Introduction

Nursing students are expected to graduate as lifelong learners who continuously integrate evolving knowledge into their practice, remain open to professional development, and deliver high-quality patient care [1]. Clinical learning environments are inherently complex and often expose students to significant stress as they engage with real-world patient care scenarios [2,3]. Consequently, students must consistently evaluate their performance, regulate their learning processes, and adapt to dynamic clinical settings [4,5]. This continuous cycle of self-assessment, learning management, and environmental adaptation is essential for navigating the challenges of clinical education [2,6].

Self-regulation is a fundamental competency for nursing students, encompassing the ability to control thoughts, emotions, and behaviors to achieve academic and professional success. This skill is particularly critical within the rigorous demands of nursing education, where students must balance intensive coursework with clinical responsibilities. Effective self-regulation not only enhances academic achievement but also improves patient care and personal well-being, underscoring its significance in nursing education frameworks [7,8].

The theoretical foundations of self-regulation in nursing education are rooted in various learning theories. Zimmerman’s self-regulated learning (SRL) theory highlights metacognition, motivation, and strategic action as essential components, conceptualizing SRL as a cyclical process involving forethought (planning and goal-setting), performance (monitoring and strategy implementation), and self-reflection (evaluation and adaptation) [9]. Pintrich’s model similarly emphasizes four phases of SRL: planning and activating, monitoring, control, and reflection, across cognitive, motivational/affective, behavioral, and contextual domains [10]. Panadero’s comprehensive review of SRL models further underscores SRL’s multidimensional nature, encompassing cognitive, metacognitive, motivational, behavioral, and increasingly social processes [11]. Additional frameworks, such as cognitive and behavioral learning theories, further elucidate how self-regulation can be cultivated through active engagement and deliberate skill development [12].

The Self-Regulated Learning Scale for Clinical Nursing Practice (SRLS-CNP) was specifically designed to capture selected motivational and strategic aspects of learning that are particularly relevant to clinical nursing education. The five factors of the scale map onto key components of SRL theory as follows: Intrinsic and Achievement Motivation align with the motivational phase of Pintrich’s model, reflecting the driving force for engagement in clinical learning; Synthesized Knowledge and Nursing Skills corresponds to cognitive strategy use; Multidimensional Thinking reflects metacognitive processes including critical reflection and perspective-taking; and Effort Control represents behavioral persistence in the face of challenges [3]. Importantly, the scale does not claim to capture the full breadth of SRL as defined in comprehensive theoretical models—such as the full cyclical processes of planning, monitoring, self-evaluation, and regulation based on feedback—but rather focuses on specific motivational and strategic dimensions that are particularly salient in the clinical learning context. This targeted approach is consistent with the scale’s original development by Iyama and Maeda (2018), who prioritized practical clinical relevance over comprehensive theoretical coverage [3].

The Self-Regulated Learning Scale for Clinical Nursing Practice (SRLS-CNP) has been developed and validated as a reliable instrument for assessing nursing students’ self-regulated learning in clinical settings. This scale incorporates subscales measuring motivation and learning strategies, providing a comprehensive evaluation of self-regulation in clinical education [13]. With strong reliability and validity, the SRLS-CNP facilitates the assessment of self-regulated learning across different academic years, offering valuable insights into students’ developmental progress throughout nursing education [3].

Nursing education has undergone significant transformations since 2019, driven by factors including the COVID-19 pandemic [14]. This crisis prompted a shift to online teaching methods and emphasized critical care [15], alongside advancements in technology [16] and an increased reliance on simulation and AI tools [17]. Furthermore, changes in students’ social routines and a move towards self-regulated learning have led to more flexible, interactive teaching approaches, making nursing education more adaptable to the evolving demands of healthcare [18].

In this context, nurse educators play a pivotal role in promoting SRL, grounded in self-regulation theory, by modeling and teaching strategic learning practices [19]. These methods empower students to take ownership of their education, enhancing skills such as time management, goal setting, and progress monitoring [20]. However, despite growing recognition of self-regulation’s importance in nursing education, a significant research gap remains regarding nursing students’ self-regulatory behaviors [21].

Understanding SRL is particularly important for nursing students, as it directly affects their academic performance, clinical competence, and professional growth [21,22]. A thorough literature review indicates that effective self-regulatory practices are linked to improved learning outcomes and enhanced adaptation to the challenges of nursing education [23,24]. Nevertheless, in the Iranian context, the lack of culturally relevant assessment tools hinders our understanding of students’ self-regulatory behaviors.

Existing scales for assessing self-regulated learning may fall short in addressing the unique cultural and educational challenges faced by Iranian nursing students. Adapting the SRLS-CNP scale, originally developed by Iyama and Maeda (2018) in Japan [3], can fill this critical gap. This 16-item scale comprises two subscales: Motivation (including intrinsic motivation and achievement motivation) and Learning Strategies (including synthesized knowledge and nursing skill, multidimensional thinking, and effort control).

By adapting the SRLS-CNP, we aim to create a reliable and culturally relevant tool that fits the specific context of Iranian nursing students. This adaptation is essential for developing targeted educational interventions and promoting a culture of self-regulation tailored to the unique challenges these students face. A rigorous psychometric evaluation of this adapted scale will determine its reliability and validity, enabling precise assessments and effective support to enhance self-regulatory skills in nursing education. Hence, this study aims to advance educational strategies in Iranian nursing by providing insights into self-regulated learning and its assessment, ultimately improving student outcomes and professional development.

Materials and methods

Psychometric testing

Face validity.

The face validity of the SRLS-CNP-P was assessed using a two-stage qualitative and quantitative approach. In the qualitative phase, 20 nursing students were purposively selected to evaluate each item’s clarity, relevance, and appropriateness to the target construct. Participants provided feedback on wording, comprehensibility, and structural coherence, leading to minor modifications to enhance precision and cultural adaptation. Qualitative feedback was systematically recorded and reviewed; all items were judged relevant and appropriate, with minor phrasing adjustments to improve naturalness in Persian (e.g., simplifying complex sentences, using more familiar clinical terminology).

For the quantitative assessment, the Item Impact Score (IIS) was calculated to determine the perceived significance of each item. The nursing students rated the Importance of each statement on a 5-point Likert scale (1 = not important to 5 = extremely important). The IIS was derived using the formula:

Frequency (%) reflected the proportion of respondents assigning a rating of 4 or 5, and Importance denoted the mean score of each item. All items with an IIS ≥ 1.5 were retained, confirming their suitability for inclusion in the final scale [31].

Content validity.

To ensure the relevance, clarity, and theoretical alignment of the SRLS-CNP-P, content validity was evaluated through qualitative expert review and quantitative indices. A panel of 10 experts was purposively selected based on the following criteria: holding a doctoral degree in nursing education, clinical nursing, or psychometrics; having at least 5 years of experience in nursing education or clinical practice; and expertise in instrument development or psychometric evaluation. The panel consisted of 5 nursing educators (with PhDs in Nursing Education), 3 clinical nursing specialists (with MSc in Nursing and >10 years of clinical experience), and 2 psychometricians. Their mean years of experience were 12.4 years (SD = 4.2). The experts assessed each item’s representativeness, clarity, and fit with self-regulated learning (SRL) in clinical nursing education. Their qualitative feedback guided refinements to wording and conceptual precision to better capture the intended construct [32]. Content comprehensiveness was assessed by asking experts whether the scale’s items adequately covered all important aspects of self-regulated learning in clinical nursing practice as defined by the theoretical framework [25]. Experts confirmed that the items comprehensively addressed the key domains (motivation and learning strategies), with no major content gaps identified. Minor concerns regarding item wording were resolved through iterative revisions.

For quantitative validation, two key indices were computed. The Item Content Validity Index (I-CVI) was determined by having experts rate each item’s relevance on a 4-point scale, ranging from 1 (not relevant) to 4 (highly relevant). The I-CVI score for each item was calculated as the proportion of experts assigning ratings of 3 or 4, and items scoring ≥ 0.78 were retained due to their strong relevance. Modified Kappa () statistics were also calculated for each item to adjust for potential chance agreement among raters. values exceeding 0.74 indicated excellent inter-expert consensus, further supporting the scale’s content validity [33]. Scale-level content validity was assessed using two indices: the average scale-level content validity index (S-CVI/Ave), calculated as the mean I-CVI across all items, with a recommended threshold of ≥ 0.90; and the universal agreement scale-level content validity index (S-CVI/UA), defined as the proportion of items rated as relevant by all experts, with an acceptable threshold of ≥ 0.80 [33].

Construct validity.

The assessment of construct validity began with an Exploratory Factor Analysis (EFA) using Maximum Likelihood Estimation (MLE) with Promax rotation, chosen to accommodate anticipated correlations among factors. The analysis applied multiple criteria for factor extraction, including eigenvalues greater than 1.0, a review of scree plot inflection points, and parallel analysis using 1000 random datasets generated with the same number of variables and sample size as the original data (n = 160, 16 items). In parallel analysis, factors are retained when eigenvalues from the actual data exceed the corresponding 95th percentile eigenvalues from randomly generated data, providing a more accurate and conservative approach to factor retention than the eigenvalue > 1.0 rule alone. The adequacy of the sample was confirmed through the Kaiser-Meyer-Olkin (KMO) measure and Bartlett’s test of sphericity. Items were retained when they demonstrated a primary factor loading of ≥ 0.50 and all secondary (cross) loadings were < 0.40. The factor correlation matrix resulting from the Promax rotation was also examined to assess the relationships among the extracted factors. Extraction converged within a maximum of 25 iterations, and the number of iterations required for convergence was recorded [34].

Given the ordinal nature of the 5-point Likert-scale items, we first evaluated the data’s distributional properties. We examined item-level descriptive statistics, including means, standard deviations, skewness, and kurtosis, to assess univariate distributions. All items demonstrated acceptable skewness (ranging from −0.58 to 0.33) and kurtosis (ranging from −0.52 to 0.42), well within the acceptable range of ±2 [35], supporting univariate normality. However, to account for the ordinal nature of the data and to ensure robust parameter estimation, we employed two complementary approaches for the CFA. First, we used Maximum Likelihood (ML) estimation with bootstrapping (1000 bootstrap samples) to obtain robust standard errors and confidence intervals, as ML with bootstrapping is relatively robust to mild violations of normality when sample sizes are adequate (n = 200) [36]. Second, we conducted a sensitivity analysis using the Weighted Least Squares Mean and Variance Adjusted (WLSMV) estimator, which is specifically designed for ordinal data and utilizes polychoric correlations [37,38]. In Confirmatory Factor Analysis (CFA), the hypothesized five-factor structure was tested using both ML estimation with 200 independent participants and WLSMV estimation as a sensitivity analysis, applying established thresholds for model evaluation: acceptable factor loadings ≥ 0.50, critical ratios (C.R.) > 1.96 (p < 0.05), and goodness-of-fit indices including absolute fit (χ²/df < 3 acceptable, < 2 excellent; RMSEA < 0.08 acceptable, < 0.05 excellent with 90% CI upper limit < 0.10), incremental fit (CFI > 0.90 acceptable, > 0.95 excellent; IFI > 0.90; TLI > 0.90), and parsimony indices (PGFI > 0.50; PCFI > 0.50) [39] for the ML estimation. For the WLSMV estimation, we evaluated model fit using the robust chi-square (χ²_robust), RMSEA, CFI, and TLI, with the same threshold criteria [38].

We defined the five-factor model according to its original structure [3], allowing for correlations among all factors. We did not include correlated residuals or make any post-hoc modifications because the initial model already showed an excellent fit. To assess the hierarchical structure, we tested four competing models: (1) a one-factor model; (2) a two-factor model (Motivation and Learning Strategies); (3) a first-order five-factor model (as hypothesized); and (4) a higher-order model with two second-order factors. We compared the models using chi-square difference tests and information criteria, including AIC and BIC [27].

Convergent and discriminant validity.

Convergent validity was assessed through two primary criteria: composite reliability (CR) values were calculated with acceptable thresholds set at > 0.70 for adequate reliability, while average variance extracted (AVE) values were evaluated against the > 0.50 benchmark, indicating sufficient item convergence on each latent construct [40]. For discriminant validity, three complementary approaches were employed: The Fornell-Larcker criterion requiring each construct’s AVE to exceed its squared correlations with other constructs, comparison of Average Shared Variance (ASV) and Maximum Shared Variance (MSV) against AVE values, and Heterotrait-Monotrait Ratio (HTMT) analysis with a conservative threshold of < 0.85 and bootstrapped 95% confidence intervals (5000 bootstrap samples) [41].

Reliability.

The internal consistency of the SRLS-CNP-P was assessed using multiple indices: Cronbach’s alpha (α), a classic measure of item interrelatedness; McDonald’s omega (), which accounts for factor structure and is less biased with non-tau-equivalent items; Coefficient H, reflecting the construct’s maximal reliability in a multidimensional scale; and the Greatest Lower Bound (GLB), providing the strictest estimate of reliability by considering shared variance. All values exceeded 0.70, confirming strong internal consistency [42–44]. For the two-item factors (Multidimensional Thinking and Effort Control), we additionally report inter-item correlations and Spearman-Brown coefficients [45,46]. Original developers [3] conceptualized the scale as comprising two broader subscales with five first-order factors and did not explicitly validate a single total score. Accordingly, we focus our interpretation on factor and subscale scores. Total-scale coefficients are provided for descriptive purposes only.

To assess the temporal stability of the SRLS-CNP-P, we calculated the Intraclass Correlation Coefficient (ICC) with a 95% confidence interval (CI) following a 2-week retest administered to 30 nursing students. For the test-retest reliability analysis, we selected a subsample of 30 participants using simple random sampling from the original 360 participants who completed the initial assessment. No significant differences were found between the retest subsample and the remaining participants in terms of age (t = 0.42, p = 0.675), gender (χ² = 0.18, p = 0.671), or academic year (χ² = 0.94, p = 0.625), indicating that the retest sample was representative of the overall study population. We calculated the ICC using a two-way mixed-effects model to assess absolute agreement, and reported single-measure ICC values. The ICC surpassed the recommended threshold of 0.75, indicating good to excellent test-retest reliability [47].

Results

Characteristics of the participants

The study included 360 participants, divided into samples for exploratory factor analysis (EFA, n = 160) and confirmatory factor analysis (CFA, n = 200). The overall sample had a mean age of 21.7 years (SD = 2.01) and an average GPA of 16.52/20 (SD = 3.35). The majority of participants were female (66.7%, n = 240), single (86.1%, n = 310), and distributed across academic years as follows: second year (33.3%, n = 120), third year (30.6%, n = 110), and fourth year (36.1%, n = 130) (Table 1).

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Demographic characteristics of the participants (n = 360).

https://doi.org/10.1371/journal.pone.0358790.t001

Item-level descriptive statistics, including means, standard deviations, skewness, kurtosis, and floor/ceiling effects, are presented in(S4 File). All items demonstrated acceptable normality with skewness values ranging from −0.58 to 0.33 and kurtosis values from −0.52 to 0.42, well within the acceptable range of ±2. No floor or ceiling effects were observed, as all item means fell within the mid-range of the 5-point Likert scale (2.42 to 3.81), and the percentage of responses at the extreme values (scores 1 and 5) ranged from 0.3% to 8.6% for floor and 1.9% to 11.4% for ceiling, all well below the 15% threshold.

Face validity

Participants evaluated the translated items’ clarity, relevance, and cultural appropriateness in the qualitative phase. Their feedback indicated that the items were clear and relevant to self-regulated learning in clinical contexts. However, they suggested minor linguistic adjustments to make the phrasing more natural in Persian. Examples of modifications included: changing “I confirm clinical nursing skills” to “I review and verify clinical nursing skills” for greater clarity; revising “I would like to develop my way of thinking” to “I seek to expand my thinking approach” to better reflect the reflective intent; and simplifying “I try to associate what I have experienced through clinical nursing practice with the knowledge I previously acquired” to “I connect my clinical experiences with my prior knowledge” for improved readability. These changes were made without altering the original meaning of the items.

The Item Impact Score was calculated for the quantitative assessment, with scores ranging from 2.5 to 3.2 (≥ 1.5). All items exceeded the threshold, confirming their suitability for the intended construct. No items were eliminated at this stage (Table 2).

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Results of the face and content validity of the SRLS-CNP-P.

https://doi.org/10.1371/journal.pone.0358790.t002

Construct validity

The results of EFA showed excellent factorability (KMO = 0.87, Bartlett’s test of sphericity (χ² = 1284.32, p < 0.001)), revealing five clear factors explaining 81.49% of the variance. All three retention methods supported a five-factor solution. The eigenvalue > 1.0 criterion identified five factors (eigenvalues: 3.117, 2.871, 2.512, 2.341, and 2.198; the sixth eigenvalue was 0.987). Scree plot inspection revealed a clear inflection point after the fifth factor, with the curve leveling off thereafter, consistent with a five-factor structure. Parallel analysis confirmed the retention of five factors, as the eigenvalues obtained from the actual data exceeded the 95th percentile random eigenvalues for the first five factors (actual eigenvalues: 3.117, 2.871, 2.512, 2.341, 2.198 vs. random 95th percentile eigenvalues: 1.562, 1.488, 1.421, 1.365, 1.298). For the sixth factor, the actual eigenvalue (0.987) fell below the random 95th percentile eigenvalue (1.245), supporting the five-factor solution. This convergence between all three retention methods strengthens confidence in the factor retention decision.

The pattern matrix from the Promax rotation demonstrated that all items loaded strongly on their respective factors, with primary loadings ranging from 0.75 to 0.85. All secondary loadings (cross-loadings) were below 0.40, confirming the clarity of the factor structure. Table 3 presents the complete pattern matrix, including primary loadings. The factor correlation matrix from the Promax rotation (Table 6) revealed moderate correlations among factors (ranging from 0.021 to 0.414), supporting the use of Promax rotation, which allows correlated factors. These correlations were consistent with those obtained from the CFA model, further confirming the interrelated yet distinct nature of the five factors. The extraction algorithm converged successfully after 8 iterations, indicating stable factor extraction. The structure showed good communalities (0.63–0.77) and conceptually coherent patterns, supporting the scale’s validity for measuring nursing students’ self-regulated learning in clinical settings (Table 3).

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. The results of the exploratory factor analysis of the SRLS-CNP-P (n = 160).

https://doi.org/10.1371/journal.pone.0358790.t003

The results of CFA demonstrated excellent model fit, with all factor loadings ranging from 0.66 to 0.96, indicating strong relationships between items and their respective latent constructs (Fig 1). The critical ratio values ranged from 5.698 (S.E. = 0.164) to 21.485 (S.E. = 0.042), demonstrating statistically significant factor loadings (p < 0.001) (Table 4). The model fit indices revealed χ² = 139.739 with 94 degrees of freedom (χ²/df = 1.487), RMSEA = 0.044 (90% CI: 0.035–0.053), SRMR = 0.041, GFI = 0.935, CFI = 0.980, TLI = 0.975, RFI = 0.928, and IFI = 0.981, all exceeding recommended thresholds for good model fit. The parsimony indices (PGFI = 0.646, PCFI = 0.768) and close-fit test (PCLOSE = 0.725) further confirmed the model’s appropriate balance between complexity and fit.

Download:

• PNG
  larger image
• TIFF
  original image

Table 4. Standardized factor loadings, standard errors, and critical ratios for the CFA model of the SRLS-CNP-P (n = 200).

https://doi.org/10.1371/journal.pone.0358790.t004

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. The final model of the first-order CFA of the SRLS-CNP-P (n = 200).

https://doi.org/10.1371/journal.pone.0358790.g001

To assess the robustness of our findings to the ordinal nature of the data, we conducted a sensitivity analysis using the WLSMV estimator, which is specifically designed for ordinal Likert-scale data and uses polychoric correlations. The WLSMV model also demonstrated excellent fit: χ²_robust = 152.847, df = 94, RMSEA = 0.048 (90% CI: 0.039–0.057), CFI = 0.976, and TLI = 0.971. All factor loadings remained statistically significant (p < 0.001) and were comparable to those obtained from the ML estimation (ranging from 0.64 to 0.95). The similarity between the ML and WLSMV results confirms that our findings are robust to the choice of estimator and supports the appropriateness of ML estimation with bootstrapping for this dataset.

The one-factor model showed poor fit (χ² = 689.432, df = 104, CFI = 0.812, RMSEA = 0.142, SRMR = 0.118). The two-factor model demonstrated suboptimal fit (χ² = 412.876, df = 103, CFI = 0.902, RMSEA = 0.098, SRMR = 0.087). The higher-order model showed acceptable fit (χ² = 152.654, df = 96, CFI = 0.976, RMSEA = 0.049, SRMR = 0.045). However, the five-factor first-order model exhibited good fit (Δχ² = 12.915, Δdf = 2, p < 0.01; AIC = 267.739 vs. 276.654; BIC = 384.162 vs. 396.877), confirming it as the best-fitting and most parsimonious model.

Convergent and discriminant validity

The SRLS-CNP-P demonstrated strong convergent validity, with all factors achieving CR values between 0.828 and 0.913, all exceeding the recommended threshold of 0.70. Additionally, the AVE ranged from 0.548 to 0.839, with all AVE values surpassing the 0.50 benchmark.

The SRLS-CNP-P demonstrated strong convergent validity, with all factors achieving CR values between 0.828 and 0.913, all exceeding the recommended threshold of 0.70. Additionally, the AVE ranged from 0.548 to 0.839, with all AVE values surpassing the 0.50 benchmark (Table 5). Discriminant validity was confirmed through multiple criteria (Tables 5 and 6). First, according to the Fornell-Larcker criterion, for each factor, the square root of AVE exceeded the correlations between that factor and all other factors (off-diagonal values), confirming that each construct shared more variance with its indicators than with other constructs. Second, for each factor, AVE exceeded both MSV and ASV, further supporting discriminant validity. Third, all HTMT ratios were below the 0.85 threshold, with values ranging from 0.023 to 0.456, and bootstrapped 95% confidence intervals did not contain 1.00 (Table 6). The highest HTMT value was observed between Intrinsic Motivation and Effort Control (0.456; 95% CI: 0.365–0.547), while the lowest was between Achievement Motivation and Multidimensional Thinking (0.023; 95% CI: 0.001–0.052). These findings provide strong evidence that all five factors are empirically distinct despite their theoretical interrelatedness within the self-regulated learning framework.

Download:

• PNG
  larger image
• TIFF
  original image

Table 5. Convergent validity and Fornell-Larcker criterion of the SRLS-CNP-P (n = 200).

https://doi.org/10.1371/journal.pone.0358790.t005

Download:

• PNG
  larger image
• TIFF
  original image

Table 6. Heterotrait-Monotrait Ratio (HTMT) and factor correlations of the SRLS-CNP-P (n = 200).

https://doi.org/10.1371/journal.pone.0358790.t006

Reliability

The SRLS-CNP-P demonstrated excellent internal consistency, with Cronbach’s alpha (α) ranging from 0.723 to 0.911, McDonald’s omega coefficient () between 0.731 and 0.920, Coefficient H values of 0.755–0.931, and Greatest Lower Bound (GLB) estimates of 0.762–0.943 across all factors. All values exceeded 0.70, confirming strong internal consistency. For the total SRLS-CNP-P, reliability coefficients were particularly strong (α = 0.911, = 0.920, H = 0.931, GLB = 0.943), confirming the instrument’s high internal consistency at both factor and SRLS-CNP-P levels. These total-scale coefficients are provided for descriptive purposes only, as the scale is multidimensional and the original developers did not validate a single total score. Interpretation should focus on the validated factor scores and subscale scores.

For the two-item factors, the inter-item correlations were r = 0.721 for Multidimensional Thinking and r = 0.578 for Effort Control, with Spearman-Brown coefficients of 0.838 and 0.733, respectively, indicating acceptable reliability for these brief factors.

Test-retest reliability was assessed in a subsample of 30 nursing students with a two-week interval. The total SRLS-CNP-P demonstrated excellent stability (ICC = 0.872, 95% CI: 0.803–0.947). Although the sample size for test-retest analysis was modest (n = 30), all ICC values exceeded the recommended threshold of 0.75, indicating good to excellent reliability. The confidence intervals estimate precision; while some intervals are relatively wide due to the limited sample size (e.g., total scale: 0.803–0.947), the lower bounds of all ICC confidence intervals remain above 0.70, supporting adequate stability (Table 7).

Download:

• PNG
  larger image
• TIFF
  original image

Table 7. The results of the reliability of the SRLS-CNP-P.

https://doi.org/10.1371/journal.pone.0358790.t007

Discussion

This study examined the psychometric properties of the SRLS-CNP-P among Iranian nursing students using a methodology explicitly aligned with COSMIN guidelines for systematic instrument validation [48]. The research evaluates the scale’s validity, reliability, and factor structure through COSMIN-recommended steps: (1) cross-cultural adaptation with linguistic equivalence testing (COSMIN domain 1: Content validity), (2) structural validity via EFA/CFA (Domain 2: Structural validity), and (3) comprehensive reliability testing (Domain 3: Internal consistency and reliability). The findings provide a scientific basis for more accurate evaluation of nursing students’ self-regulated learning abilities in real clinical settings within the study population. Practical applications may include identifying students’ strengths and weaknesses in learning management, designing targeted educational interventions to enhance self-regulation, and utilizing this validated tool in future nursing education research. For educators and administrators, this scale represents a promising instrument for developing professional training programs and improving the quality of clinical nursing education within similar contexts. Future research is needed to confirm its utility for advancing nursing students’ clinical competencies and lifelong learning skills. The study’s outcomes may offer insights for curriculum development and teaching methodologies in similar nursing education programs.

A further consideration concerns the normative framing of self-regulated learning throughout this manuscript. SRL is consistently presented as a desirable attribute associated with academic success, clinical competence, and professional development, which is common in the literature. However, this perspective risks transforming SRL from an analytical construct into a normative ideal. The discussion may imply that improved learning outcomes are primarily a function of individual self-regulation, while contextual factors—such as supervision quality, clinical learning environment characteristics, and institutional support—receive less attention. Research has demonstrated that SRL is shaped by contextual factors, including feedback from clinical educators, the structure of learning opportunities, and the prevailing educational culture [49,50]. A more balanced perspective acknowledges that self-regulation develops through dynamic interaction between individual and environmental influences [9].

Furthermore, several items in the scale (e.g., those emphasizing effort, persistence, independent practice, and achievement) may reflect broader expectations associated with being a “good” nursing student rather than self-regulated learning specifically. High scores could partly capture alignment with professional socialization norms or conformity with educational expectations rather than self-regulated learning alone. This is particularly relevant in clinical education, where students are implicitly and explicitly socialized into professional roles and behaviors [51]. While the scale assesses observable behaviors relevant to clinical learning, users should be aware that scores may reflect both genuine self-regulatory processes and the internalization of professional norms. Future research employing qualitative methods could help disentangle these influences and provide richer insight into what high scores represent in different educational contexts.

Another contextual consideration is the sample’s demographic composition, which was predominantly female (66.7%) and unmarried (86.1%). While this distribution reflects the demographic composition of nursing programs in Iran, where nursing remains a female-dominated profession, the manuscript does not consider whether gendered professional expectations, social roles, or broader cultural factors may shape SRL. Research suggests that gender-specific socialization processes, including expectations around caregiving, responsibility, and academic diligence, can influence self-regulation in educational settings [52,53]. In the Iranian cultural context, where nursing is strongly associated with feminine caregiving roles, female students may experience different motivational pressures and professional socialization compared to their male counterparts [54]. Similarly, marital status may influence students’ available time, social support, and competing responsibilities, potentially affecting their engagement in self-regulatory behaviors. While these demographic factors were not the focus of this validation study, future research should examine whether the SRLS-CNP-P functions equivalently across gender and marital status groups through measurement invariance testing, and explore how social and cultural contexts shape self-regulated learning in clinical nursing education.

An important distinction should be made between linguistic adaptation and cultural equivalence. While our translation procedure ensured semantic, idiomatic, and conceptual equivalence between the English and Persian versions, this does not fully establish that the underlying construct of self-regulated learning is interpreted identically across Japanese, English, and Persian cultural contexts. The original SRLS-CNP was developed in Japan [3], where clinical nursing education emphasizes hierarchical mentor-student relationships, collective learning, and implicit knowledge transmission. In contrast, Iranian nursing education operates within a different cultural framework, characterized by more direct educator-student interactions, explicit performance evaluation, and varying motivational orientations shaped by collectivist values and educational structures. These cultural differences may influence how students interpret items related to achievement motivation (e.g., “getting better grades than classmates”) or synthesized knowledge (e.g., “learning patient materials before practice”). Our findings demonstrate that the scale is understandable and psychometrically stable in the Iranian context; however, true cultural equivalence—the extent to which the construct of SRL carries the same meaning across cultures—remains to be fully established. Future cross-cultural studies employing measurement invariance testing across Japanese, Turkish, and Persian samples would provide stronger evidence of cultural comparability.

In our Persian adaptation, participants confirmed the clarity and relevance of items during face validity assessment, though minor linguistic adjustments were needed to enhance natural phrasing in Persian. The quantitative face validity assessment showed robust Item Impact Scores (2.5–3.2), all exceeding the 1.5 threshold, indicating comprehensibility, as per COSMIN’s emphasis on respondent feedback (Domain 1). Similarly, the Turkish version reported good readability and comprehension in its pilot test, though it did not provide specific quantitative face validity metrics. Both adapted versions showed excellent content validity results, with our Persian version exhibiting I-CVI values of 0.8–1.0 and modified Kappa statistics of 0.79–1.0, while the Turkish version reported CVI values of 0.81–1.0. Notably, while both adaptations retained all original items due to high content validity indices, our Persian version incorporated expert suggestions to refine item wording. In contrast, the Turkish version maintained the items verbatim. These differences likely reflect cultural-linguistic variations between Persian and English compared to Turkish and English, with Persian requiring more nuanced phrasing adjustments to maintain conceptual equivalence while ensuring natural expression. The methodological approach also differed: our study used qualitative feedback and quantitative metrics (including Kappa-adjusted CVIs), providing a more comprehensive validation process than the Turkish version’s [55] reliance on qualitative pilot testing and CVI values alone. These variations do not indicate the superiority of either approach; rather, they show how cultural context and linguistic distance from the original English version influence adaptation strategies. The excellent content validity in both adaptations confirms the fundamental cross-cultural applicability of the scale’s constructs. At the same time, the observed differences in implementation highlight the importance of tailored approaches to instrument adaptation based on target language characteristics and local clinical education contexts. These findings suggest that while the core concepts of self-regulated learning in clinical nursing practice are transferable across cultures, careful attention to linguistic nuances and local educational frameworks remains essential for optimal scale performance in different settings.

Our Persian version demonstrated strong psychometric properties, with exploratory factor analysis extracting five distinct factors with strong loadings (0.75–0.85) accounting for 81.49% of the variance, compared to the Turkish version’s [55] broader range of factor loadings (0.34–0.91). Confirmatory factor analysis showed excellent model fit in our Persian version (CFI = 0.980, RMSEA = 0.044, χ²/df = 1.487), outperforming both the original (CFI = 0.927, RMSEA = 0.066) [3] and Turkish versions (CFI = 0.99, RMSEA = 0.050). The Persian version showed stronger convergent validity (CR = 0.828–0.913; AVE = 0.548–0.839) and was the only version to establish discriminant validity comprehensively across multiple criteria (AVE > MSV/ASV, HTMT < 0.85). These differences in psychometric indices across the three versions should be interpreted with caution, as they may reflect variations in sample characteristics (e.g., age distribution, clinical experience levels), cultural and educational contexts (e.g., different clinical training environments, teaching methodologies), or methodological approaches rather than inherent superiority of any particular adaptation. For instance, the Persian version’s more rigorous cultural-linguistic adaptation process, including cognitive interviews and multiple revisions, may have contributed to stronger factor loadings and a clearer factor structure. Additionally, the application of more advanced psychometric analytical techniques (e.g., multiple reliability indices, comprehensive discriminant validity assessment) in the Persian validation may account for some of the observed differences in reported values. Despite these variations, the overall similarity in factor structure across all three versions confirms the scale’s conceptual robustness across cultures. These findings highlight the importance of considering cultural-linguistic differences and specific clinical education contexts when translating and adapting psychometric instruments.

It should be noted that the discriminant validity established in this study reflects factorial distinctiveness—demonstrating that the five factors are distinguishable from one another within the SRLS-CNP. However, this represents a relatively restricted form of discriminant validity, as all analyses are based on factors derived from the same instrument. The study did not assess whether the SRLS-CNP-P can be distinguished from related but distinct constructs such as self-efficacy, metacognitive regulation, or clinical reasoning ability. Including external measures of related constructs would have provided a more convincing test of construct validity by demonstrating that the scale captures self-regulated learning specifically, rather than overlapping with other theoretically adjacent concepts [56]. Future research should examine discriminant validity against external measures to more fully establish the scale’s construct validity.

An important psychometric observation is the notably high AVE (0.839) for the two-item Multidimensional Thinking factor. While AVE > 0.50 is desirable, values exceeding 0.80 may signal item redundancy [57]. This high AVE likely reflects that the factor has only two items, making the estimate sensitive to inter-item correlations, and that both items (Item 13: developing one’s way of thinking; Item 14: questioning other ways of thinking) share conceptual proximity as they both tap into metacognitive reflection. Nevertheless, the two items are theoretically distinct—Item 13 captures a forward-looking, developmental orientation, while Item 14 reflects a reflective, questioning orientation—consistent with the original framework where multidimensional thinking encompasses both generative and reflective cognitive processes [3]. The original developers also retained this two-item factor, confirming its theoretical importance.

Our Persian version demonstrated strong reliability coefficients, with Cronbach’s alpha ranging from 0.723 to 0.911, McDonald’s omega between 0.731 and 0.920, Coefficient H values of 0.755–0.931, and GLB estimates of 0.762–0.943. These results were comparable to the original version’s [3] reported Cronbach’s alpha of 0.853 for the total scale (with subscales 0.711–0.838) and the Turkish version’s [55] alpha coefficients (total scale 0.898; subscales 0.758–0.871). Test-retest reliability in our Persian version (ICC 0.783–0.872) was consistent with the original version’s correlation of 0.77 and similar to the Turkish version’s [55] 0.878 for the total scale. The observed coefficients in the Persian version were slightly higher in some instances compared to the original and Turkish versions, which may be attributed to several factors. Our cultural adaptation process, including cognitive interviews and multiple pilot tests, may have yielded clearer item phrasing and, consequently, higher internal consistency. The application of advanced reliability measures (omega, Coefficient H, GLB) that better account for scale dimensionality may also explain our more comprehensive estimates compared to studies relying solely on Cronbach’s alpha. However, these differences should not be interpreted as indicating the superiority of any particular version; rather, they likely reflect a combination of methodological, cultural, and sample-related factors that are inherent to cross-cultural validation research.

Conclusion

This study validated the SRLS-CNP-P among undergraduate nursing students at Guilan University of Medical Sciences, demonstrating strong psychometric properties, including excellent reliability, construct validity, and cross-cultural adaptation. The five-factor structure (intrinsic motivation, achievement motivation, synthesized knowledge, multidimensional thinking, and effort control) was confirmed, aligning with the original scale while ensuring cultural and linguistic appropriateness for this population. Furthermore, adherence to COSMIN guidelines and the rigorous translation and validation process enhance confidence in the scale’s applicability in this context. These findings highlight the SRLS-CNP-P as a reliable and valid tool for assessing nursing students’ self-regulated learning behaviors in this setting, offering educators and researchers a means to evaluate and enhance clinical training strategies. However, further validation studies across diverse Persian-speaking populations and educational settings are needed to establish broader generalizability.

Future research should expand on this validation study by examining the scale’s performance across diverse Persian-speaking populations and educational settings to strengthen generalizability. Longitudinal investigations could provide valuable insights into how self-regulated learning skills develop throughout nursing education and their long-term impact on clinical competency. Researchers might also explore the relationship between scale scores and objective clinical performance measures to establish criterion validity. Intervention studies utilizing the SRLS-CNP-P to evaluate educational programs that enhance self-regulated learning could yield important practical applications. Comparative studies with other validated versions of the scale could illuminate cultural influences on self-regulated learning in nursing education. The potential utility of this instrument for assessing practicing nurses’ self-regulated learning behaviors in continuing education contexts also warrants exploration. Finally, qualitative studies could complement these quantitative findings by better understanding the contextual factors influencing self-regulated learning in clinical nursing practice. These future directions would collectively help optimize nursing education and clinical training by improving knowledge and measurement of self-regulated learning processes.

Supporting information

Acknowledgments

We would like to sincerely thank all the nursing students who participated in this study. We sincerely thank the nursing faculty and clinical educators at Guilan University of Medical Sciences for their support and collaboration throughout the data collection process. We are especially grateful to the expert panel who played a vital role in the translation and validation process, ensuring the cultural and linguistic accuracy of the Persian version of the scale. Additionally, we would like to acknowledge the Research and Technology Deputy of Guilan University of Medical Sciences for their administrative and ethical oversight.