leaders, traditional healers, and religious leaders—serve as primary sources of health signals, thereby reducing noise while maintaining sensitivity [
7]. However, previous studies have documented several barriers to effective community engagement in CBS, including limited awareness, low trust in health systems, cultural beliefs, time constraints, and inadequate feedback mechanisms [
8,
9]. These barriers underscore the need to systematically examine the psychological and social factors that influence participation.
In Malaysia, delays in reporting health events to the authorities have persisted, with events often being reported within communities or by the media before health officials are informed [
10]. These delays hinder timely intervention and increase the risk of outbreaks. Malaysia presents several unique challenges for CBS implementation, including a diverse multiethnic population, varying levels of health literacy across rural and urban areas, and a dual healthcare system comprising both public and private sectors. In addition, the country’s tropical climate and biodiversity create favorable conditions for the emergence of vector-borne and zoonotic diseases, making early community detection particularly important [
11]. Limited community participation in health event reporting suggests a need to better understand the underlying factors.
Although the importance of CBS is increasingly recognized, substantial gaps remain in understanding the behavioral determinants of community participation, particularly in Southeast Asian settings. Although previous studies have examined challenges in CBS implementation [
12], few have systematically investigated the interplay among knowledge, attitudes, and perceptions using integrated theoretical frameworks. Furthermore, the mechanisms through which psychosocial factors influence participation intentions and engagement behaviors remain poorly understood. Several psychological theories have been developed to understand and predict health behaviors, including the Health Belief Model (HBM) [
13], the Theory of Reasoned Action (TRA) [
14], and the Theory of Planned Behavior (TPB) [
15]. These theories identify factors that influence intentions and behaviors related to specific health actions. Recent studies in Southeast Asia have demonstrated the applicability of these frameworks to health behaviors related to disease prevention and surveillance [
16], yet their integration in CBS contexts remains limited. In this study, we integrated the TRA and HBM to evaluate factors associated with community participation in CBS, with a focus on schoolteachers as community leaders. The study used a validated questionnaire comprising 3 domains: knowledge (infectious diseases, CBS, and community-level case definitions), attitudes (based on the TRA, including subjective norms, negative attitudes, intention to participate, and behavioral likelihood), and perceptions (based on the HBM, including perceived benefits, perceived barriers, perceived susceptibility, and self-efficacy) [
17].
Objectives
This study aimed to examine the interrelationships among social and health behavioral factors and their influence on the behavioral likelihood that community leaders would engage in CBS activities. Specifically, we hypothesized that (1) knowledge, subjective norms, and attitudes influence intention to participate in CBS; (2) intention, perceived benefits, perceived barriers, perceived susceptibility, and self-efficacy influence the behavioral likelihood of engaging in CBS; and (3) these factors have both direct and indirect effects on CBS participation. The findings are expected to inform the development of evidence-based educational programs and policy interventions to strengthen CBS implementation in Malaysia and similar settings, particularly through targeted training curricula for community leaders and strategic approaches to addressing identified barriers.
Materials and Methods
Study Design and Setting
A cross-sectional study design was used. Although this design allows efficient examination of relationships among multiple variables, causal inferences cannot be drawn from cross-sectional data, and temporal relationships among knowledge, attitudes, perceptions, and behavioral likelihood cannot be established definitively. The study was conducted in Kelantan, Malaysia, from March to June 2024. Four districts were selected: Kota Bharu and Bachok represented urban areas, whereas Kuala Krai and Pasir Puteh represented rural areas. Data were collected in both Islamic and public schools across these districts. Teachers in both school types serve as respected community leaders and regularly interact with families and community members, making them valuable key informants for CBS. Including both school types allowed representation of different educational structures and community engagement patterns that may influence CBS participation.
Participants
Participants were schoolteachers from Islamic and public schools in Kelantan. Schools were randomly selected from each district, whereas teachers within each school were selected by convenience sampling based on their availability during researcher visits. This approach was adopted for logistical feasibility because teachers’ availability varied considerably according to teaching schedules, administrative duties, and school activities. Although this method facilitated implementation and may have improved response rates, it may also have introduced selection bias, as participating teachers could have differed systematically from non-participating teachers in motivation, interest in health topics, or available time. This potential bias should be considered when interpreting the generalizability of the findings to the broader teacher population.
Eligibility criteria were as follows: (1) active employment as a teacher in a selected school, (2) willingness to participate voluntarily, (3) provision of informed consent, and (4) completion of the questionnaire. Teachers who were absent during data collection or declined participation were excluded.
Participants who met the inclusion criteria were briefed on the study, and both verbal and written informed consent were obtained before questionnaire completion. Schools involved in earlier stages of the study, including pilot testing and exploratory factor analysis, were excluded from this phase to minimize response bias.
Variables
Outcome variables
• Intention to participate in CBS (INT): motivation and conscious decision to engage in CBS behaviors
• Behavioral likelihood of engaging in CBS (BL): probability of actual CBS engagement
Predictor variables
• Knowledge of infectious diseases (KID1): understanding of transmission, prevention, and control measures
• Knowledge of CBS (KID2): awareness of CBS concepts and reporting systems
• Community-level case definition (KID3): ability to recognize reportable disease signals
• Negative attitudes (ATT): unfavorable beliefs about CBS activities
• Subjective norms (SN): perceived social pressure from influential others
• Perceived susceptibility (SUS): beliefs about disease likelihood without CBS participation
• Perceived benefits (BEN): beliefs about the effectiveness of CBS in disease prevention
• Perceived barriers (BARR): obstacles to CBS participation
• Self-efficacy (SE): confidence in performing CBS activities
Data Sources/Measurement
Data were collected using the validated KAP-CBS-ID questionnaire [
17,
18], a self-administered instrument comprising 3 main sections. The questionnaire was originally developed in English and underwent rigorous cultural adaptation and validation for the Malaysian context, including forward and backward translation by bilingual experts, cognitive interviewing with representatives of the target population, and pilot testing to ensure cultural appropriateness and comprehensibility [
17].
Knowledge domains
• Knowledge of infectious diseases (KID1): 18 dichotomous items assessing understanding of infectious diseases
• Knowledge of CBS (KID2): 3 dichotomous items measuring awareness of CBS
• Community-level case definition (KID3): 10 dichotomous items assessing the ability to recognize reportable signals
Attitudes domains (TRA-based)
• Negative attitudes (ATT): 6 items on a 5-point Likert scale (5 items retained after factor analysis)
• Subjective norms (SN): 4 items on a 5-point Likert scale
• Intention to participate (INT): 6 items on a 5-point Likert scale
• Behavioral likelihood (BL): 7 items on a 5-point Likert scale (6 items retained after factor analysis)
Perception domains (HBM-based)
• Perceived susceptibility (SUS): 5 items on a 5-point Likert scale
• Perceived benefits (BEN): 5 items on a 5-point Likert scale
• Perceived barriers (BARR): 3 items on a 5-point Likert scale (reduced from 7 after confirmatory factor analysis [CFA])
• Self-efficacy (SE): 4 items on a 5-point Likert scale
The questionnaire underwent rigorous validation, including content validation by 11 experts in public health and epidemiology, face validation, pilot study with 30 participants, exploratory factor analysis, and CFA.
Bias
To minimize selection bias, schools were randomly selected from each district. To reduce response bias, researchers emphasized confidentiality and encouraged honest responses during briefing sessions. Social desirability bias was addressed by including reverse-coded items and ensuring anonymous responses. However, residual bias related to self-report may have remained despite these measures, as participants may still have provided socially desirable responses or had limited insight into their own attitudes and behaviors. Schools involved in the pilot testing and exploratory factor analysis phases were excluded to prevent familiarity bias.
Study Size
Sample size was determined using 2 methods. First, applying the rule of 5 participants per item (N:P=5:1) to the 75-item questionnaire yielded a required sample of 375 participants. After accounting for an anticipated dropout rate of 20%, the adjusted sample size was 468 using Arifin’s formula [
19]:
sample size with dropouts=(calculated sample size)/(1−proportion of dropouts)=375/(1−0.2)=468.7. Second, Arifin’s web-based calculator (2017) estimated a required sample of 463 participants [
20]. On the basis of these calculations, 470 participants were recruited. According to Kline [
21], although 200 participants would generally suffice, the required sample size depends on model complexity. The final sample of 470 exceeded the minimum requirement for structural equation modeling (SEM) as recommended by Hair et al. [
22]. SEM was selected because it allows simultaneous examination of multiple interrelated pathways among constructs, accounts for measurement error, and enables testing of both direct and indirect effects within a single comprehensive model. Given the complexity of the hypothesized relationships among knowledge, attitudes, perceptions, intentions, and behaviors, SEM was more appropriate than traditional regression approaches.
Quantitative Variables
Knowledge items were scored dichotomously (correct=1, incorrect=0). Items from the 3 knowledge scales were grouped into 3 parcels for SEM analysis to achieve a more parsimonious model and more stable parameter estimates [
23]. The 3 knowledge domains (KID1, infectious disease knowledge; KID2, CBS knowledge; and KID3, community-level case definitions) were modeled as a single second-order latent construct representing overall knowledge because they theoretically reflect different facets of the comprehensive knowledge base required for effective CBS participation. This approach is consistent with hierarchical conceptualizations of health knowledge and reduces model complexity while maintaining theoretical coherence. The 3 parcels (PK1, PK2, and PK3) were formed by assigning items from each respective knowledge domain: PK1 comprised items from KID1, PK2 comprised items from KID2, and PK3 comprised items from KID3, thereby preserving the domain-specific content within each parcel. Attitude and perception items were measured on 5-point Likert scales (1=strongly disagree to 5=strongly agree). Reverse-coded items were recoded as appropriate. Factor scores were computed for each construct on the basis of the CFA results.
Statistical Methods
Descriptive statistics were generated using JAMOVI version 2.4.11 (
https://www.jamovi.org/). Numerical variables were summarized as means and standard deviations (SDs), whereas categorical variables were summarized as frequencies and percentages.
CFA was conducted in R within the RStudio (R Foundation for Statistical Computing) integrated development environment to validate the measurement model. Model fit was evaluated using the following criteria: comparative fit index (CFI) >0.92, Tucker-Lewis index (TLI) >0.92, root mean square error of approximation (RMSEA) <0.07, and standardized root mean square residual (SRMR) <0.08 [
22].
Structural equation modeling was performed using Mplus version 7.4 (Muthen & Muthen) to examine interrelationships among the study variables. Multivariate normality was assessed using Mardia’s test. Because the assumption of multivariate normality was not met, the maximum likelihood robust (MLR) estimator was used, as proposed by Yuan and Bentler [
24]. MLR was selected because it provides robust standard errors and chi-square test statistics corrected for non-normality, making it appropriate when data violate the assumption of multivariate normality. Unlike standard maximum likelihood estimation, MLR yields accurate parameter estimates and valid inference even with nonnormal data, which are common in social and behavioral research using Likert-scale measures.
The initial hypothesized model included 11 variables: KID1, KID2, KID3, ATT, SN, INT, BL, BARR, BEN, SUS, and SE. We hypothesized that KID1–KID3, SN, and ATT would influence INT, whereas INT, BARR, BEN, SUS, and SE would influence BL, as guided by the TRA and HBM (
Figure 1). Model modifications were made on the basis of modification indices and theoretical justification. Error correlations were added between items within the same factors only when the modification indices suggested substantial improvement and when a theoretical rationale supported shared method variance or conceptual overlap. All modifications were theory-driven to ensure that the final model remained interpretable and theoretically coherent rather than simply achieving optimal statistical fit.
Direct, indirect, and total effects were calculated to examine mediation pathways. Statistical significance was set at p<0.05 (2-tailed).
Ethics Statement
The study protocol was approved by the Human Research Ethics Committee of Universiti Sains Malaysia (Reference No: USM/JEPeM/22050317). Participants were briefed about the study objectives, procedures, and their rights. Written informed consent was obtained from all participants prior to data collection. Participation was voluntary, and participants were assured of confidentiality and anonymity. All completed questionnaires and electronic data files are stored securely in password-protected computers and locked filing cabinets accessible only to the research team. Data will be retained for 5 years following publication, after which they will be securely destroyed. No personal identifiers were collected, ensuring participant anonymity is maintained throughout data storage and analysis. Questionnaire completion required approximately 20–25 minutes.
Results
Participants
A total of 470 schoolteachers participated in the study. All 470 participants who completed the questionnaires were included in the analysis, with no loss to follow-up and no missing data requiring exclusion.
Table 1 presents the demographic and background characteristics of the participants.
The mean age of the participants was 43.3 years (SD, 9.5). Most participants were female (59.6%, n=280), of Malay ethnicity (99.4%, n=467), and married (79.4%, n=373), and a large proportion were government servants (85.5%, n=402). Regarding educational attainment, 87.2% (n=410) had a university education or higher, reflecting the educational requirements for teaching positions. In terms of role, 52.8% (n=248) were Islamic schoolteachers and 47.2% (n=222) were public schoolteachers. In addition, 64% (n=301) resided in urban areas, whereas 36% (n=169) resided in rural areas. These demographic characteristics have several implications for CBS participation. The relatively high educational level may facilitate understanding of infectious disease concepts and accurate reporting. The predominance of government-employed teachers may reflect greater exposure to structured communication channels and public health messaging. The inclusion of both Islamic and public schoolteachers provides insight into participation across different educational environments, whereas the urban-rural distribution helps contextualize potential differences in access to health information and surveillance systems. Overall, these characteristics suggest a sample with a strong baseline capacity to engage in CBS activities effectively.
Descriptive Data
Table 2 presents the CFA results across all domains. In the knowledge section, information on infectious diseases showed excellent composite reliability (CR, 0.92) but slightly low convergent validity (average variance extracted [AVE]=0.45), whereas CBS knowledge (CR, 0.78; AVE, 0.58) and community-level case definition (CR, 0.93; AVE, 0.50) demonstrated adequate psychometric properties. In the attitude section, subjective norms showed the strongest performance (CR, 0.895; AVE, 0.69), followed by behavioral likelihood (CR, 0.860; AVE, 0.58). Intention (CR, 0.714; AVE, 0.31) and negative attitudes (CR, 0.546; AVE, 0.20) showed suboptimal convergent validity but were retained because of their theoretical importance. The lower convergent validity of intention and negative attitudes suggests that these constructs may capture somewhat heterogeneous aspects of attitudes and intentions related to CBS participation. This may reflect the multifaceted nature of behavioral intentions and the complexity of attitudes in real-world contexts. Although these metrics were below ideal thresholds, the constructs were retained because they are theoretically central to the TRA and demonstrated acceptable reliability. Findings related to these constructs should therefore be interpreted with this measurement limitation in mind. In the perception domain, psychometric properties were generally good, with perceived benefits showing the strongest performance (CR, 0.904; AVE, 0.65), followed by perceived barriers (CR, 0.808; AVE, 0.58), self-efficacy (CR, 0.820; AVE, 0.53), and perceived susceptibility (CR, 0.816; AVE, 0.545).
The CFA models demonstrated acceptable to good fit across all domains: knowledge (RMSEA, 0.028; CFI, 0.945; TLI, 0.941), attitude (SRMR, 0.067; RMSEA, 0.055; CFI, 0.937; TLI, 0.927), and perception (SRMR, 0.055; RMSEA, 0.059; CFI, 0.962; TLI, 0.954). Factor loadings ranged from 0.33 to 0.98.
Main Results
Table 3 summarizes the model fit indices from the initial model (M0) to the final model (M3), showing clear improvement across iterations. Models M0 and M1 showed poor fit (CFI<0.86, TLI<0.85), whereas M2 and M3 achieved acceptable fit (CFI, 0.921; TLI, 0.913; RMSEA, 0.040 [90% CI, 0.036–0.044]; SRMR, 0.071).
The initial model (M0) included 3 separate knowledge factors but showed suboptimal fit and therefore required respecification. In M1, knowledge was treated as a single construct measured by 3 parcels (PK1, PK2, and PK3), which improved fit but remained inadequate. M2 incorporated error correlations between items within the same factors on the basis of modification indices and achieved acceptable fit. M3, the final model, added theoretically justified pathways to improve coherence and parsimony while maintaining acceptable fit.
Figure 2 presents the final structural model (M3) with standardized path coefficients, illustrating the relationships among all study variables.
Table 4 presents the path relationships in the final model. Knowledge significantly increased intention (β=0.419; 95% CI, 0.249–0.622;
p<0.001). Subjective norms positively influenced intention (β=0.235; 95% CI, 0.049–0.457;
p=0.038), whereas negative attitudes reduced intention (β=–0.432; 95% CI, –0.667 to –0.235;
p<0.001). Intention positively predicted behavioral likelihood (β=0.347; 95% CI, 0.230–0.486;
p<0.001). Perceived susceptibility (β=0.310; 95% CI, 0.175–0.471;
p<0.001) and perceived benefits (β=0.198; 95% CI, 0.097–0.319;
p=0.001) were positively associated with behavioral likelihood, whereas perceived barriers were negatively associated with behavioral likelihood (β=–0.132; 95% CI, –0.229 to –0.034;
p=0.008).
Additional significant paths included a positive association between perceived barriers and perceived susceptibility (β=0.205; 95% CI, 0.128–0.296; p<0.001). Self-efficacy significantly increased perceived benefits (β=0.785; 95% CI, 0.732–0.850; p<0.001) and perceived susceptibility (β=0.765; 95% CI, 0.673–0.874; p<0.001).
The final model explained 46.1% of the variance in behavioral likelihood, 57.0% of the variance in intention, 70.4% of the variance in perceived susceptibility, and 61.7% of the variance in perceived benefits.
Table 5 presents the direct and indirect relationship results. Perceived barriers showed a significant negative direct effect on behavioral likelihood (β=–0.132,
p=0.008) and a positive indirect effect through perceived susceptibility (β=0.063,
p=0.004), resulting in a non-significant total effect. Self-efficacy demonstrated a strong positive total effect on behavioral likelihood (β=0.393,
p<0.001), mediated through perceived susceptibility (β=0.237,
p=0.001) and perceived benefits (β=0.156,
p=0.002). Negative attitudes exerted a significant negative indirect effect on behavioral likelihood through intention (β=–0.150,
p=0.005). Knowledge positively affected behavioral likelihood through a significant indirect effect via intention (β=0.145,
p=0.001). Subjective norms showed non-significant total and indirect effects (
p=0.076).
Discussion
This study examined the interrelationships among social and health behavioral factors and the behavioral likelihood that community leaders would engage in CBS activities. The structural model revealed 10 significant hypothesized path relationships, explaining 46.1% of the variance in the behavioral likelihood of participating in CBS activities.
The CFA showed satisfactory psychometric properties for most constructs, supporting the multidimensional nature of CBS-related behaviors. The knowledge factors demonstrated excellent CR, particularly for information on infectious diseases, although convergent validity was slightly low. This finding is consistent with previous research suggesting that knowledge constructs can be difficult to measure because they often comprise diverse and partly independent components [
25]. The attitude and perception domains generally showed stronger psychometric properties, with subjective norms and perceived benefits demonstrating particularly robust measurement characteristics, consistent with the TRA and HBM [
13,
14].
Knowledge of infectious diseases significantly and positively influenced intention, emphasizing the importance of informed community representatives in promoting CBS engagement. This finding is consistent with studies identifying knowledge as a key determinant of health behavior intentions [
26]. However, this relationship may differ in resource-limited settings where access to health information is substantially lower than in Malaysia’s educational context. In contrast with studies from sub-Saharan Africa, where knowledge gaps were reported as major barriers to participation in disease surveillance [
27], the highly educated sample in this study (87.2% with university education) suggests that knowledge may be necessary but not sufficient for behavioral engagement. These findings support the value of educational programs that focus on disease awareness and the benefits of CBS in strengthening community participation [
28].
Subjective norms also positively influenced intention to participate in CBS, suggesting that social expectations and encouragement from peers or authorities promote participation. This finding is consistent with the TRA and TPB, which emphasize the role of subjective norms in shaping intentions [
13,
14]. Leveraging social influence through campaigns and endorsements may therefore strengthen CBS participation [
29–
31].
Negative attitudes were negatively associated with intention, underscoring the importance of addressing unfavorable perceptions to improve CBS participation. This finding is also consistent with the TRA, which emphasizes the influence of attitudes on health behavior intentions [
32]. The relatively strong magnitude of this association (β=–0.432) is consistent with findings from other Southeast Asian settings, where cultural beliefs and trust in reporting systems substantially influenced willingness to participate [
33]. Negative attitudes have similarly been shown to hinder public health initiatives in other contexts [
27].
Intention emerged as a direct predictor of behavioral likelihood. Notably, intention had a stronger direct effect than the individual perception variables of perceived susceptibility and perceived benefits, confirming its central mediating role and aligning with the TPB, which positions intention as a primary determinant of behavior [
34]. This finding suggests that intervention strategies should prioritize intention formation as a key leverage point. In practical terms, motivational messaging and communication campaigns designed to strengthen intentions to participate in CBS may be more effective than approaches that focus solely on perceptions of disease risk or surveillance benefits.
Perceived susceptibility and perceived benefits were positively associated with behavioral likelihood, whereas perceived barriers were negatively associated with behavioral likelihood (β=–0.132,
p=0.008). These direct effects are consistent with the predictions of the HBM. The negative association of perceived barriers with behavioral likelihood highlights the importance of reducing obstacles such as time constraints and perceived burden. This finding is consistent with previous studies reporting adverse effects of barriers on health behaviors [
13]. Conversely, the positive association between perceived benefits and behavioral likelihood suggests that recognizing the advantages of CBS encourages participation.
A notable finding was that self-efficacy influenced behavioral likelihood indirectly rather than directly. Although self-efficacy often shows direct associations with health behaviors in diverse populations [
35], this study suggests that it operates indirectly through perceived benefits and perceived susceptibility. This pattern indicates that confidence in performing CBS activities does not automatically translate into behavioral engagement; rather, self-efficacy may first strengthen beliefs about disease risk and the effectiveness of surveillance, which in turn increase behavioral likelihood. This indirect mechanism has important implications for intervention design. Rather than addressing self-efficacy in isolation through skills-based training alone, interventions may need to combine confidence-building activities with messaging that clearly explains how such capabilities can reduce disease risk and enhance community protection. This interpretation is consistent with Koerniawan [
36], who reported that self-efficacy had no direct effect on health-promoting behavior during the COVID-19 pandemic.
Perceived barriers demonstrated a complex relationship with behavioral likelihood, with both negative direct effects and positive indirect effects through perceived susceptibility. This counterintuitive finding suggests that individuals who perceive greater barriers to CBS participation may simultaneously report higher perceived susceptibility to disease, possibly because awareness of such obstacles heightens concern about surveillance gaps and disease risk. This pathway suggests that interventions aimed at reducing barriers should also acknowledge and address these heightened perceptions of risk so that barriers do not ultimately discourage participation.
Several indirect pathways influenced behavioral likelihood through intention: negative attitudes and subjective norms affected behavioral likelihood through intention, consistent with studies using the TRA and TPB [
26,
37]. Knowledge also indirectly influenced behavioral likelihood through intention, highlighting the role of awareness in engagement. Collectively, these indirect pathways account for a substantial proportion of the model’s explanatory power, underscoring that CBS participation is shaped by multifactorial pathways rather than isolated direct effects. For intervention design, this finding suggests the need for integrated approaches that simultaneously strengthen knowledge, reshape attitudes, mobilize social support, and build confidence in ways that influence intention and downstream behavior.
Limitations
This study has several limitations. First, the cross-sectional design precludes causal inferences regarding the relationships among knowledge, attitudes, perceptions, and behavioral likelihood. Longitudinal studies are needed to establish temporal relationships. Second, convenience sampling of teachers within schools may limit generalizability to other teachers and to other types of community leaders. Third, the self-reported nature of the data may have introduced social desirability bias, although researchers emphasized confidentiality and encouraged honest responses during briefing sessions. In addition, self-reported Likert-scale and knowledge items may be subject to measurement error, acquiescence bias, and variation in how participants interpret response options. Dichotomous knowledge items may also fail to capture the nuanced understanding required for effective CBS participation and may be influenced by guessing. Fourth, the study focused exclusively on schoolteachers in Kelantan, Malaysia, which limits generalizability to other community leaders, such as village leaders, religious leaders, and traditional healers. Fifth, the study did not account for potential confounders, including prior CBS training, personal health experiences, or organizational support, that might influence CBS participation. Sixth, the absence of qualitative data is a limitation because interviews or focus groups could have provided richer contextual understanding of the attitudes, barriers, and facilitators identified quantitatively. Such insights could help clarify the cultural, social, and institutional factors underlying the observed relationships and strengthen interpretation of the findings.
Generalizability
The findings are most applicable to schoolteachers in similar cultural and geographic contexts, particularly in Southeast Asian countries with comparable educational systems and community structures. The high educational level of the participants (87.2% with university education) may limit generalizability to community leaders with lower educational attainment. This high educational level may reflect stronger baseline knowledge, critical thinking, and health risk perception than would be expected among community leaders with less formal education. As a result, the relationships among knowledge, attitudes, and behaviors may differ in other populations; for example, leaders with lower educational attainment may rely more on subjective norms than on knowledge. Caution is therefore warranted when generalizing these findings, and interventions should be adapted to varying levels of health literacy and education. However, the TRA and HBM have demonstrated cross-cultural validity, suggesting potential applicability to other settings with appropriate cultural adaptation. Future research should include a wider range of community leaders across Malaysia and Southeast Asia and should examine CBS participation in non-school settings, such as villages, religious institutions, and healthcare facilities, to improve external validity and practical relevance.
