Predicting the Determinants of Colorectal Cancer Screening Behaviors Using Protection Motivation Theory: A Cross‐Sectional Partial Least Squares Structural Equation Modeling Analysis

ABSTRACT

Background

Colorectal cancer (CRC) screening reduces mortality by 40%–60%, yet participation remains low globally. In this study, we applied protection motivation theory (PMT) to identify determinants of CRC screening behavior in Iran using partial least squares structural equation modeling (PLS‐SEM).

Design and Methods

This cross‐sectional study was conducted on 433 adults over 50 years of age using stratified sampling in Neyshabur, Iran in 2024. Participants completed a validated 43‐item PMT questionnaire assessing key cognitive factors (perceived sensitivity, severity, self‐efficacy, response efficacy, response costs, fear, rewards, and behavioral intention), with self‐reported screening behavior as the outcome.

Results

Among the PMT constructs, self‐efficacy had the highest mean score (69.43 ± 18.97 out of 100), while actual screening behavior had the lowest (26.12 ± 9.93 out of 100). PLS‐SEM analysis revealed significant pathways: perceived severity (β = 0.118, p = 0.030) and response efficacy (β = 0.172, p = 0.003) positively influenced behavioral intention, while perceived rewards negatively impacted intention (β = −0.197, p < 0.001). Fear mediated sensitivity/severity effects on intention (β = 0.155, p = 0.003). Notably, self‐efficacy and response costs showed nonsignificant relationships with intention.

Conclusion

Participants demonstrated moderate intention (54.8 of 100) yet low screening behavior (26.1 of 100), highlighting a critical intention–behavior gap. While PMT constructs effectively predicted screening intention, their limited ability to explain behavior underscores the influence of contextual barriers beyond cognitive appraisals in this Iranian cohort. Future interventions should integrate PMT‐based education targeting threat appraisal with system‐level strategies (e.g., mailed test kits, navigational support) to bridge this implementation gap.

Abbreviations

CRC

colorectal cancer

PMT

protection motivation theory

PLS‐SEM

partial least squares structural equation modeling

FIT

fecal immunochemical tests

1. Introduction

Colorectal cancer (CRC) represents one of the most prevalent and lethal malignancies globally, imposing a substantial burden on healthcare systems and accounting for millions of deaths annually [1]. Robust evidence indicates that early detection through regular screening can significantly reduce CRC mortality by approximately 40%–60% [2]. Despite this proven effectiveness, participation rates in CRC screening programs remain suboptimal in many populations [3]. This underscores the critical need for a deeper understanding of the psychological, social, and demographic determinants of screening behaviors as a research and public health priority [4].

Protection motivation theory (PMT) offers a robust framework for understanding preventive health behaviors, including cancer screening, with core constructs of threat appraisal (severity and perceived sensitivity, balanced against maladaptive rewards) and coping appraisal (response efficacy, self‐efficacy, and perceived barriers or response costs) [5, 6, 7, 8, 9, 10, 11]. Norman et al. and other studies support PMT's usefulness in predicting cancer screening intentions and behaviors, highlighting self‐efficacy and perceived severity as key predictors [7, 12, 13, 14, 15, 16].

In the context of CRC screening, PMT posits higher likelihood of participation when individuals view CRC as a severe personal threat (high perceived severity and sensitivity), believe screening reduces risk (high response efficacy), and feel capable of completing the process (high self‐efficacy), while greater barriers and rewards of inaction can reduce motivation [8, 9, 10, 11, 17, 18, 19]. Consequently, PMT has guided interventions to boost screening uptake by enhancing threat perceptions, strengthening coping appraisals (clarifying benefits, building self‐efficacy), and addressing barriers [10, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29]. Evidence suggests PMT‐based interventions that focus on knowledge and barriers can raise screening participation, with self‐efficacy and response efficacy consistently predicting screening intentions across populations [9, 26].

Analyzing the complex, multilevel relationships between latent PMT constructs (e.g., perceived severity, sensitivity, self‐efficacy, response efficacy, and response costs) and actual screening behavior requires advanced multivariate methods like structural equation modeling (SEM) [30]. Within SEM, two primary approaches exist: covariance‐based SEM (CB‐SEM) and partial least squares SEM (PLS‐SEM) [30]. CB‐SEM is ideal for confirming complex causal theories, assuming multivariate normality and larger sample sizes. PLS‐SEM, however, offers greater flexibility, being robust to violations of distributional assumptions, suitable for smaller samples, and focused on maximizing variance explained in dependent variables (prediction) [30, 31].

Comparative studies, such as Hair et al., indicate PLS‐SEM often performs better under non‐normal data or smaller sample conditions, while CB‐SEM excels in rigorous causal theory testing under ideal conditions [30]. PLS‐SEM has also proven effective in modeling complex predictors of health behaviors, including screening [32].

Despite numerous studies applying PMT to understand CRC screening behaviors, the specific use of PLS‐SEM for predicting CRC screening determinants within the PMT framework remains underexplored. This gap may lead researchers to select analytical methods less suited to their data characteristics or predictive goals.

Therefore, this study aims to address this gap by evaluating the performance of PLS‐SEM in predicting determinants of CRC screening behavior based on PMT. This study is significant for several reasons: First, by potentially identifying key PMT predictors of CRC screening behavior more accurately, it can inform the design of more effective interventions to increase participation. Second, by rigorously applying PLS‐SEM in a health context, it provides clearer guidance for researchers on selecting the appropriate SEM approach (CB‐SEM for theory confirmation vs. PLS‐SEM for prediction) based on data characteristics and research objectives. Third, it contributes to the methodological and theoretical literature on cancer prevention by integrating PMT with advanced analytical techniques.

2. Methods

This cross‐sectional study will be conducted in 2024 among adults aged ≥ 50 years residing in Neyshabur, the second‐largest city in Khorasan Razavi Province, Iran, with an estimated population of 300,000. Inclusion criteria comprised willingness to participate, ability to comprehend and respond to the questionnaire, adequate literacy, and no personal history of CRC. Participants completing less than 50% of the questionnaire were excluded. A stratified sampling method with proportional allocation was employed. First, Neyshabur was divided into 11 strata corresponding to its comprehensive health centers. Participants were then systematically selected from each stratum proportionate to its population size. The sample size was calculated based on recommendations for partial least squares structural equation modeling (PLS‐SEM), requiring 5–10 participants per questionnaire item and a minimum of 300 participants [33]. The questionnaire contained 43 items (39 measuring PMT constructs and 4 assessing behavior), leading to a minimum target sample size of 430; ultimately, 433 participants completed the survey.

Data were collected using a validated PMT questionnaire [26]. This instrument measures eight latent constructs: perceived sensitivity (four items, Cronbach's α = 0.91), perceived severity (seven items, α = 0.88), fear (four items, α = 0.70), perceived reward (three items, α = 0.83), response efficacy (six items, α = 0.79), self‐efficacy (six items, α = 0.89), response cost (six items, α = 0.74), and behavioral intention (three items, α = 0.93). All items for these constructs used a 5‐point Likert scale (1 = strongly disagree to 5 = strongly agree). Actual CRC screening behavior was assessed using four separate dichotomous (yes/no) items. The study received ethical approval from the Regional Ethics Committee of Mashhad University of Medical Sciences (IR.MUMS.FHMPM.REC.1404.030), and written informed consent was obtained from all participants before enrollment. All study procedures complied with the principles of the Declaration of Helsinki and applicable institutional guidelines.

PLS‐SEM was used to analyze the relationships between PMT constructs and CRC screening behavior. This method was selected due to its suitability for predictive and exploratory research, flexibility with non‐normal data distributions and smaller sample sizes, and primary focus on maximizing explained variance in the dependent variables [31, 34]. Cohen's f² was used to assess the effect size of the model. f² values of 0.02–0.15 indicate a small effect, 0.15–0.35 a medium effect, and ≥ 0.35 a large effect on the dependent variable [35]. The coefficient of determination, R², was used to estimate the proportion of variance in the dependent variables explained by the model. R² values are interpreted as follows: < 0.19 = very weak; 0.19–0.33 = weak; 0.33–0.67 = moderate; ≥ 0.67 = substantial [36]. In this study, all participants responded to every item in the questionnaires, and there were no missing data in the questionnaires; analyses were conducted using complete data. Analyses were performed using SmartPLS software (version 4.0) [31]. The statistical significance level was set at α = 0.10 for Full PLS‐SEM and α = 0.05 for all other hypothesis tests. All statistical tests were conducted two‐sided.

3. Results

The study included 433 participants with a mean age of 58.26 ± 5.41 years. The majority were female (61.9%, n = 268), married (72.3%, n = 313), had three to five children (55.2%, n = 239), possessed less than a high school diploma (45.3%, n = 196), and identified as housewives (43.2%, n = 187). Detailed demographic characteristics are presented in Table 1.

Table 1.

Demographic characteristics of participants (n = 433).

Characteristic	Category	Number	%
Gender	Female	268	61.9
	Male	165	38.1
Marital status	Single	5	1.2
	Married	313	72.3
	Divorced	17	3.9
	Widowed	98	22.6
Number of children	No children	12	2.8
	1–2 children	128	29.6
	3–5 children	239	55.2
	> 5 children	54	12.5
Educational level	Illiterate	92	21.2
	< High school diploma	196	45.3
	High school diploma	116	26.8
	University education	29	6.7
Occupation	Worker	47	10.9
	Employee	35	8.1
	Freelance	85	19.6
	Housewife	187	43.2
	Retired	64	14.8
	Unemployed	4	0.9
	Farmer	11	2.5

Among the PMT constructs, self‐efficacy exhibited the highest mean score (69.43 ± 18.97), while the lowest score was observed for the actual screening behavior construct (26.12 ± 9.93). Scores for all constructs, standardized to a 0–100 scale, are detailed in Table 2.

Table 2.

Scores of PMT constructs (standardized to 0–100 scale).

Construct	Mean ± standard deviation	Median (first quartile, third quartile)
Perceived sensitivity	41.86 ± 13.6	43.75 (31.25, 50.00)
Perceived severity	66.48 ± 20.22	67.86 (53.57, 78.57)
Fear	45.71 ± 18.72	43.75 (31.25, 56.25)
Perceived reward	42.71 ± 21.48	41.67 (25.00, 58.33)
Response effectiveness	67.72 ± 17.84	66.67 (54.17, 79.17)
Self‐efficacy	69.43 ± 18.97	70.83 (54.17, 79.17)
Response cost	39.23 ± 18.86	37.50 (25.00, 50.00)
Behavioral intention	54.83 ± 18.28	50.00 (50.00, 66.67)
Behavior	9.93 ± 26.12	0.00 (0.00, 0.00)

Initial PLS‐SEM analysis fitting the full PMT model revealed Nonsignificant paths (p > 0.05) from response cost (β = −0.011, p = 0.988), self‐efficacy (β = 0.054, p = 0.396), and perceived sensitivity (β = −0.011, p = 0.843) to behavioral intention (Table 3, Figure 1). Consequently, these paths were removed, and a refined model was fitted (Figure 2, Table 4).

Table 3.

PLS‐SEM path coefficients—full PMT model.

Path	β ± SE	t value (p value)	f 2
Behavioral intention → Behavior	0.063 ± 0.016	3.818 (< 0.001)	0.041
Fear → Behavioral intention	0.152 ± 0.055	2.797 (0.005)	0.023
Perceived reward → Behavioral intention	−0.176 ± 0.056	3.173 (0.002)	0.001
Perceived sensitivity → Behavioral intention	−0.011 ± 0.053	0.198 (0.843)	0.009
Perceived sensitivity → Fear	0.293 ± 0.057	5.066 (< 0.001)	0.088
Perceived severity → Behavioral intention	0.109 ± 0.06	1.82 (0.069)	0.009
Perceived severity → Fear	0.254 ± 0.053	4.783 (< 0.001)	0.068
Response cost → Behavioral intention	−0.011 ± 0.067	0.015 (0.988)	0.001
Response effectiveness → Behavioral intention	0.159 ± 0.06	2.64 (0.008)	0.020
Self‐efficacy → Behavioral intention	0.054 ± 0.067	0.848 (0.396)	0.002

Note: Bolded values are nonsignificant paths at the 0.05 level and were removed in the refined PMT model.

Abbreviations: β, standardized regression coefficient; SE, standard error of the parameter estimate.

Figure 1.

Full PLS‐SEM fitting based on the full model of PMT. The numbers on the paths represent regression coefficients and associated p‐values, while the numbers within the constructs denote the coefficients of determination (R²).

Figure 2.

PLS‐SEM fitting based on the modified model of PMT after removing nonsignificant paths. The numbers on the paths represent regression coefficients and associated p‐values, while the numbers within the constructs denote the coefficients of determination (R²).

Table 4.

PLS‐SEM path coefficients—refined PMT model (nonsignificant paths removed).

Path	β ± SE	t value (p value)	f 2
Behavioral intention → Behavior	0.063 ± 0.016	3.815 (< 0.001)	0.041
Fear → Behavioral intention	0.155 ± 0.052	2.986 (0.003)	0.026
Perceived reward → Behavioral intention	−0.197 ± 0.049	3.971 (< 0.001)	0.038
Perceived sensitivity → Fear	0.294 ± 0.057	5.106 (< 0.001)	0.088
Perceived severity → Behavioral intention	0.118 ± 0.054	2.172 (0.030)	0.012
Perceived severity → Fear	0.254 ± 0.053	4.795 (< 0.001)	0.068
Response effectiveness → Behavioral intention	0.172 ± 0.057	2.993 (0.003)	0.024

Abbreviations: β, standardized regression coefficient; SE, standard error of the parameter estimate.

The refined model explained 20.5% of the variance in fear (R² = 0.205), 22.0% of the variance in behavioral intention (R² = 0.220), and 3.9% of the variance in actual screening behavior (R² = 0.039).

4. Discussion and Conclusion

The present study revealed a pronounced intention–behavior gap in CRC screening, with PMT constructs explaining only 3.9% of the variance in actual screening behavior, despite accounting for 22.0% of the behavioral intention. This substantial discrepancy between intention formation and behavioral execution aligns with findings from Wuhan, China, where PMT constructs predicted 24% of screening intention, but actual behavior was not measured [9]. This suggests a fundamental limitation in translating cognitive appraisals into preventive action. The low uptake of screening (mean score 26.12/100) contrasts sharply with high self‐efficacy scores (69.43/100), potentially indicating social desirability bias in self‐reported measures—a methodological limitation noted in multiple PMT studies [37]. This bias could be mitigated through objective verification using electronic health records, as implemented in Kaiser Permanente's research programs [38].

Notably, self‐efficacy (β = 0.054, p = 0.396) and response costs (β = −0.011, p = 0.988) demonstrated nonsignificant relationships with behavioral intention, contradicting global PMT literature where these factors are typically strong predictors [39]. This divergence may reflect the contextual specificities of Iran's fragmented healthcare system, where structural barriers potentially overshadow individual confidence. Conversely, urban Chinese populations have shown self‐efficacy as a dominant predictor (β = 0.36, p < 0.01) [9], while studies from Saudi Arabia identify response costs—such as procedural discomfort and privacy concerns—as primary deterrents [40]. The prominence of fear (β = 0.155, p = 0.003) and response efficacy (β = 0.172, p = 0.003) as key drivers of intention in our study echoes qualitative findings from Saudi populations [40], where awareness of cancer severity motivated screening consideration despite procedural fears.

The observed positive effect of fear on behavioral intention (β = 0.155, p = 0.003) aligns with contemporary refinements of PMT, which posit that fear enhances motivation when individuals perceive high self‐efficacy and response efficacy; conditions present in our study [41]. This counters classic views of fear‐induced avoidance by demonstrating its adaptive potential in health contexts where threat appraisals are coupled with strong coping appraisals [42]. Specifically, in cancer screening behaviors, fear functions as a mediator that translates perceived susceptibility and severity into protective intention when cultural frameworks (e.g., viewing cancer as a “stealth enemy”) amplify the perceived value of vigilance [43, 44]. Thus, our findings reflect a context‐dependent manifestation of PMT principles rather than a theoretical paradox [39].

Methodologically, the application of PLS‐SEM was justified by the nonnormal data distribution and moderate sample size (N = 433), consistent with recommendations for predictive modeling in health behavior research [45]. This approach proved advantageous over covariance‐based SEM for our exploratory aims, as demonstrated in other CRC screening studies [1]. However, the exceedingly low explained variance for actual behavior (R² = 0.039) underscores potential measurement limitations in operationalizing this construct via self‐reported dichotomous items, highlighting the need for future studies to incorporate multimodal assessments, such as medical record verification [38].

The sociodemographic profile—predominantly female (61.9%) and low educational attainment (66.5% with less than a high school diploma)—reflects gendered healthcare engagement patterns and literacy barriers documented in Middle Eastern cancer screening literature [40]. These findings resonate with Saudi studies identifying female‐specific barriers [40] and mirror socioeconomic gradients observed in American Indian communities, where education level and trust in providers significantly influence screening participation [46]. Iran's opportunistic screening approach starkly contrasts with Korea's National Cancer Screening Program, which achieved a 45.5% colonoscopy uptake through systematic invitations and subsidized services [47]. This comparison suggests that environmental enablers are likely prerequisites for PMT constructs to effectively translate into action.

This study possesses several strengths. First, the employment of PLS‐SEM was appropriate given the focus on predictive accuracy rather than theory confirmation and considering the moderate sample size (n = 433). PLS‐SEM's robustness in managing non‐normal data distributions and complex latent variables is well established in health behavior research [31]. Second, the stratified sampling across 11 health districts in Neyshabur, Iran, enhanced the representativeness of the sample. The sample size exceeded the minimum requirement—based on the rule of 5–10 participants per questionnaire item—aligning with Kass and Tinsley's guidelines [33]. Such methodological rigor addresses gaps in applying PMT within underrepresented populations, contrasting with prior research predominantly conducted in Western or East Asian contexts [9]. The significant predictors—response efficacy (β = 0.172, p = 0.003) and perceived severity (β = 0.118, p = 0.030)—are consistent with global PMT findings. A culturally specific finding of note was the lack of significance of self‐efficacy, diverging from results in urban Chinese populations [9].

Empirical work across health behaviors suggests that contextual factors—such as access to care, health system navigation, social norms, and logistical barriers—can reshape threat and coping appraisals and, in turn, behavioral outcomes (e.g., cancer screening uptake) beyond individual cognitions alone. Integrating PMT with models that account for environmental and policy‐level influences has shown promise in improving explanatory power and informing multilevel interventions. For instance, frameworks that combine health behavior theory with social ecological considerations have yielded stronger associations with preventive behaviors and guided more effective implementation strategies in population health research published in high‐impact journals. Future work in this domain could operationalize an integrated model by adding measurable environmental/structural constructs (e.g., access to screening, reminder systems, transportation, and system navigation) to PMT, enabling more precise identification of leverage points for policy and programmatic interventions [48, 49, 50].

The large gap between intention and action (44.9 points) in this study can be explained by structural barriers in resource‐limited settings, especially in low‐ and middle‐income countries [51, 52, 53]. Structural barriers may overshadow individual psychological factors in preventing action [54]. PMT explains how intentions form but not how they lead to action [39]; the Health Action Process Approach suggests that planned actions (like action planning) help turn intentions into behavior [55]. This is supported by the weak link between self‐efficacy and behavioral intention in our data (β = 0.054). Psychological factors such as procrastination seem to interact with systemic barriers, indicating that individual‐focused interventions (e.g., planning tools) must be paired with structural changes to reduce barriers to action [56].

Like other studies, this study has limitations. The cross‐sectional design restricts causal inference regarding the relationships between PMT constructs and behaviors. Furthermore, the very low variance explained in actual screening behavior (R² = 0.039), despite a relatively higher explanation of intention (22.0%), illustrates a persistent intention‐behavior gap—a challenge frequently encountered in CRC screening research [9, 57]. Reliance on self‐reported, dichotomous behavioral measures may have introduced social desirability bias, explaining the discrepancy between high self‐efficacy scores and low actual screening rates. Another limitation of this study was measuring behavior as a dichotomous outcome (yes/no). It is suggested that future research consider objective or continuous criteria for screening behavior. This study focused on PMT; future research should compare PMT with alternative models, such as the health belief model (HBM).

Future research should employ longitudinal designs to clarify how PMT constructs influence behavior over time, as demonstrated in Korea's National Cancer Screening Program evaluations [47, 58]. Incorporating system‐level variables—such as healthcare accessibility and provider recommendations—may enhance predictive accuracy. Additionally, replicating PMT analyses across diverse populations, including rural and underserved groups, can improve generalizability. Tailored interventions are also critical; for example, community navigator programs to bolster self‐efficacy, modeled after successful initiatives within U.S. American Indian populations, could be adapted for the Iranian context. It is also valuable to compare PMT's utility with other frameworks, such as the HBM, particularly in regions with organized screening programs (e.g., the Netherlands), versus opportunistic systems [59].

Finally, evaluating the cost‐effectiveness of PMT‐based interventions—like educational campaigns versus mailed fecal immunochemical tests (FIT)—would inform policy choices and resource allocation. A comprehensive approach integrating behavioral theory with system‐level modifications holds promise for improving screening uptake globally.

5. Implications for Practice

Our findings support targeted messaging emphasizing perceived severity and response efficacy, which are significant predictors of intention. For instance, PMT‐based educational campaigns that highlight screening effectiveness successfully increased FIT participation in Iran, yielding an 18.6% uptake increase [10]. However, addressing the weak translation from intention to behavior necessitates multilevel strategies: individual‐level navigational support to overcome self‐efficacy barriers, system‐level interventions like mailed FIT kits [38], and community engagement to change cultural perceptions [46]. Iran can transform opportunistic screening into proactive prevention by embedding mailed FIT within Primary Healthcare Centers systems and budget allocation. Early pilots show 40%–71.7% participation is feasible, with colonoscopy access as the critical scalability determinant [60, 61, 62]. Future research that combines longitudinal tracking of multiple factors, utilizes objective behavioral assessments, and tests integrated models—including healthcare access variables—could better elucidate the complex interplay of determinants influencing CRC screening behaviors across diverse settings.

Author Contributions

Mahdi Gholian‐Aval: conceptualization, writing – review and editing. Bahareh Behrouzi: conceptualization, supervision, writing – original draft. Jamshid Jamali: data curation, writing – review and editing. All authors participated in drafting the manuscript and approved the final version.

Conflicts of Interest

The authors declare no conflicts of interest.

Transparency Statement

The lead author Jamshid Jamali affirms that this manuscript is an honest, accurate, and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned (and, if relevant, registered) have been explained.

Data Availability Statement

The data produced and examined in this study can be obtained from the corresponding author upon a reasonable request, contingent upon approval from the Mashhad Regional Committee for Medical Research Ethics.