Modeling breast cancer screening hesitancy among Chinese women: integrating the Knowledge-Attitude-Practice, health belief, and 5 C frameworks.

Introduction
Breast cancer is the most common malignancy among women worldwide and remains a major public health concern [1]. Despite advances in early detection and treatment, incidence and mortality continue to rise, especially in low- and middle-income countries [2]. Early screening is the most effective strategy for reducing mortality [3]. Breast self-examination (BSE), a simple and low-cost method, is particularly important in resource-limited settings [4]. Unlike mammography or clinical breast examination, BSE depends on individual motivation and perceived capability, providing a useful context for examining the psychological processes underlying preventive behavior. However, participation remains low even among knowledgeable groups, revealing a persistent gap between awareness and action [5].
To explain this gap, the concept of breast cancer screening hesitancy (BCSH) has been introduced [6]. BCSH describes a psychological state in which individuals recognize screening benefits yet remain reluctant to participate, reflecting internal ambivalence rather than structural barriers. Originating from vaccine hesitancy theory [7–9], hesitancy represents a position between acceptance and refusal, shaped by uncertainty and emotional conflict. Clarifying this construct is essential for understanding the cognitive and motivational factors that disrupt the translation of intention into screening behavior.
Previous research has primarily used the Knowledge–Attitude–Practice (KAP) model or the Health Belief Model (HBM) to explain screening behavior. However, screening hesitancy measures derived from these frameworks often lack cultural adaptation and do not capture the full emotional and motivational dimensions of decision-making [10]. Several studies have attempted to integrate health belief–based frameworks with other cognitive or behavioral theories to explain cancer screening behaviors, typically through partial construct-level integration or outcome-focused modeling [11]; however, these efforts have largely emphasized selected variables or specific behavioral endpoints rather than offering a comprehensive, mechanism-oriented account of screening hesitancy, a limitation that has also been highlighted in recent conceptual syntheses of cancer screening behavior models [12]. The KAP model assumes a linear progression from knowledge to behavior, which cannot explain non-engagement among well-informed individuals [13, 14]. Although the HBM incorporates key cognitive factors, it offers limited explanatory power for sociocultural influences and motivational conflict [15, 16]. To address these limitations, this study adapts the 5 C psychological model [8] to the context of breast cancer screening. The five components of the model, namely confidence, complacency, convenience, constraints, and risk and responsibility calculation, offer a broader framework that integrates cognitive, emotional, and contextual influences. This approach is particularly meaningful in the Chinese sociocultural environment. Building on prior integrative efforts in health behavior research, this study proposes a culturally informed integrative model rather than a simple aggregation of existing frameworks. At the model level, the KAP, HBM, and adapted 5 C frameworks are conceptualized as sequential components within a unified decision-making process, reflecting a cognitive–belief–motivation pathway. Knowledge and practice serve as background exposure variables shaping cognitive readiness, health beliefs function as proximal psychological mechanisms translating exposure into motivation, and the adapted 5 C dimensions represent outcome-level manifestations of screening hesitancy. This sequential structure is theoretically motivated by the need to explain persistent hesitation among well-informed individuals, which single-framework approaches fail to capture.
At the construct level, specific psychological variables are explicitly mapped onto salient aspects of the Chinese sociocultural context. Self-efficacy reflects women’s perceived agency within family-centered decision-making structures, perceived severity captures culturally shaped interpretations of illness consequences, and perceived barriers correspond to socially patterned concerns related to time, privacy, and access. In parallel, hesitancy-related constructs such as confidence, constraints, and risk and responsibility calculation map onto trust in healthcare institutions, feasibility within everyday social roles, and moral deliberations regarding individual versus familial health responsibility. Importantly, existing evidence suggests that these psychological mechanisms do not operate uniformly across individuals but are shaped by differences in perceived agency and access to socioeconomic resources.
This layered integration reflects a cognitive–belief–motivation pathway that is especially relevant in the Chinese context, where screening information has become more accessible, yet participation remains strongly shaped by subjective beliefs, perceived self-capability, and sociocultural expectations [17, 18]. In this context, sequential integration is theoretically necessary rather than optional, as cognitive exposure, belief formation, and motivational conflict represent interdependent stages of decision-making shaped by sociocultural norms. Within this framework, psychological resources, such as self-efficacy, and socioeconomic conditions, such as education and income, are conceptualized as contextual factors that shape the strength and direction of the translation from health beliefs to hesitancy-related outcomes, thereby providing a theoretical basis for examining conditional effects in the proposed model.
Building on these perspectives, this study integrates the KAP, HBM, and adapted 5 C frameworks into a unified cognitive–belief–motivation model to explain BSE hesitancy. BSE was chosen because it captures the transition from knowledge to action and offers a strong paradigm for examining motivational pathways [19]. Using structural equation modeling (SEM), the study examines how knowledge and practice influence hesitancy through health beliefs, identifying the latent pathways linking cognition, belief, and motivation. This integrative approach offers mechanistic insight into the psychological architecture of screening hesitancy and introduces a model that bridges behavioral science and preventive medicine.
Grounded in this framework, we hypothesize that higher levels of knowledge and BSE practice are associated with lower hesitancy. In addition, stronger health belief constructs are expected to be associated with lower hesitancy and to mediate the associations between knowledge, practice, and hesitancy.

Method

Study design
A multistage cross-sectional survey was conducted to assess psychological and behavioral determinants of BSE hesitancy. SEM analyses were subsequently performed in a staged manner, including mediation, moderation, and multi-group analyses.

Study population and sampling
Women aged 18 to 70 years living in Guizhou Province, China, were recruited using a two-stage region-based sampling approach to enhance geographic representation. In the first stage, five prefecture-level cities were selected to reflect geographic dispersion and variation in urban–rural contexts across the province. In the second stage, one administrative district was randomly selected within each city as the data collection site.
Participant recruitment was conducted at or in connection with routine outpatient visits and community health activities in each selected district, including those taking place at community health centers in urban areas and township health centers in rural areas. These settings served solely as recruitment venues and were not involved in study design, data analysis, or interpretation.
This sampling approach was intended to ensure coverage of heterogeneous urban and rural contexts rather than to generate a province-wide probability sample, and is appropriate for structural equation modeling studies focusing on psychological and behavioral mechanisms. Eligible women were informed about the study by trained research assistants or healthcare staff and invited to participate on a voluntary basis.
Sample size was calculated using a medium effect size of 0.3 [20], with α = 0.05 and power = 0.80, yielding a minimum of 700 participants. Allowing for nonresponse, the target sample was 849. Eligibility criteria included Chinese citizenship, at least one year of residence in Guizhou, and informed consent. Women with a current or previous diagnosis of breast or other cancers were excluded.

Instruments
Three validated Chinese-language instruments were used. The Knowledge–Practice (KP) component was derived from a previously validated KAP questionnaire [21]. The knowledge component included 16 items assessing awareness of breast cancer risk factors, symptoms, and screening methods. Responses were dichotomous (Yes/No/Not Sure) and coded as 1 = Correct, 0 = False. Negatively worded items were reverse-scored to maintain consistency. The practice component comprised 4 items assessing engagement in BSE–related behaviors. Responses were dichotomous (Yes/No/Not Sure) and coded as 1 = Good, 0 = Poor. Negatively framed items were reverse-coded for consistency.
The HBM scale [22] consisted of 22 items covering six dimensions: perceived susceptibility, perceived severity, perceived benefits, perceived barriers, self-efficacy, and cues to action. All HBM items were rated on a five-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree).
The 5 C BCSH scale included 20 items assessing five psychological dimensions: confidence, complacency, convenience, constraints, and risk and responsibility calculation. All items were measured using a five-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree). For both the HBM and 5 C scales, negatively worded items were reverse-coded prior to analysis so that higher scores consistently reflected higher levels of the underlying constructs. All instruments were previously validated among Chinese women to ensure linguistic and cultural suitability. The full item wording for all instruments is provided in Appendix A.

Participant recruitment and data collection
Participants were recruited using a voluntary convenience sampling approach through primary health care settings in Guizhou Province, China, including community health centers in urban areas and township health centers in rural areas. During routine outpatient visits or community health activities, eligible women were informed about the study objectives and procedures by trained research assistants or healthcare staff and invited to participate on a voluntary basis.
No predefined sampling frame, clinic list, or contact registry was used; therefore, the total number of individuals approached could not be systematically enumerated. Recruitment and data collection were conducted between March and May 2025. A total of N = 849 women provided informed consent and completed the questionnaire.
Data were collected anonymously via Wenjuanxing, an online survey platform. The survey required about 20 to 25 min to complete. IP restrictions ensured responses originated within Guizhou, with one submission per device. No personal identifiers were collected. Combining online and offline recruitment helped minimize selection bias.

Statistical analysis
Descriptive statistics were conducted using SPSS 26.0. Continuous variables were summarized as means and standard deviations, and categorical variables as frequencies and percentages. All model-based analyses (CFA, SEM, moderation, and multi-group SEM) were performed in R 4.5.1 using the lavaan and semTools packages.
Measurement properties of the latent constructs were evaluated prior to structural modeling. Internal consistency was assessed using Cronbach’s α and composite reliability (CR). Convergent validity was examined using average variance extracted (AVE), and discriminant validity was evaluated using the heterotrait–monotrait (HTMT) ratio. Recommended cutoff criteria were applied to determine the adequacy of reliability and validity.
SEM examined how knowledge and practice predicted psychological hesitancy (5 C model) through the mediating role of HBM constructs. The attitude component of the KAP model was excluded to avoid overlap with attitudinal elements already captured in the HBM and 5 C frameworks. CFA supported the latent structure of all instruments. The structural model was specified to represent how multiple health belief components jointly relate to distinct dimensions of screening hesitancy within a unified decision-making framework.
The mediation model specified knowledge and practice as predictors of six HBM constructs, which subsequently predicted the five 5 C dimensions. A full mediation model and a partial mediation model were compared. At the item level, both knowledge and practice were modeled as latent exogenous variables using WLSMV; only the full mediation model converged. To facilitate mediation, moderation, and multi-group structural equation modeling and to improve estimation stability, an alternative composite-based specification was subsequently estimated. Composite scores for knowledge and practice were constructed by averaging their respective items and standardizing the resulting scores. These composite predictors were treated as continuous variables and estimated using the maximum likelihood estimator with robust standard errors (MLR). The composite-based model yielded a structural pattern of associations that was consistent with the item-level results, supporting the robustness of the full mediation structure. Moderation and multi-group SEM were conducted using the composite-based model to ensure convergence and stable estimation.
To assess potential multicollinearity among the exogenous predictors in the structural model, variance inflation factors (VIFs) were examined. All VIF values were below commonly recommended thresholds, indicating no evidence of problematic multicollinearity.
Self-efficacy (latent), income (observed), and education (observed) were examined as moderators of the structural relationships between the HBM and 5 C constructs. Self-efficacy was modeled as a latent moderator, and latent interaction terms were constructed using the indProd() function. Because latent interaction modeling is not supported under WLSMV estimation, moderation analyses involving latent interactions were estimated using the MLR estimator. Income and education were included as observed moderators. Income was measured using ordered income categories and treated as an ordinal variable. For moderation analyses, income was was treated as an approximately continuous variable and standardized (mean-centered and scaled to unit variance). Education level was originally measured as an ordinal categorical variable with six ordered categories (primary school, junior high school, high school or vocational school, associate degree, bachelor’s degree, and postgraduate degree or above) and was likewise standardized prior to analysis. Interaction terms were generated using the standardized moderator variables and incorporated into the structural model.
Multi-group structural equation modeling (MGSEM) was conducted to examine whether structural relationships differed across residential area (urban vs. rural) and education level. For MGSEM, education was dichotomized a priori into lower education (associate degree or below) and higher education (bachelor’s degree or above) groups to facilitate subgroup comparisons. Two nested models were compared: an unconstrained model with freely estimated structural paths and a constrained model with regression paths constrained to equality across groups. Model differences were evaluated using the Satorra–Bentler scaled chi-square difference test implemented via lavTestLRT() under MLR estimation. Only structural paths were constrained, as measurement invariance for all instruments had been established previously.

Results

Participant characteristics
A total of 849 participants were included. The mean age was 39.73 years (SD = 14.35; range 18–70), and 74% were between 18 and 50 years. Over half lived in urban areas (53.83%), followed by rural (35.10%) and suburban areas (11.07%). Most participants held at least an associate degree (59.84%) and were employed (70.08%). The median monthly household income was 3597 RMB (IQR: 1000–5000.5). Most lived in owned homes (84.60%) and in households of three to four members (52.30%). Only 4.00% reported chronic diseases, and 15.50% had a family history of breast cancer. Details are presented in Table 1.

Measurement model assessment
The measurement model demonstrated satisfactory psychometric properties, with acceptable internal consistency, convergent validity, and discriminant validity across all latent constructs (Tables S1–S3).
CFA showed good model fit (χ² = 6319.575, df = 1810, χ²/df = 3.49, CFI = 0.962, TLI = 0.959, RMSEA = 0.054 (90% CI = 0.053–0.056), SRMR = 0.055). All standardized loadings were significant and above 0.63, supporting convergent validity (Table S4- S5).

Structural model assessment
The item-level structural equation model showed acceptable fit (χ² (1837) = 9189.282, χ²/df = 5.00, CFI = 0.938, TLI = 0.934, RMSEA = 0.069 [90% CI: 0.067–0.070], SRMR = 0.085) (Table 2). All hypothesized paths were significant (p < 0.05), with most at p < 0.0001 (Table 3). Knowledge demonstrated consistently positive associations with all HBM constructs, including perceived severity, perceived susceptibility, perceived benefits, perceived barriers, self-efficacy, and cues to action. In contrast, practice-related behaviors exhibited predominantly negative associations with perceived severity, perceived susceptibility, and perceived benefits, while showing positive associations with perceived barriers, self-efficacy, and cues to action. The HBM constructs showed differentiated associations with the five dimensions of the 5 C framework. Higher levels of perceived severity and perceived susceptibility were generally associated with increased levels of hesitancy-related dimensions, whereas perceived benefits, self-efficacy, and cues to action were consistently associated with reduced hesitancy across multiple 5 C outcomes. Perceived barriers emerged as the strongest positive predictor of constraints, highlighting its central role in shaping structural and practical obstacles to breast cancer screening. Although several standardized coefficients exceeded 1.0, the overall structural pattern was theoretically coherent and aligned with the satisfactory global model fit, supporting the robustness of the hypothesized pathways.

Table 4 presents the R² estimates for the endogenous constructs. Among the HBM dimensions, self-efficacy exhibited the highest explained variance. Within the 5 C framework, confidence exhibited the highest explained variance among the five dimensions.

At the item level, the full mediation model converged successfully, whereas the partial mediation model did not converge. Accordingly, the item-level full mediation model was retained. In addition, a composite-based mediation model was estimated as a robustness check.
As expected given that knowledge and practice were treated as aggregated background exposure variables in the composite specification, the explained variance for the HBM constructs was minimal (R² ≤ 0.01). In contrast, moderate levels of explained variance were observed for the 5 C outcomes (Table 5). The standardized regression estimates are presented in Table 6. The overall structural pattern of associations was consistent with that observed in the item-level model, supporting the robustness of the full mediation structure.

Mediation models
To examine whether the effects of knowledge and practice on the 5 C outcomes were fully or only partially mediated by the HBM constructs, full and partial mediation models were compared. The item-level partial mediation model failed to converge, but the composite-based models provided stable estimates with acceptable fit (χ²/df ≈ 3.5, CFI ≈ 0.891, RMSEA = 0.059–0.064; Table S6). The chi-square difference test was not significant (Δχ² = 16.178, Δdf = 10, p = 0. 09465; Table S7), indicating no advantage of the partial mediation model. Therefore, the item-level full mediation model was retained as the final structural model, with the composite results supporting its robustness. The final structural model is presented in Fig. 1.

Moderator effects
This section examines the moderating roles of self-efficacy, income, and education level within the unified BSE model rather than across subgroups.
Self-Efficacy: Self-efficacy significantly moderated multiple HBM-5 C pathways (Table S8). A significant negative interaction was observed between self-efficacy and perceived susceptibility in predicting convenience. For constraints, self-efficacy showed significant interaction effects with perceived severity, perceived susceptibility, perceived barriers, and cues to action, with both positive and negative directions observed. Regarding complacency, self-efficacy significantly moderated the associations with perceived severity, perceived susceptibility, and perceived barriers. In addition, a significant negative interaction was found between self-efficacy and perceived susceptibility in relation to risk and responsibility calculation.
Income: Income moderated three pathways (Table S9). A significant negative interaction was observed between income and perceived susceptibility in predicting complacency. In contrast, positive interaction effects were found for perceived barriers in relation to complacency and for cues to action in relation to confidence.
Education: Education level moderated several structural pathways (Table S10). A negative interaction was observed between education level and perceived severity in predicting constraints, whereas a positive interaction was found between education level and perceived barriers in relation to constraints. In addition, education level showed significant positive interaction effects with perceived severity, self-efficacy, and knowledge in predicting risk and responsibility calculation. A further positive interaction was observed between education level and perceived severity in relation to confidence.

Measurement invariance
Structural invariance was tested across residential settings and education levels (Table S11). For residential settings, the constrained model did not differ significantly from the unconstrained model (Δχ² = 51.701, Δdf = 42, p = 0.145), indicating invariant structural pathways between urban and rural groups. Similarly, no significant difference was found across education levels (Δχ² = 49.049, Δdf = 42, p = 0.2224), supporting structural invariance across both subgroups.

Discussion

Summary of key findings
This study developed and validated an integrated cognitive–belief–motivation model combining the KAP framework, the HBM, and an adapted 5 C psychological framework to examine BSE hesitancy among Chinese women. The final structural model supports a sequential process in which health beliefs function as a central psychological mechanism linking cognitive exposure, reflected by knowledge and practice, to differentiated motivational expressions of screening hesitancy.
From a theoretical perspective, these findings indicate that the HBM may be more appropriately understood as a mediating belief system rather than solely as an outcome-oriented explanatory model when addressing hesitancy-related phenomena. This conceptualization helps to clarify why high levels of knowledge do not necessarily translate into screening engagement, a pattern that has been repeatedly observed but insufficiently explained in prior screening research. By explicitly positioning health beliefs as proximal mechanisms rather than distal predictors, the present findings engage with long-standing theoretical debates regarding the limitations of linear knowledge–behavior models, including traditional applications of the KAP framework in cancer screening research. By situating the adapted 5 C dimensions at the motivational outcome level, the proposed framework provides a mechanism-oriented account of screening hesitancy that captures both internal ambivalence and contextual constraint.
Within the Chinese sociocultural context, this integrative structure offers a coherent lens through which cognitive readiness, belief formation, and motivational conflict can be jointly interpreted, without reducing screening behavior to either informational deficits or structural barriers alone. Notably, the prominence of self-efficacy and perceived barriers over threat-based beliefs reflects context-specific decision patterns in which screening-related motivation is shaped more strongly by subjective capability, role expectations, and practical feasibility than by fear-based risk appraisal alone. This contextualized pattern provides empirical support for the cultural adaptation of the integrated framework in explaining BSE hesitancy among Chinese women.

Mediation mechanisms
A more fine-grained examination of the mediation effects indicates that knowledge and practice do not influence screening hesitancy uniformly through all HBM constructs. Rather, their effects are primarily transmitted through a subset of higher-impact belief components. Specifically, self-efficacy, perceived barriers, and cues to action emerged as the most stable and influential mediators linking knowledge and practice to multiple 5 C dimensions, constituting the core psychological pathways through which cognitive exposure and prior experience are translated into hesitancy-related outcomes [23].
Among these, self-efficacy played a particularly central mediating role across confidence, convenience, and risk and responsibility calculation. This pattern suggests that individuals’ judgments of their own capability are a critical prerequisite for transforming knowledge advantages and prior practice into lower levels of psychological hesitancy. Such a finding is especially plausible in the context of breast self-examination (BSE), which is inherently skill-dependent and self-initiated. Even when adequate knowledge is present, insufficient confidence in one’s ability to perform BSE correctly may sustain elevated levels of hesitancy [24, 25].
In contrast, perceived barriers primarily mediated hesitancy dimensions related to constraints and convenience, underscoring the foundational role of subjective and objective obstacles, such as time costs, procedural inconvenience, and psychological discomfort, in screening decision-making. Cues to action exerted a stronger influence on hesitancy dimensions associated with behavioral initiation and maintenance, highlighting the triggering function of external reminders, social support, and medical recommendations in converting latent intentions into action. By comparison, perceived severity and susceptibility demonstrated weaker or more selective mediation effects across pathways, indicating that threat appraisal alone is insufficient to account for screening hesitancy in the BSE context.
Taken together, these findings suggest that knowledge and practice reduce screening hesitancy not primarily by amplifying disease threat perceptions, but by strengthening beliefs related to action readiness, including perceived capability, feasibility as sessments, and behavioral triggers. This mechanism-oriented pattern provides a clear theoretical basis for intervention design, indicating that efforts to reduce BSE hesitancy should prioritize enhancing self-efficacy, lowering perceived barriers, and reinforcing effective cues to action, rather than focusing solely on disease risk communication [26, 27].

Moderation analysis
The moderation analyses further illustrate that the translation of health beliefs into screening hesitancy is not uniform, but contingent upon both psychological resources and socioeconomic conditions. Across multiple pathways, self-efficacy, income, and education systematically shaped the strength and direction of associations between HBM constructs and the 5 C dimensions, underscoring the conditional nature of belief-to-hesitancy processes [28, 29]. At the psychological level, self-efficacy emerged as a key contextual factor that modulated how threat perceptions and perceived barriers were reflected in hesitancy-related outcomes. Higher self-efficacy generally attenuated reliance on convenience and deliberative considerations, suggesting that women who perceive themselves as capable may engage less in extended evaluation when responding to perceived risk, a pattern that aligns with prior work on agency and health decision-making [28, 29]. At the same time, self-efficacy exhibited differentiated moderating patterns across motivational dimensions, highlighting the complexity of agency in shaping screening-related decision processes.
At the socioeconomic level, income and education functioned as structural contexts that conditioned the expression of health beliefs. Higher socioeconomic resources appeared to buffer the demotivating effects of threat on complacency and constraints, while simultaneously amplifying the salience of perceived barriers and informational cues in shaping confidence and responsibility-related deliberation [30–32]. Education, in particular, strengthened the linkage between threat-related beliefs, self-efficacy, and responsibility-oriented evaluation, consistent with previous evidence that educational attainment reshapes how health information and perceived risk are cognitively processed [32, 33]. Taken together, these findings reinforce the view that screening hesitancy emerges from a context-dependent decision process in which psychological agency and socioeconomic position jointly shape how health beliefs are enacted, rather than from fixed or uniform belief–behavior relationships.

Structural path invariance
Structural invariance across residential and educational groups indicates that the core cognitive–motivational processes underlying BSE hesitancy are robust across sociodemographic contexts [33, 34]. This demonstrates the universality of the proposed model while emphasizing the importance of differentiated implementation. Although mechanisms remain stable, subgroups differ in baseline knowledge, access to information, and motivational readiness. Interventions should therefore maintain a unified theoretical foundation while tailoring delivery strategies—for example, strengthening basic knowledge and skills among rural or lower-education women and addressing structural or psychological barriers among higher-education or urban groups [34].

Public health implications for improving BSE practices
The SEM findings highlight the central mediating role of health beliefs in translating knowledge and intention into sustained BSE practice. These findings suggest that information provision alone may be insufficient, and that interventions may benefit from simultaneously addressing perceived barriers, self-efficacy, and perceived benefits to support the transition from intention to regular and standardized behavior [31, 32].
From an HBM perspective, programs may consider targeting both threat appraisal (perceived severity and susceptibility) and coping appraisal (self-efficacy, benefits, and barriers) [22]. Risk communication could emphasize early detection while providing clear, feasible, and non-intimidating guidance. Coping-oriented strategies, including skills training, behavioral modeling, and emotional support, appear particularly relevant for strengthening confidence and reducing hesitation.
Within the 5 C framework, multiple theoretically informed intervention entry points can be identified. Confidence may be improved through credible information and hands-on skills practice; convenience may be supported by emphasizing the simplicity and privacy of BSE; constraints may be addressed through reminders, step-by-step guidance, and behavioral support; complacency may be reduced through targeted education and survivor narratives; and risk and responsibility calculation may be reinforced by framing BSE as both a personal and familial health responsibility [6, 7].
Findings were largely consistent across subgroups, with moderation analyses indicating subgroup-specific nuances (self-efficacy buffering, income-linked sensitivity, and education-linked deliberation), supporting a unified theoretical framework with scope for tailored delivery. For example, clinical programs could incorporate brief BSE coaching during routine visits, community campaigns might combine risk-awareness messages with coping strategies and skill demonstrations, and policy initiatives may prioritize low-literacy materials, peer navigators, and mobile outreach services.
Overall, the SEM results provide a theoretically grounded basis for identifying potential intervention targets. The proposed implications should therefore be interpreted as theory-informed considerations derived from cross-sectional structural associations, and future longitudinal and intervention studies are needed to empirically test these pathways. Beyond BSE, this integrative framework offers a transferable conceptual model for understanding hesitancy in other preventive health behaviors, bridging psychological theory and public health practice.

Limitations
This study has several limitations. First, the cross-sectional design precludes causal inference, and the relationships identified should be interpreted as associative rather than temporal. Second, the data were collected through self-reported online questionnaires, which may be subject to recall or social desirability bias. Third, participants were recruited using a voluntary convenience sampling approach without a predefined sampling frame, which precluded the calculation of a formal response rate and may have introduced potential non-response bias. Fourth, although the sample covered both urban and rural areas in Guizhou Province, generalizability to other regions of China may be limited due to contextual and cultural differences. Fourth, although multiple HBM constructs were modeled as parallel mediators, potential interactions or reciprocal influences among these belief components were not explicitly specified. In health behavior theories, belief constructs may influence one another dynamically, and future studies using longitudinal or more flexible modeling approaches could further examine these interrelationships. Finally, some SEM estimation issues required the use of parceling to ensure model stability, which may limit the granularity of item-level interpretations. Future longitudinal and mixed-method studies are recommended to validate the temporal and contextual stability of the proposed model.

Conclusion
This study identified self-efficacy, perceived barriers, and cues to action as pivotal determinants of breast self-examination hesitancy. Integrating the KAP framework, HBM and the 5 C psychological framework within a structural equation model demonstrated that health beliefs serve as key mediators linking knowledge and practice to psychological hesitancy.
Beyond the context of breast cancer screening, this integrative cognitive–belief–motivation model provides a unified paradigm for understanding preventive health behaviors more broadly. The theoretical value of this framework lies in clarifying why awareness does not automatically translate into action, while its practical relevance lies in informing mechanism-based approaches to reducing behavioral hesitancy.
From a public health perspective, the findings suggest that efforts to address BSE hesitancy may benefit from strengthening women’s confidence and skills, reducing structural and psychological barriers, and supporting ongoing cues to action through healthcare providers and digital platforms. Community-based education and low-literacy materials are particularly crucial for rural or lower-education groups. These findings provide a theoretical and empirical basis for considering stratified yet integrated screening promotion efforts in China. Future studies should extend this framework to other screening modalities and use longitudinal designs to further validate its predictive capacity and policy relevance.

Supplementary Information