Psychometric properties of the Chinese version of Nurses’ Perception of Talent Management Scale(NPTMS-C): a structural equation modeling analysis

Abstract

Background

Nurse talent management is critical for workforce stability amid shortages and burnout. Scientifically assessing nurses’ perceptions of talent management strategies is significant for optimizing human resource management and enhancing retention.

Objective

This study translated the Nurse Perceived Talent Management Scale (NPTMS) into Chinese and evaluated the psychometric properties of the Chinese version (NPTMS-C) among clinical nurses from tertiary hospitals in Liaoning Province, China.

Methods

A convenience sample of clinical nurses was recruited from 12 tertiary hospitals. The 26-item NPTMS-C was evaluated through item analysis (normality testing, critical ratio, CITC), structural validity (EFA with promax rotation and CFA), convergent validity (AVE and CR), criterion-related validity (POSS as criterion), reliability (Cronbach’s α, McDonald’s ω, split-half, test-retest ICC), and measurement invariance (MG-CFA across gender).

Results

Item analysis retained all 26 items (critical ratio: 27.487–63.484, P < 0.001; CITC: 0.512–0.792). EFA yielded KMO = 0.989, Bartlett’s χ² = 15009.908 (P < 0.001), factor loadings = 0.610–0.893, explaining 55.480% of variance. CFA showed good fit: χ²/df = 1.300, RMSEA = 0.020, CFI = 0.993, TLI = 0.992, GFI = 0.966, NFI = 0.967, IFI = 0.970. Convergent validity: AVE = 0.500, CR = 0.954. Criterion correlation (NPTMS-C vs. POSS) = 0.631 (P < 0.001). Reliability: Cronbach’s α = 0.956, ω = 0.955, split-half = 0.902, test-retest ICC = 0.887. MG-CFA supported configural, metric, scalar, and strict invariance across gender, though caution is warranted due to sample gender imbalance.

Conclusion

The NPTMS-C demonstrates sound item quality, stable factor structure, and good reliability and validity among tertiary hospital nurses in Liaoning. It is a culturally adapted, scientifically rigorous tool suitable for nursing human resource management research and practice in similar Chinese clinical settings. Future studies with probability-based sampling across diverse regions and hospital levels are needed to confirm broader generalizability.

Clinical trial number

Not applicable.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12912-026-04972-7.

Introduction

Nurses constitute the core human resources of healthcare systems, with their quantity and quality directly influencing patient safety and the overall quality of healthcare services [1]. Nevertheless, nurse shortages have emerged as a global public health concern [2]. According to data published by the World Health Organization, the global nursing deficit exceeds 5.9 million and is projected to widen further over the coming decade [3]. In China, the accelerating transition toward an aging society, the sustained growth in healthcare demand, and the ongoing advancement of the “Healthy China 2030” initiative have rendered the need for high-quality nursing talent increasingly urgent [4]. Concurrently, persistently elevated nurse turnover rates and prominent occupational burnout have positioned the effective attraction, development, and retention of exceptional nursing professionals as a pressing core issue demanding resolution within the field of health administration [5].

Although no universally accepted definition of talent exists, it is generally characterized as encompassing systematically developed or acquired skills, high potential, exceptional performance, distinctive and rare qualities, inimitable core competencies, and strategic value [6]. In a study conducted by Haines [7], nursing talent was specifically defined as the capacity of nurses to apply their leadership qualities through the utilization of professional knowledge and clinical skills. Since Michaels et al. [8]. introduced the concept of the “war for talent” in 2001, talent management has progressively emerged as a prominent research focus within the field of organizational management. Existing evidence indicates that effective talent management strategies not only contribute to enhancing employee job satisfaction, organizational commitment, and retention intention, but also strengthen the core competitiveness of organizations, ultimately facilitating improvements in overall organizational performance [9]. Within the healthcare sector, the strategic value of talent management is equally noteworthy. Studies have indicated that healthcare institutions implementing systematic talent management demonstrate significantly lower nurse turnover rates compared to those without such strategies in place, and exhibit comparatively superior patient satisfaction and nursing quality indicators [10]. However, compared with fields such as business and education, research on nursing talent management remains relatively nascent, and its theoretical framework has yet to be fully established [6, 11]. The highly specialized nature of nursing work, its emotional labor intensity, and the particular demands of shift-based scheduling collectively determine that generic talent management theories and instruments cannot be directly transposed to nursing practice contexts [12]. From the perspective of employee perception, nurses’ subjective perceptions of the talent management practices implemented by their employing institutions tend to exert a more direct influence on their occupational behaviors and attitudes than the objective investments made by management. Existing research has confirmed significant associations between nurses’ perceived talent management and their work engagement, professional identity, and turnover intention, suggesting that the incorporation of nurses’ subjective perceptions into talent management evaluation frameworks carries considerable theoretical and practical value [13, 14].

Nevertheless, instruments for measuring nurses’ perceived talent management remain considerably limited at present [6, 15]. Existing studies have predominantly employed self-designed, non-standardized questionnaires or directly adapted generic scales from other industries, both of which exhibit notable limitations in terms of applicability and scientific rigor within nursing contexts. Although certain scales have been utilized in nursing management research, their item construction fails to adequately encompass the core dimensions of nursing talent management—including nursing-specific elements such as tiered training for clinical nurses [16], specialty competency development, and human resource allocation for night and rotating shifts—thereby rendering the validity of their measurement outcomes questionable [17]. Furthermore, the majority of existing instruments lack comprehensive psychometric validation reports, with insufficient reliability and validity data to meet the demands of scientific inquiry and evidence-based management. Consequently, the development or introduction of a talent management perception measurement instrument specifically tailored to the characteristics of nursing work and subjected to rigorous psychometric evaluation constitutes an essential prerequisite for advancing research progress in this domain.

To address the aforementioned research gaps, Duygu and colleagues developed the Nurse Perceived Talent Management Scale (NPTMS) in 2025 [6]. Grounded in theoretical frameworks including the resource-based view and the talent pipeline model, the scale was constructed through a systematic process encompassing literature review, expert consultation, pilot testing, and large-sample (n = 918) psychometric validation, ultimately yielding a unidimensional 26-item instrument. While talent management conceptually encompasses multiple domains (e.g. identification, development, and retention), the original developers demonstrated that these dimensions, as perceived by nurses, reflect a single coherent latent construct rather than distinct orthogonal factors. This is theoretically consistent with the resource-based view and talent pipeline model, which frame talent management as an integrated, unified organizational process. Findings demonstrated that the scale possesses satisfactory reliability and validity, establishing it as a measurement instrument of considerable scientific value within the nursing field. However, cultural adaptability represents an indispensable consideration in cross-cultural research that cannot be overlooked [18]. Substantial differences exist between China and Turkey with respect to cultural background, healthcare systems, nursing management models, and nurses’ professional identity, and whether the original scale can maintain its psychometric properties within Chinese nursing contexts remains to be verified.

Prior to undertaking the study, a review of existing domestic and international research employing structural equation modeling (SEM) to examine the psychometric properties and nomological relationships of talent management perception in nursing was conducted, with the aim of situating the current study within the broader empirical literature and identifying pertinent research gaps. Internationally, the original NPTMS development study by Duygu utilized both exploratory and confirmatory factor analysis within an SEM framework, demonstrating a unidimensional structure with acceptable model fit indice in a Turkish nursing sample [6]. Beyond scale development, SEM has been widely applied in the broader talent management literature to examine structural relationships between perceived talent management and key employee outcomes. For instance, studies using PLS-SEM have demonstrated that talent management exerts significant positive effects on work engagement and intention to stay, with perceived organizational support serving as a mediating mechanism [19]. Similarly, SEM-based investigations have confirmed that talent management practices positively predict organizational commitment while reducing turnover intention, with employee engagement functioning as a significant mediator in this pathway [20, 21]. Within nursing-specific contexts, a systematic review identified talent management as one of the significant predictors of nurse work engagement, though the majority of included studies relied on self-constructed, non-standardized measures rather than validated instruments with robust SEM-based psychometric evidence [22]. Domestically, SEM has been widely adopted in Chinese nursing research to validate measurement instruments and examine complex latent variable relationships—for example, in validating scales of nurses’ psychological capital, humanistic literacy, and work environment—yet no study has thus far applied SEM to examine the factor structure or construct validity of a nursing-specific talent management perception instrument within Chinese clinical settings [23, 24].

A closer examination of representative findings from existing studies illuminates both the theoretical and empirical contributions of talent management perception research in nursing, while also revealing critical gaps that justify the present study.

At the measurement level, Duygu et al. reported that the NPTMS demonstrated a Cronbach’s α of 0.97 and an average variance extracted (AVE) value of 0.51 in a Turkish nursing sample, confirming satisfactory internal consistency and convergent validity [6]. At the outcome level, Lerotholi and Bezuidenhout conducted a cross-sectional survey among professional nurses in South African public hospitals and found that talent acquisition significantly predicted nurses’ dedication and absorption (r = 0.112, p < 0.05), while management commitment was a significant predictor of absorption, underscoring the practical utility of talent management practices for nursing work engagement [25]. Taken together, these studies establish that perceived talent management is a meaningful predictor of key nursing outcomes across different national contexts.

Nevertheless, three interconnected gaps remain unresolved in the current literature, and the present study is specifically designed to address them. First, although the NPTMS has shown sound psychometric properties in Turkey, no validated, nursing-specific instrument for measuring talent management perception has been adapted or psychometrically evaluated within Chinese clinical nursing populations—leaving researchers and administrators without a culturally and linguistically appropriate tool for this context. This gap directly motivates our first objective: to perform a rigorous cross-cultural adaptation of the NPTMS for Chinese nurses. Second, while the NPTMS was originally proposed as a unidimensional scale, no study has systematically tested whether this factor structure holds across different cultural and healthcare system contexts, raising unresolved questions about measurement equivalence. This gap underpins our second objective: to evaluate the factorial validity of the adapted Chinese version through confirmatory factor analysis. Third, beyond psychometric properties, it remains unknown whether the NPTMS scores demonstrate criterion-related validity in predicting work engagement among Chinese nurses—a gap that informs our third objective: to test the predictive validity of the scale in a local clinical setting. By systematically addressing these gaps, this study not only advances cross-cultural measurement science in nursing management but also provides Chinese healthcare administrators and policymakers with a reliable, evidence-based tool to guide talent management improvements, ultimately contributing to enhanced nurse retention, workforce sustainability, and patient care quality within the Chinese healthcare system.

Methods

Translation procedure

The research team contacted Dr. Duygu Gül, the original developer of the NPTMS, via email and obtained her formal authorization to proceed. Forward translation was conducted in accordance with Brislin’s guidelines [26]: the NPTMS was independently translated into Chinese by a nursing expert proficient in English and a doctoral-level translator with fluency in English. Upon completion of the translations, both translators convened with the research team to compare the two versions, discuss discrepancies, and reconcile differences, thereby producing a preliminary Chinese version for subsequent back-translation. Back-translation was subsequently performed independently by a doctoral candidate in nursing science and a master’s degree holder in clinical medicine, neither of whom had prior exposure to the original scale. Discrepancies identified between the two back-translated versions were reviewed and resolved through discussion. The finalized back-translated version was then submitted to Dr. Duygu Gül for review and appraisal.

Cultural adaptation

To ensure the cultural appropriateness and measurement suitability of the scale within the Chinese nursing context [27], the research team invited nine experts from the fields of nursing management or clinical nursing to participate in content validity evaluation. All invited experts held associate senior professional titles or above, with years of professional experience ranging from 27 to 43 years; among them, two held doctoral degrees and seven held master’s degrees, collectively representing a breadth of professional expertise and extensive clinical experience. The expert evaluation encompassed three dimensions: conceptual equivalence (assessing whether each item captured the same underlying construct within the Chinese nursing context), content relevance (determining whether items were applicable to Chinese healthcare settings), and clarity of expression (ensuring linguistic appropriateness for the target population). This multidimensional evaluation ensured that item adaptation extended beyond wording adjustments to address deeper cultural and conceptual considerations. Following the consolidation of all expert feedback, the research team made targeted revisions to individual items based on the evaluation outcomes and modification recommendations, with particular attention to linguistic refinement of items characterized by ambiguous expression or semantic complexity, thereby enhancing the accuracy and readability of item wording.

Subsequently, a pilot test incorporating cognitive interviewing techniques was administered to 30 clinical nurses. Participants were invited to articulate their understanding of each item and identify any expressions they found ambiguous or culturally incongruent. This process revealed several items requiring further contextual refinement beyond linguistic accuracy. Multiple iterative rounds of revision were conducted based on participant feedback to progressively resolve interpretive ambiguities, ensuring semantic equivalence between the original and adapted versions. Upon completion of this systematic and standardized cross-cultural adaptation procedure, the Chinese version of the NPTMS (NPTMS-C) was formally established.

The theoretical frameworks underlying the NPTMS—the resource-based view and the talent pipeline model—have been extensively applied in Chinese nursing management research, providing a sound basis for theoretical equivalence in the cross-cultural adaptation. Additionally, the review and endorsement of the final back-translated version by the original developer, Dr. Duygu Gül, further confirmed that the conceptual integrity of the scale was maintained throughout the adaptation process.

Study design and setting

A large-sample cross-sectional study, utilizing convenience sampling to recruit 1,728 clinical nurses from 12 tertiary-level hospitals in Liaoning Province between August 2025 and March 2026. Inclusion criteria were as follows: possession of a valid Nurse Practitioner Certificate of the People’s Republic of China with legitimate licensure; active engagement in clinical nursing duties within hospital clinical departments during the survey period; a minimum of six months of continuous employment at the current institution, ensuring sufficient familiarity with and exposure to organizational talent management practices; and voluntary participation in the study with provision of written informed consent. Exclusion criteria comprised: nursing administrators at the head nurse level or above, given that their professional perspective differs fundamentally from that of frontline nurses and may introduce measurement bias; nurses on maternity leave, sick leave, secondment, or external training programs during the survey period; and student nurses or nurses undergoing standardized residency training.

Sample size estimation

Sample size was determined through the combined application of multiple calculation methods. For the EFA component, based on the Monte Carlo simulation findings of Wolf [28] and the 10:1 item-to-respondent ratio recommended by Mundfrom [29], a minimum effective sample size of 260 participants was required. For the CFA component, power analysis based on RMSEA as proposed by MacCallum [30]. (α = 0.05, 1 − β = 0.90, df = 300, RMSEA_H₀ = 0.10, RMSEA_H₁ = 0.05), verified using the semTools package in R, indicated a minimum effective sample size of 250 participants. As the total sample was randomly divided into equal halves for EFA and CFA respectively, a combined minimum effective sample of 510 participants was required; after incorporating a 15% redundancy allowance, the final target recruitment figure was set at no fewer than 600 participants. Detailed calculation procedures are provided in Supplementary Material 1. A total of 1,728 valid cases were ultimately collected, representing 2.88 times the theoretical minimum threshold. The use of a large sample contributes to enhanced precision in parameter estimation and greater stability of model fit indices [31].

Instruments

General Information Questionnaire A demographic questionnaire was developed to collect information on the following variables: sex, age, educational level, marital status, professional title, clinical department, and years of work experience.

Chinese Version of the Nurses’ Perception of Talant Management Scale (NPTMS-C): The NPTMS-C is a unidimensional scale comprising 26 items rated on a five-point Likert scale. Item responses range from 1 (strongly disagree) to 5 (strongly agree), with higher total scores indicating a higher level of perceived talent management.

Perceived Organizational Support Scale (POSS): The present study utilized the Nurse Perceived Organizational Support Scale revised by Zuo Hongmei in 2009 [32]. This scale comprises a total of 13 items encompassing two dimensions: emotional support and instrumental support. Each item is rated on a five-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree). Total scores range from 13 to 65, with higher scores indicating a stronger perception of organizational support among nurses. In the present study, the Cronbach’s α coefficient for this scale was 0.882, demonstrating satisfactory reliability and validity.

Data collection and quality control

Data were collected via the Wenjuan Xing online platform (https://www.wjx.cn/), a professional web-based survey system widely utilized in healthcare research in China that incorporates data encryption and storage, duplicate response prevention, and logical validation functions.

To ensure data quality, systematic quality control measures were implemented across three levels: questionnaire design, data collection, and data cleaning. At the questionnaire design level, all items were designated as mandatory to prevent missing data. At the data collection level, the Wenjuan Xing platform employed IP address recognition to restrict duplicate submissions from the same device, thereby ensuring that each respondent completed the questionnaire only once; additionally, a minimum completion time threshold of three minutes was established, with submissions falling below this threshold classified as invalid responses and subsequently excluded. At the data cleaning level, the research team conducted systematic screening of all retrieved data, applying the following exclusion criteria: (1) presence of patterned responding (e.g., identical option selection across all items); (2) abnormally brief completion times; and (3) missing data on key demographic variables. A total of 1,800 questionnaires were distributed; following the aforementioned quality control procedures, 1,728 valid questionnaires were retained for final analysis, yielding an effective response rate of 96.00%. The total sample was subsequently randomized into two subsamples: 860 cases (P1) for exploratory factor analysis and 868 cases (P2) for confirmatory factor analysis.

Statistical analysis

Statistical analyses were performed using IBM SPSS 27.0, AMOS 28.0 and R 4.5.1. Categorical data are presented as frequencies (n) and percentages (%). Quantitative data conforming to a normal distribution are expressed as mean ± standard deviation (Mean ± SD), while quantitative data not conforming to a normal distribution are presented as median and interquartile range (IQR). Data were considered normally distributed when skewness and kurtosis values fell within the range of − 2 to + 2 [33]. Statistical significance was defined as P < 0.05.

Item analysis

Critical ratio method

The retrieved scale responses were ranked in descending order of total score, with the upper 27% designated as the high-score group and the lower 27% as the low-score group. An independent samples t-test was conducted between the two groups. Items yielding a critical ratio value below 3 with a corresponding P > 0.05 were considered statistically non-significant, indicative of poor discriminability, and were flagged for potential deletion [34].

Correlation coefficient method

Normality testing was conducted on total scale scores and individual item scores. Pearson correlation was applied when normality assumptions were satisfied, while Spearman correlation was employed when data deviated from normality. Items with a correlation coefficient below 0.4 were considered for deletion [35].

Corrected item-total correlation (CITC)

Items with a corrected item-total correlation coefficient exceeding 0.3 were retained [36], while those falling below this threshold were considered for deletion.

Validity analysis

Content validity

Nine experts from relevant fields were invited to evaluate the items and content validity of the NPTMS-C using a four-point Likert rating scale, scored as follows: “not relevant” = 1, “somewhat relevant” = 2, “quite relevant” = 3, and “highly relevant” = 4. Both the item-level content validity index (I-CVI) and the scale-level content validity index (S-CVI) were calculated. The adopted evaluation criteria were I-CVI ≥ 0.780 and S-CVI ≥ 0.900 [37].

Structural validity

The suitability of factor analysis was assessed using the Kaiser-Meyer-Olkin (KMO) measure, with a value > 0.8 required, alongside Bartlett’s test of sphericity, with a significance threshold of P < 0.05 [38]. Exploratory factor analysis (EFA) was conducted using the Principal Axis Factoring (PAF) method combined with Promax rotation. PAF was selected for its capacity to extract common variance among variables, rendering it more appropriate for identifying underlying latent structures. Given the theoretical possibility of intercorrelations among scale dimensions, Promax oblique rotation was applied to yield a clearer factor structure. The number of factors to retain was determined based on eigenvalues greater than 1.0, in conjunction with examination of the scree plot. Factors with loadings exceeding 0.40 were retained [39]. Confirmatory factor analysis (CFA) was subsequently employed to evaluate overall model fit. Model fit was assessed based on the following indices: a chi-square to degrees of freedom ratio (CMIN/DF) of less than 3; a root mean square error of approximation (RMSEA) of less than 0.05; and values exceeding 0.9 for the goodness-of-fit index (GFI), normed fit index (NFI), incremental fit index (IFI), Tucker-Lewis index (TLI), and comparative fit index (CFI). Satisfaction of all aforementioned criteria was taken to indicate adequate model fit to the observed data [40].

Convergent validity

Composite reliability (CR) and average variance extracted (AVE) are commonly employed indices for evaluating convergent validity. It is generally accepted that when CR > 0.7 and AVE > 0.5 [41], the observed indicators of each latent variable converge effectively upon their corresponding latent construct, thereby indicating satisfactory convergent validity of the model.

Criterion-related validity

The Perceived Organizational Support Scale was selected as the criterion measure in the present study. According to organizational support theory [42], employees’ perceived organizational support derives from organizational management practices pertaining to resource provision, developmental opportunities, and emotional care [43]. The Nurse Perceived Talent Management Scale primarily reflects organizational practices in the areas of cultivation, motivation, and support within human resource management, which constitute important sources of organizational support [6]. Accordingly, a significant positive correlation between the two constructs is theoretically anticipated, rendering the POSS an appropriate instrument for criterion-related validity examination.

Reliability analysis

Cronbach’s α coefficient and test-retest reliability were selected to evaluate the internal consistency of the scale. An ideal measurement instrument requires an internal consistency reliability coefficient exceeding 0.7 for the total scale and exceeding 0.6 for each individual dimension [44]. Two weeks following the initial administration, 30 participants were randomly selected from the total sample for re-administration of the questionnaire. The intraclass correlation coefficient (ICC) was calculated to determine test-retest reliability, with a value exceeding 0.7 generally considered indicative of satisfactory temporal stability of the scale [45].

Measurement invariance testing

To evaluate the applicability and comparability of the scale across gender groups, the present study employed Multigroup Confirmatory Factor Analysis (MG-CFA) to systematically examine measurement invariance across male and female participants. The testing procedure followed a hierarchical sequence from less to more constrained models, in which four nested models were specified and compared: (1) the Configural Model, which constrained both groups to share the same factor structure without imposing any equality restrictions on parameters, serving as the baseline model for subsequent comparisons; (2) the Metric Model, which additionally constrained factor loadings to be equal across groups to examine whether each item reflected the latent construct to the same degree across groups; (3) the Scalar Model, which further constrained item intercepts to be equal across groups to examine whether male and female respondents shared the same baseline response levels on each item; and (4) the Strict Model, which additionally constrained residual variances to be equal across groups to examine the cross-group consistency of measurement error.

Overall model fit was evaluated using the CFI, the TLI, the RMSEA, and the SRMR. Measurement invariance at each level was determined based on changes in model fit indices, with ΔCFI, ΔTLI < 0.010 and ΔRMSEA < 0.015 serving as the criteria for supporting invariance [46, 47]. Given that the chi-square difference test (Δχ²) is highly sensitive to sample size under large-sample conditions and prone to erroneous conclusions, it was not adopted as the primary criterion in the present study. All analyses were conducted using the lavaan package in R 4.5.1.

Results

Translation and cross-cultural adaptation

Based on considerations of cultural appropriateness, linguistic conventions, and the principles of semantic and content equivalence within the Chinese nursing context, the following modifications were made to the scale items. “Talented nurses are made to feel valued.” was revised to “Talented nurses are made to feel that they are valued”. “Talented nurses are discovered through events for students and new graduates.” was revised to “Talented nurses are discovered or find or search through events for students and new graduates”.

Baseline characteristics of the study participants

A total of 1,800 questionnaires were distributed, of which 1,728 valid responses were retained, yielding an effective response rate of 96.00%. As shown in Table 1, the sample was predominantly female (94.8%), with the majority of participants falling within the 25ཞ34 age range (38.1%). Bachelor’s degree holders accounted for the largest proportion of educational backgrounds (63.9%). In terms of professional title, nurse practitioners represented the highest proportion (40.0%), followed by registered nurses (28.0%). The 10ཞ19 year work experience category was most prevalent (36.1%), indicating a sample with substantial clinical experience. Unmarried participants slightly outnumbered married participants (52.0% vs. 43.0%). Detailed information can be found in Table 1.

Table 1.

Baseline characteristics of the study participants

Variables		Frequencies	Percentages(%)
Gender	Females	1639	94.8
	Males	89	5.2
Age	< 25	166	9.6
	25ཞ34	658	38.1
	35ཞ44	510	29.5
	45ཞ54	280	16.2
	55ཞ59	98	5.7
	> 60	16	0.9
Educational level	Junior college diploma	553	32.0
	Bachelor’s degree	1105	63.9
	Master’s degree	70	4.1
Professional title	Registered Nurse	484	28.0
	Nurse Practitioner	692	40.0
	Senior Staff Nurse	466	27.0
	Associate Chief Nurse	52	3.0
	Chief Nurse	34	2.0
Work experience(year)	< 5	242	14.0
	5ཞ9	345	20.0
	10ཞ19	623	36.1
	20ཞ29	397	23.0
	> 30	121	7.0
Marital status	Married	743	43.0
	Unmarried	899	52.0
	Others	86	5.0

Item analysis

Item analysis was conducted on all 26 items of the NPTMS-C, yielding the following results (Table 2). To address potential item redundancy, the Cronbach’s α if item deleted was computed for all 26 items. The resulting values ranged from 0.953 to 0.956, none of which exceeded the overall scale Cronbach’s α of 0.956, indicating that the removal of any single item would not improve internal consistency. This finding provides direct evidence that no individual item was redundant or detrimental to the overall reliability of the scale, and that all 26 items contribute meaningfully to the measurement of the construct.

Table 2.

Results of item analysis for the NPTMS-C

Items	Skewness	Kurtosis	Critical Ratio	Pearson correlation coefficient	Corrected Item-Total Correlation	The Cronbach Alpha after the deletion items	P
Q1	0.172	-1.492	43.132	0.716	0.685	0.954	< 0.001
Q2	0.480	-1.113	43.256	0.739	0.714	0.954	< 0.001
Q3	0.083	-1.381	33.704	0.669	0.640	0.954	< 0.001
Q4	-0.263	-1.511	50.168	0.739	0.709	0.954	< 0.001
Q5	-0.158	-1.561	50.072	0.733	0.703	0.954	< 0.001
Q6	0.400	-1.378	35.825	0.664	0.630	0.954	< 0.001
Q7	-0.134	-1.368	38.553	0.676	0.644	0.954	< 0.001
Q8	0.510	-0.486	43.556	0.773	0.753	0.953	< 0.001
Q9	-0.529	-1.115	28.058	0.580	0.540	0.955	< 0.001
Q10	-0.805	-0.948	28.193	0.590	0.548	0.955	< 0.001
Q11	-0.166	-0.118	40.289	0.807	0.792	0.953	< 0.001
Q12	0.370	-1.329	33.681	0.648	0.613	0.955	< 0.001
Q13	-0.041	-0.848	31.244	0.648	0.621	0.954	< 0.001
Q14	0.208	-1.402	51.807	0.762	0.736	0.953	< 0.001
Q15	0.101	-1.591	27.487	0.556	0.512	0.956	< 0.001
Q16	0.020	-1.408	51.812	0.783	0.760	0.953	< 0.001
Q17	0.083	-1.471	30.249	0.595	0.556	0.955	< 0.001
Q18	0.129	-1.261	36.998	0.695	0.664	0.954	< 0.001
Q19	-0.233	-1.459	48.119	0.739	0.711	0.954	< 0.001
Q20	-0.020	-1.621	63.484	0.791	0.767	0.953	< 0.001
Q21	0.219	-1.273	39.885	0.695	0.664	0.954	< 0.001
Q22	-0.374	-1.476	47.790	0.708	0.675	0.954	< 0.001
Q23	0.182	-1.486	44.721	0.708	0.675	0.954	< 0.001
Q24	0.375	-1.183	42.685	0.731	0.704	0.954	< 0.001
Q25	-0.149	-0.796	32.598	0.671	0.643	0.954	< 0.001
Q26	0.395	-0.116	29.540	0.631	0.603	0.955	< 0.001

Normality testing

The absolute values of skewness for individual items ranged from 0.020 to 0.805, while the absolute values of kurtosis ranged from 0.116 to 1.621. All values fell within the accepted normality criterion of − 2 to + 2 for skewness and kurtosis [33], indicating that the distribution of each item approximated normality and satisfied the requirements for subsequent statistical analyses.

Critical ratio (CR) values

Discriminability of each item was examined using the extreme groups comparison method, with the upper and lower 27% of total scores designated as the high- and low-score groups, respectively. Results indicated that CR values for all 26 items ranged from 27.487 to 63.484, substantially exceeding the critical threshold of 3.00 [48], with all differences reaching statistical significance (P < 0.001). These findings demonstrate that each item possesses satisfactory discriminative capacity, effectively differentiating between high- and low-scoring respondents; accordingly, all items were retained.

Pearson correlation coefficients

Pearson correlation coefficients between individual item scores and the total scale score ranged from 0.556 to 0.807, all exceeding the retention criterion of 0.40 and achieving statistical significance (P < 0.001) [35]. These results indicate satisfactory consistency between each item and the overall scale, with all items adequately reflecting the global construct measured by the scale.

Corrected item-total correlation (CITC)

Corrected item-total correlation coefficients ranged from 0.512 to 0.792, all surpassing the minimum acceptable threshold of 0.30 [36]. These findings further confirm strong internal consistency between individual items and the overall scale, with no items warranting deletion.

Content validity

As presented in Table 3, I-CVI values for individual items ranged from 0.888 to 1.000. The majority of items achieved an I-CVI of 1.000, while a minority yielded values of 0.888. The overall S-CVI for the scale was 0.957. These findings indicate that all items received a high degree of endorsement from the expert panel, confirming that the items effectively represent the conceptual domain intended to be measured (Table 3).

Table 3.

Results of content validity for the NPTMS-C

Professors										Content validity
Items	P1	P2	P3	P4	P5	P6	P7	P8	P9	CVI	S-CVI
Q1	4	4	4	4	4	4	4	4	4	1.000	0.957
Q2	4	3	4	4	4	4	4	4	4	1.000
Q3	4	3	4	4	4	4	4	4	4	1.000
Q4	4	3	4	4	4	2	4	3	4	0.888
Q5	4	3	4	4	4	4	4	4	4	1.000
Q6	4	4	4	4	4	4	4	4	4	1.000
Q7	4	4	4	4	3	4	4	4	4	1.000
Q8	4	4	4	3	4	2	4	4	4	0.888
Q9	4	4	4	4	4	4	4	4	4	1.000
Q10	4	4	4	4	4	4	4	4	4	1.000
Q11	3	4	4	4	4	4	4	4	4	0.888
Q12	4	4	4	4	4	4	4	4	4	1.000
Q13	4	3	4	4	4	4	4	4	4	0.888
Q14	4	4	4	4	4	4	4	4	4	1.000
Q15	4	4	3	4	4	4	4	4	4	1.000
Q16	4	4	4	3	4	4	4	4	4	1.000
Q17	4	4	4	3	4	4	4	4	4	1.000
Q18	4	4	4	4	4	4	4	4	3	0.888
Q19	4	4	4	4	4	4	4	4	4	1.000
Q20	4	4	4	4	4	4	4	4	4	1.000
Q21	3	4	4	4	4	4	4	4	4	0.888
Q22	4	4	4	4	4	4	4	4	4	1.000
Q23	4	4	4	3	4	4	3	4	4	0.888
Q24	4	4	4	4	4	4	4	3	4	0.888
Q25	4	4	4	4	4	4	3	4	4	0.888
Q26	4	2	4	4	4	3	3	4	4	0.888

Exploratory factor analysis

Prior to conducting the EFA, the suitability of the data for factor analysis was evaluated. The KMO measure of sampling adequacy yielded a value of 0.989, which substantially exceeded the recommended threshold of 0.80, indicating excellent factorability of the correlation matrix. Bartlett’s test of sphericity was statistically significant (χ² = 15009.908, df = 325, P < 0.001), confirming that the inter-item correlation matrix was significantly different from an identity matrix and that the data were appropriate for factor analysis. EFA was performed using principal axis factoring with promax rotation. The number of factors to retain was determined based on eigenvalues greater than 1.0 and examination of the scree plot. The analysis revealed a clear single-factor structure (Fig. 1), with one factor extracted, explaining 55.480% of the total variance. Only one eigenvalue exceeded 1.0, and the scree plot showed a sharp drop after the first factor, with subsequent eigenvalues leveling off, further supporting the retention of a single factor. which meets the generally accepted criterion of ≥ 50% for scale development in the social and health sciences [49]. All 26 items demonstrated satisfactory factor loadings ranging from 0.610 to 0.893, surpassing the recommended minimum threshold of 0.40 [50]. No item exhibited cross-loading issues that would warrant removal (Table 4).

Fig. 1.

Scree plot of the NPTMS-C

Table 4.

Factor loading matrix from exploratory factor analysis of the NPTMS-C

Items	Factor 1
Q1	0.804
Q2	0.722
Q3	0.743
Q4	0.661
Q5	0.642
Q6	0.803
Q7	0.669
Q8	0.763
Q9	0.740
Q10	0.665
Q11	0.820
Q12	0.698
Q13	0.681
Q14	0.835
Q15	0.790
Q16	0.776
Q17	0.711
Q18	0.803
Q19	0.839
Q20	0.794
Q21	0.649
Q22	0.730
Q23	0.740
Q24	0.628
Q25	0.668
Q26	0.610

Note: Eigenvalues is 14.425; Cumulative variance is 55.480%

Confirmatory factor analysis

To further evaluate the factorial validity of the NPTMS-C, CFA was conducted using the structural equation modeling approach (Fig. 2). Multiple goodness-of-fit indices were examined to comprehensively assess the adequacy of the hypothesized factor structure, as no single index is sufficient to determine model fit. The results demonstrated that the proposed factor model achieved excellent fit to the observed data across all evaluated indices. The CMIN/DF = 1.300 was well below the recommended threshold of 3.0, indicating an acceptable ratio of model chi-square to degrees of freedom. The RMSEA = 0.020 fell substantially below the stringent cutoff of 0.08, suggesting a close fit between the hypothesized model and the population covariance matrix. With respect to the incremental fit indices, the CFI = 0.993, TLI = 0.992, NFI = 0.967, and IFI = 0.970 all exceeded the recommended criterion of 0.90. Additionally, the GFI = 0.966 similarly surpassed the 0.90 threshold. A summary of all fit indices and their corresponding thresholds is presented in Table 5.

Fig. 2.

Standardized structural equation model diagram from CFA of the NPTMS-C

Table 5.

Model fit indices from confirmatory factor analysis of the NPTMS-C

Index	Threshold	Value in the Present Study
CMIN/DF	< 3	1.300
RMSEA	< 0.08	0.020
CFI	> 0.9	0.993
GFI	> 0.9	0.966
TLI	> 0.9	0.992
NFI	> 0.9	0.967
IFI	> 0.9	0.970

Convergent validity

Convergent validity of the NPTMS-C was assessed by examining the standardized factor loadings, AVE and CR for each item within the confirmatory factor analysis framework. As presented in Table 6, all 26 items yielded statistically significant standardized factor loadings (P < 0.001), ranging from 0.454 to 0.858, with the majority of items exceeding the recommended threshold of 0.40. These results indicate that each item shared a substantial proportion of variance with its underlying latent construct, providing item-level evidence of convergent validity. At the construct level, the AVE value was 0.500, meeting the minimum acceptable criterion of ≥ 0.50 as recommended by Fornell and Larcker [51]. This suggests that, on average, the latent construct explained at least half of the variance in its observed indicators, demonstrating adequate convergent validity. Furthermore, the CR value was 0.954, considerably exceeding the recommended threshold of 0.70, which reflects a high degree of internal consistency and indicates that the items reliably and consistently measured the intended latent construct (Table 6).

Table 6.

AVE and CR values of the NPTMS-C

Itmes	Standardized Estimate	S.E.	AVE	CR	P
Q1	0.585	0.057	0.500	0.954	< 0.001
Q2	0.777	0.055			< 0.001
Q3	0.586	0.072			< 0.001
Q4	0.814	0.079			< 0.001
Q5	0.825	0.070			< 0.001
Q6	0.491	0.059			< 0.001
Q7	0.637	0.060			< 0.001
Q8	0.787	0.048			< 0.001
Q9	0.618	0.059			< 0.001
Q10	0.527	0.043			< 0.001
Q11	0.800	0.055			< 0.001
Q12	0.574	0.043			< 0.001
Q13	0.709	0.065			< 0.001
Q14	0.656	0.074			< 0.001
Q15	0.454	0.070			< 0.001
Q16	0.815	0.072			< 0.001
Q17	0.518	0.057			< 0.001
Q18	0.517	0.063			< 0.001
Q19	0.573	0.075			< 0.001
Q20	0.821	0.059			< 0.001
Q21	0.815	0.065			< 0.001
Q22	0.703	0.072			< 0.001
Q23	0.701	0.065			< 0.001
Q24	0.858	0.044			< 0.001
Q25	0.641	0.045			< 0.001
Q26	0.719	0.057			< 0.001

Criterion-related validity

The POSS was employed as the criterion measure to examine its correlation with the NPTMS-C. Results revealed a significant positive correlation between the total scores of the two scales (r = 0.631, P < 0.001).

Reliability

The reliability of the NPTMS-C was evaluated using four complementary indices: Cronbach’s α coefficient (overall reliability), McDonald’s Omega, split-half reliability, and test-retest reliability. The results are summarized in Table 7.The overall Cronbach’s α coefficient for the NPTMS-C was 0.956, substantially exceeding the widely accepted threshold of 0.70 [44], and indicative of excellent internal consistency among the 26 items. This finding suggests that the items of the scale consistently and cohesively measure the same underlying construct.

Table 7.

Reliability of the NPTMS-C

Reliability
Overall Reliability	McDonald’s Omega	Split-half Reliability	Test-retest Reliability
0.956	0.955	0.902	0.887

To supplement Cronbach’s α coefficient, which has been critiqued for its sensitivity to the number of items and assumptions of tau-equivalence, McDonald’s Omega was additionally computed. The Omega coefficient was 0.955, closely corroborating the Cronbach’s α coefficient estimate and providing further robust evidence of high internal consistency reliability, as it makes fewer restrictive assumptions about the measurement model.

Split-half reliability was assessed by randomly dividing the 26 items into two equal halves and computing the correlation between the two half-scores, with subsequent correction using the Spearman-Brown formula. The resulting split-half reliability coefficient was 0.902, exceeding the recommended minimum of 0.70, thereby confirming that the two halves of the scale yielded highly consistent measurement outcomes.

Test-retest reliability was examined by administering the NPTMS-C to a subsample of participants on two separate occasions, with an appropriate time interval between administrations. The intraclass correlation coefficient (ICC) for test-retest reliability was 0.887, surpassing the criterion of 0.70 for good temporal stability [45]. This result demonstrates that the scale produces stable and reproducible scores over time, supporting its suitability for longitudinal and repeated-measures research designs.

Measurement invariance testing

To examine the measurement equivalence of the scale across gender groups, the present study employed MG-CFA to sequentially test configural invariance, metric invariance, scalar invariance, and strict invariance. The model comparison results are presented in Table 8.

Table 8.

Measurement invariance results of the NPTMS-C across gender groups

Model	χ²	df	CFI	TLI	RMSEA	SRMR	ΔCFI	ΔTLI	ΔRMSEA
Configural	3302.52	598	0.896	0.887	0.072	0.057	-	-	-
Metric	3325.72	623	0.896	0.892	0.070	0.058	0.000	0.005	-0.002
Scalar	3343.43	648	0.897	0.896	0.069	0.058	0.001	0.004	-0.001
Strict	3369.92	674	0.897	0.901	0.068	0.058	0.000	0.005	-0.001
Partial Scalar	3337.03	642	0.897	0.895	0.069	0.058	0.000	0.001	-0.001

The configural model indicated that the same factor structure was specified in both male and female groups (CFI = 0.896, TLI = 0.887, RMSEA = 0.072, SRMR = 0.057). It should be noted that the CFI and TLI values fell marginally below the conventional threshold of 0.90, which may be partially attributable to the substantial gender imbalance in the present sample (females: n = 1,639; males: n = 89). Given the well-documented sensitivity of absolute fit indices to sample size inequality in multi-group models, these values are interpreted with caution, and configural invariance is considered only tentatively supported.

Building on the configural model, factor loadings were constrained to be equal across groups to test metric invariance. Changes in model fit indices were negligible (ΔCFI = 0.000, ΔTLI = 0.005, ΔRMSEA = − 0.002), all falling well within the established criteria (ΔCFI < 0.010, ΔRMSEA < 0.015), providing support for metric invariance. This suggests that the factor loadings of the scale items on the latent construct were equivalent across gender groups.

Scalar invariance was subsequently tested by additionally constraining item intercepts to be equal across groups. Changes in model fit remained within acceptable limits (ΔCFI = 0.000, ΔTLI = 0.005, ΔRMSEA = − 0.001), supporting scalar invariance. Finally, residual variances were constrained to be equal to test strict invariance, and model fit showed minimal change (ΔCFI < 0.001, ΔTLI = 0.001, ΔRMSEA = − 0.001), further supporting strict invariance.

Discussion

Translation and cross-cultural adaptation

During the cross-cultural adaptation process, two items underwent linguistic modification based on the recommendations of the expert panel and feedback collected during pilot testing, with the aim of enhancing conceptual clarity and cultural appropriateness within the Chinese nursing context. Item NPTMS11 was originally worded as “Talented nurses are made to feel valued.” Following expert review, this item was revised to “Talented nurses are made to feel that they are valued.” The rationale for this modification lies in the subtle yet meaningful difference in emphasis between the two formulations. In Chinese cultural and linguistic conventions, the explicit inclusion of a reference (i.e., “they”) serves to reinforce the subjective internalization of organizational recognition, aligning more closely with the Confucian-influenced value system in which an individual’s sense of self-worth is closely intertwined with perceived social and organizational affirmation [52]. Without the pronoun, Chinese respondents may interpret the original phrasing as referring to an externally observable outcome rather than a genuinely internalized psychological experience. The revised wording therefore more accurately captures the intended construct of perceived organizational valuation as an intrinsic psychological state, consistent with the theoretical underpinnings of organizational support theory and self-determination theory [53]. Item NPTMS18 was originally worded as “Talented nurses are discovered through events for students and new graduates.” During the cross-cultural adaptation process, Chinese-speaking registered nurses consistently interpreted this item as encompassing both a passive process of being recognized by others and an active process of independently seeking out and engaging with professional opportunities. Accordingly, the item was revised to “Talented nurses are discovered or find or search through events for students and new graduates,” indicating that the adapted wording explicitly captures both the passive and active dimensions of talent identification as understood within the Chinese nursing context.

This discrepancy reflects a culturally embedded understanding of professional talent development among nurses in tertiary-level hospitals in Liaoning Province, China. In the Western context from which the original instrument was developed, the passive construction “are discovered” implies an organizationally driven, top-down identification of talent, consistent with institutional mechanisms of recognition. In the Chinese cultural context, however, professional advancement is more commonly understood as a bidirectional process in which individuals are simultaneously subject to organizational selection and personally motivated to demonstrate competence. This is particularly evident among registered nurses in tertiary-level hospitals in Liaoning Province, China, who frequently participate in skill competitions, academic presentations, and professional development events not only as passive candidates for recognition, but as active agents seeking visibility and career advancement.

This adaptation followed established guidelines for cross-cultural instrument development, including the forward-backward translation protocol recommended by Brislin and the criteria for semantic, conceptual, and functional equivalence outlined in international scale adaptation frameworks. The revision was reviewed and approved by an expert panel, confirming its content validity and cultural appropriateness for use with registered nurses in tertiary-level hospitals in Liaoning Province, China.

Adequacy of item quality and scale structure

Item analysis results indicated that all 26 items of the NPTMS-C demonstrated satisfactory statistical properties. The universal significance of critical ratio values (CR values ranging from 27.487 to 63.484) reveals that there are significant individual differences in talent management perceptions among nurses in tertiary-level hospitals in Liaoning Province, China, and that these differences can be effectively captured and distinguished by the scale items. This finding resonates with conclusions from existing literature [20, 54], which suggest that nurses’ perceptions of organizational talent management strategies are not homogeneous in nature; rather, nurses at different hospital levels and at varying stages of career development tend to exhibit markedly distinct perceptual patterns. This implies that the NPTMS-C functions not merely as a measurement instrument, but also as a diagnostic tool for identifying weaknesses in nursing human resource management, enabling administrators to precisely locate structural differences in talent management perceptions across distinct nursing subgroups.

The internal consistency of items as reflected by high CITC values (0.512 to 0.792) corroborates, from a measurement theory perspective, the intrinsic conceptual coherence of nurses’ perceived talent management as a unified construct [55]. Nurses’ perceptions of talent management do not constitute fragmented and independent experiential elements. They represent an integrated evaluative judgment centered upon the core cognitive appraisal of “whether the organization values, nurtures, and effectively utilizes nursing talent” [56]. This is conceptually congruent with the logic of Social Exchange Theory [57], whereby nurses tend to interpret organizational talent management behaviors across recruitment, training, promotion, and incentivization as a holistic signal, upon which they form a comprehensive perceptual judgment of the organization, subsequently influencing their professional identity and retention intention [20]. The high degree of CITC consistency serves as a measurement-level reflection of this integrated perceptual structure.

Rationality of the factor structure

The stable factor structure revealed by both EFA and CFA demonstrates the construct validity of talent management perception as an operationalizable and measurable construct within the Chinese nursing context. This finding resonates cross-culturally with the theoretical framework of the original scale, but more importantly, it reflects the indigenous structural characteristics of talent management perception within the nursing organizational environment of tertiary-level hospitals in Liaoning Province, China.

Within the domestic healthcare system, nurses’ career development pathways, promotion mechanisms, and compensation structures are characterized by a high degree of institutionalization [58], such that nurses’ perceptions of talent management are substantially embedded within hospital hierarchical structures and administrative management logic. The stability of the factor structure indicates that, notwithstanding this institutional specificity, nurses are nonetheless capable of forming systematic and multidimensional perceptions of organizational talent management strategies, and that such perceptual structures possess inherent identifiability and cross-individual consistency. This provides structural-level assurance for future researchers seeking to employ the NPTMS-C for comparative analyses across diverse hospital settings.

Particularly noteworthy is the methodological choice of Promax rotation over orthogonal rotation in the EFA, a decision that carries substantive implications beyond mere statistical consideration. Oblique rotation permits intercorrelations among factors, thereby acknowledging the potential conceptual overlap among perceptions of different talent management dimensions [59]. This is consistent with the manner in which nurses perceive talent management in actual practice—nurses typically do not draw a sharp distinction between “being trained by the organization” and “being valued by the organization”, as these two perceptions tend to mutually reinforce one another psychologically [60]. Consequently, the intercorrelated factor structure revealed through oblique rotation represents not merely a statistically optimal choice, but a faithful reflection of the psychological reality underlying nurses’ perceptions of talent management [61]. The exceptional performance of CFA fit indices further indicates that a satisfactory level of correspondence has been achieved between the theoretical construct structure of the scale and the empirical measurement outcomes.

Multidimensional evidence of validity

The convergent validity and criterion-related validity findings of the NPTMS-C collectively delineate the external validity profile of the construct of nurses’ perceived talent management from complementary perspectives. The AVE value of 0.500 meets the minimum acceptable threshold established by Fornell and Larcker (1981), but only marginally so. This borderline result suggests that while the scale items share a meaningful proportion of variance with the latent construct, the level of convergent validity is modest rather than strong. It should be noted that some standardized factor loadings were also not particularly high, which is consistent with this pattern. We acknowledge that these findings do not constitute robust evidence of convergent validity, and interpretations should be confined accordingly. The borderline AVE may partly reflect the multidimensional nature of perceived talent management as a construct; however, we refrain from drawing strong theoretical conclusions from this observation, as the present cross-sectional validation study is not designed to disentangle the sources of measurement variance. Future research employing multilevel modeling or controlling for organizational-level variables may help clarify whether contextual factors such as departmental culture or supervisory style contribute to residual item variance.

In the criterion-related validity analysis, the significant positive correlation of 0.631 between the NPTMS-C and the POSS suggests that a substantive association exists between nurses’ positive perceptions of organizational talent management strategies and their self-efficacy and proactive coping capacities. From the perspective of Positive Organizational Behavior, when nurses perceive genuine organizational investment in talent management, such perceptions may activate the accumulation of personal resources and the development of psychological capital through the mediating mechanism of perceived organizational support, subsequently enhancing nurses’ capacity for self-regulation. This correlational finding not only validates the criterion-related validity of the scale, but also provides preliminary empirical evidence for understanding how nursing human resource management influences individual nurse development through psychological pathways, carrying considerable value for advancing subsequent theoretical construction.

It is worth noting that the CR value of 0.954, while notably high, is considered acceptable and theoretically interpretable within the context of the present study. This elevated value is attributable to the unidimensional structure of the scale, the domain-specific and professionally homogeneous nature of the items, and the relatively large number of items (n = 26), all of which collectively contribute to strong inter-item covariance. Importantly, the absence of item redundancy—as evidenced by the broad range of factor loadings (0.610–0.893)—suggests that the high CR value reflects genuine construct coherence rather than item duplication.

Robustness of reliability

The four reliability indices reported in the present study—Cronbach’s α coefficient = 0.956, McDonald’s Ω = 0.955, split-half reliability = 0.902, and test-retest ICC = 0.887 — are not only satisfactory in their numerical magnitude, but more importantly, their mutual corroboration and convergence reveal the inherent robustness of the reliability structure of the NPTMS-C. This carries particular practical significance for a measurement instrument intended for application within complex nursing management contexts.

The close correspondence between Cronbach’s α and McDonald’s Ω (differing by only 0.001) provides dual methodological verification of internal consistency estimates. It is noteworthy that the high Cronbach’s α value of 0.956 is likely partially attributable to the relatively large number of items (n = 26), as α is mathematically sensitive to scale length and tends to be elevated in longer instruments. However, this alone does not constitute evidence of item redundancy. Corrected item-total correlation coefficients ranged from 0.512 to 0.792, indicating that items contributed differentially to the total scale score with no evidence of ceiling-level redundancy. To further address this concern, McDonald’s ω (0.955) was examined as a complementary reliability index. Unlike Cronbach’s α, McDonald’s ω is grounded in a measurement model framework and is less susceptible to inflation due to item accumulation, thereby providing a more robust estimate of reliability. The high degree of concordance between α (0.956) and ω (0.955) therefore provides an unambiguous signal: the elevated internal consistency of the NPTMS-C is not an artifactual product of item accumulation, but rather a genuine reflection of the scale’s coherent and valid measurement of nurses’ perceived talent management. These findings collectively suggest that the high α value reflects authentic construct coherence rather than item duplication, providing a reliable foundation for the application of the NPTMS-C in clinical nursing management practice.

The test-retest ICC value of 0.887 represents one of the most practically valuable reliability findings of the present study. Perceived talent management is not a transient emotional reaction, but rather a relatively stable cognitive appraisal formed by nurses through prolonged occupational experience with organizational management behaviors [62]. The cross-temporal measurement stability revealed by the ICC value empirically substantiates the psychological nature of perceived talent management as a “stable trait-like cognition,” as opposed to a “state-like experience” subject to situational fluctuation [42]. Furthermore, the high test-retest reliability also reflects, to a certain extent, that the manner in which talent management strategies are currently implemented within nursing organizations has already generated a relatively clear and stable perceptual impression among the nursing workforce. This phenomenon may be characterized as a double-edged sword: on one hand, it attests to the measurement reliability of the scale; on the other hand, administrators should recognize that once nurses’ perceptions of talent management are formed, they acquire a degree of cognitive stickiness. Should an organization seek to alter nurses’ negative perceptions of talent management, sustained and systematic behavioral change in management practices — rather than one-time institutional adjustments—would be required to produce meaningful effect. This insight provides valuable theoretical reference for the future development of evidence-based nursing human resource intervention strategies.

Practical application and implications

The successful adaptation and rigorous validation of the NPTMS-C among clinical nurses from tertiary hospitals in Liaoning Province carries significant implications for the field of nursing management in similar clinical contexts.

The NPTMS-C provides nursing administrators and hospital human resource departments in tertiary healthcare settings with a quantitative assessment instrument capable of systematically diagnosing nurses’ perceived levels and degree of satisfaction with organizational talent management strategies. Against the backdrop of increasingly severe nurse shortages and occupational burnout, the precise identification of nurses’ perceptual differences across various talent management domains—including recruitment, training, promotion, and incentivization—holds direct instructional value for the formulation of targeted human resource intervention strategies and the enhancement of organizational attractiveness and retention rates in tertiary hospitals. The scale may additionally serve as a component of hospital management quality evaluation systems in comparable settings, furnishing quantitative evidence for the cultivation of nursing organizational culture and the monitoring of managerial effectiveness.

Furthermore, while the current findings are based on a regional sample and therefore limit immediate broad generalization, the NPTMS-C holds promise for supporting comparative research across hospitals and regions once its validity has been further established in more diverse samples. Pending such validation, the scale may provide preliminary data support for health administrative authorities seeking to understand the current state of nursing talent management in tertiary hospital contexts and evaluate the effectiveness of related policies. In particular, within the policy agenda of advancing high-quality nursing services and establishing a comprehensive nurse career development system, standardized instruments with sound psychometric properties—validated across diverse hospital levels, specialties, and geographic regions—constitute an essential prerequisite for evidence-based decision-making.

However, several important caveats should be acknowledged when interpreting Measurement Invariance Testing. First, the marked imbalance between the male (n = 89) and female (n = 1,639) subsamples may compromise the stability and generalizability of the MG-CFA estimates for the male group, as small sample sizes are known to inflate standard errors and reduce the power of model fit evaluation. Second, given that the configural model fit indices did not fully meet the conventional 0.90 threshold, the conclusions regarding higher-order invariance levels should be interpreted with appropriate caution. The present findings are therefore best regarded as preliminary evidence of cross-gender measurement consistency, and future studies with larger and more balanced gender samples are warranted to confirm and extend these results. Accordingly, while direct cross-gender latent mean comparisons using the NPTMS-C may be considered, researchers are advised to exercise caution in the interpretation of such comparisons given the aforementioned limitations.

Conclusion

In conclusion, the present study successfully introduced and rigorously validated the NPTMS-C among clinical nurses from tertiary hospitals in Liaoning Province, China. The comprehensive psychometric evaluation demonstrated that the NPTMS-C possesses satisfactory item quality, a stable and theoretically coherent factor structure, and robust reliability and validity across multiple complementary indices. The scale exhibited excellent internal consistency, satisfactory split-half reliability, and good temporal stability across measurement occasions. Structural validity was well-supported by both exploratory and confirmatory factor analyses, while convergent and criterion-related validity were adequately established. These findings collectively indicate that the NPTMS-C is a psychometrically sound and culturally adapted instrument capable of effectively measuring the perceptions of organizational talent management strategies among clinical nurses in tertiary hospital settings. The scale is recommended as a reliable and valid tool for nursing human resource management research and evidence-based managerial practice in comparable clinical contexts, and may serve as a meaningful foundation for future investigations into the antecedents, consequences, and moderating mechanisms of perceived talent management within tertiary hospital nursing populations. However, given that the current sample was drawn exclusively from tertiary hospitals in Liaoning Province using convenience sampling, the generalizability of these findings to nurses working in secondary or primary healthcare facilities, or in other geographic regions of China, remains to be established. Future research employing probability-based sampling strategies across diverse hospital levels, specialties, and regions is strongly encouraged to further validate the broader applicability of the NPTMS-C.

Limitation

Several limitations of the present study should be acknowledged. First, it should be noted that the present study employed only a single criterion measure for criterion-related validity assessment. Future research would benefit from incorporating multiple criterion variables—such as job satisfaction, organizational commitment, and turnover intention—in order to construct a more comprehensive network of validity evidence and thoroughly examine the breadth of external validity of the NPTMS-C within nursing contexts. Second, the present study employed convenience sampling to recruit participants from selected hospitals, which may limit the generalizability of the findings to the broader population of Chinese clinical nurses. Although the sample was drawn from hospitals of varying levels and encompassed nurses across diverse clinical departments, the non-random nature of the sampling strategy means that the extent to which the psychometric properties of the NPTMS-C can be generalized to nurses working in different regions, healthcare settings, or institutional contexts remains to be established. Future studies should consider employing stratified random sampling or multi-stage probability sampling across geographically diverse regions of China to enhance the representativeness of the sample and strengthen the external validity of the scale.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

CMIN/DF

Chi-square/degree of freedom

RMSEA

Root-mean-square error of approximation

CFI

Comparative fit index

TLI

Tucker lewis index

IFI

Incremental fit index

GFI

Goodness-of-fit index

S-CVI

Scale-level content validity index

I-CVI

Item-level content validity index

KMO

The Kaiser–Meyer–Olkin

EFA

Exploratory factor analysis

CFA

Confirmatory factor analysis

CR

Critical ration

AVE

Average variance extracted

Author contributions

QZ: Manuscript writing, data collection, data analysis, project management.CL: Data collection, data analysis, project managementJZ: Manuscript revision, data collection, data analysis, project management.

Funding

This study was not funded by any grant.

Data availability

Data supporting the conclusions of this study are available from the first author upon reasonable request.

Declarations

Ethical approval

This study was conducted in accordance with the principles of the Declaration of Helsinki and was approved by the Ethics Committee of The First Affiliated Hospital of Jinzhou Medical University (KYLL2025333) to participation, all individuals provided written informed consent. Furthermore, all data collected were kept strictly confidential and anonymized, and were used solely for research purposes.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.