Latent profile analysis of clinical reasoning and its relationship with patient safety perception among nursing interns: a multicenter cross-sectional study

Abstract

Background

Clinical reasoning competency is essential for patient safety. Nursing interns, in transition from students to professionals, are particularly vulnerable to errors. This study identified distinct profiles of clinical reasoning among nursing interns, described their characteristics, and examined factors associated with profile membership.

Methods

A multicenter cross-sectional study was conducted using convenience sampling. An online survey was administered to 1,185 nursing interns from 15 institutions across China. Latent Profile Analysis was used to categorize interns based on their scores on the four subscales of the Self-Assessment of Clinical Reasoning and Reflection scale and the total score of the General Self-Efficacy scale. The latter was included as a key psychological component integral to the development and manifestation of clinical reasoning competency. Differences across profiles were analyzed using ANOVA, chi-square tests, Kruskal-Wallis H test and multinomial logistic regression to explore associated sociodemographic and perceptual variables.

Results

Using latent profile analysis based on self-report scales, this study identified three distinct competency profiles. The “foundational competency group” (24.6%) scored lowest on clinical reasoning subscales and self-efficacy, yet showed a modest positive inclination towards reflection. The “advanced competency group” (14.9%) demonstrated the highest scores in clinical reasoning and self-efficacy but had relatively lower reflective scores. The “transitional competency group” (60.5%) exhibited moderate and balanced scores across all competencies. Multivariable regression analysis revealed that gender, educational level, choosing the field out of interest, willingness to participate in future training, and patient safety perception were significantly associated with profile membership among the interns (p < 0.05). A higher level of patient safety perception was strongly associated with higher odds of belonging to the “advanced competency group” (OR = 1.364, 95% CI: 1.321–1.409).

Conclusions

Clinical reasoning competency among nursing interns demonstrates heterogeneity, with three distinct profiles. Patient safety perception is strongly associated with advanced competency. These findings provide a basis for understanding competency diversity and suggest that tailored educational strategies may support the development of clinical reasoning. Due to the cross-sectional design and self-reported data, findings represent associations, not causal relationships.

Trial registration

Not applicable.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12912-026-04523-0.

Background

Globally, approximately one in ten hospitalized patients experiences harm due to unsafe care [1]. In China, patient safety remains a national policy priority amid healthcare expansion and rising public expectations [2].Central to safe and effective practice is clinical reasoning competency, particularly among nurses, whose decisions directly influence patient outcomes [3–6]. Nursing interns, who enter a ten-month clinical placement after completing theoretical coursework [7], represent a key group during this transitional phase of professional development [8, 9]. However, due to their still-developing clinical experience and competence, they are also considered at higher risk for errors and adverse events [9–11]. Therefore, examining the developmental profiles and structural characteristics of clinical reasoning among nursing interns is crucial for improving education and strengthening safety in clinical practice.

Clinical reasoning in nursing students can be described as a holistic and recursive cognitive process, which emphasizes understanding patient problems, making decisions, and synthesizing knowledge within specific clinical contexts [3]. Research by Hunter and Arthur notes that nursing students’ clinical reasoning is continuously and dynamically influenced by external factors, including socio-cultural context and clinical situational characteristics [12]. Subsequent studies have further revealed that the development of clinical reasoning competency is shaped by both intrinsic and extrinsic factors. Regarding intrinsic factors, beyond demographic characteristics such as age and gender, cognitive and metacognitive abilities, motivation, and problem-solving skills critically impact reasoning effectiveness [13, 14]. Specifically, emotional states such as excessive anxiety can disrupt logical analysis and compromise reasoning quality [15]. Concerning extrinsic factors, the quality of clinical instructor guidance, accessibility of resources in nursing practice, and the timeliness and consistency of feedback profoundly influence the development of clinical reasoning competency by shaping learning experiences and the efficiency of knowledge acquisition [16–18].

Notably, clinical reasoning encompasses not only cognitive skills but also psychological attributes, with self-efficacy representing a key element rooted in Bandura’s social cognitive theory as the belief in one’s capacity to organize and execute actions in specific situations [19]. Self-efficacy influences knowledge activation, hypothesis generation, decision-making, and persistence amid uncertainty [19]. Research indicates it is embedded within the self-regulatory cycle of clinical reasoning, shaping how learners retrieve and integrate information, apply reflective strategies, and engage with clinical complexity [20]. Hence, self-efficacy constitutes an integral psycho-motivational component of clinical reasoning, rather than merely a distal predictor.

Most current research on clinical reasoning competency adopts a variable-centered, quantitative approach, treating it as a homogeneous construct measured and compared via a single total score. This approach overlooks the potential existence of heterogeneous subgroups within the intern population. Furthermore, although the theoretical link between clinical reasoning and patient safety is widely acknowledged, research specifically examining the association between clinical reasoning competency and patient safety perception among nursing interns remains relatively scarce. The emerging person-centered analytical method, Latent Profile Analysis (LPA), offers a robust tool for uncovering such within-group competency differences.

This approach can identify subgroups of individuals with similar clinical reasoning characteristic patterns based on their responses across multiple indicators [21, 22].

To address these gaps, this multicenter cross-sectional study employs LPA. Grounded in the theoretical interaction between cognitive skills and motivational beliefs in clinical reasoning, it utilizes combined scores from the Self-Assessment of Clinical Reasoning and Reflection scale and the General Self-Efficacy Scale to holistically delineate distinct competency profiles among nursing interns. All measures are self-reported cross-sectional data, so the study examines associations, not causation. The study aims to: (1) identify clinical reasoning profiles using LPA; (2) analyze dimensional and influencing factor differences across profiles; (3) explore variations in patient safety perception among them to clarify their relationship with clinical reasoning characteristics. The findings aim to provide educators and managers with a nuanced perspective to inform tailored clinical teaching and safety strategies.

Methods

Aims

This study aimed to address the following questions:

1. What are the categories of clinical reasoning competency among nursing interns?

2. What are the characteristics of interns across different clinical reasoning competency categories?

3. Does patient safety perception influence the clinical reasoning competency of nursing interns?

Design

This was a multicenter cross-sectional study. The design and reporting of this study adhered to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.

Participants

From November to December 2025, nursing interns from 15 nursing colleges across four provinces (Hunan, Guangdong, Jiangxi, and Hainan) in southern China were selected as survey participants using convenience sampling. The selection of these colleges was based primarily on established academic and collaborative relationships, which facilitated efficient questionnaire distribution and improved response rates. A systematic survey was conducted to assess the clinical reasoning competency levels of nursing students during their internship. Inclusion criteria were as follows: (1) aged 18 years or older; (2) full-time nursing students; (3) provision of informed consent and voluntary participation. Exclusion criteria included: (1) continuous absence from clinical practice for ≥ 14 days due to illness or personal reasons; (2) unwilling to participate in the study.

Sample size

Determining the sample size for LPA requires a balance between statistical power and model complexity. Typically, a minimum of 500 participants is recommended to ensure robust model estimation and obtain adequate statistical power for identifying meaningful latent class structures. Furthermore, larger sample sizes generally provide greater statistical power to detect the true number of latent classes [23]. Therefore, the sample size of 1185 in this study meets this requirement.

Instruments

Demographics characteristics

Drawing on previous research [13], we designed a demographic questionnaire comprising three sections. The first section collected participants’ basic demographic information, including gender, family residence, whether they were the only child, educational level, student leader experience and career interests. The second section focused on training experiences, including participation in patient safety training, clinical reasoning training, or gamified teaching. The third section investigated preferences regarding gamified patient safety training, covering willingness to participate in such training in the future, ideal training duration and frequency, and the preferred internship stage for receiving the training.

Self-assessment of clinical reasoning and reflection

The original English version of the Self-Assessment of Clinical Reasoning and Reflection (SACRR) scale, developed based on Roth’s theory of reasoning, has been widely used to assess clinical reasoning competency among healthcare professionals [24]. This study employed the Chinese version translated and adapted by Yu et al. [25]. The Chinese scale consists of 26 items covering four dimensions: Information Systematization (12 items), Problem Analysis (9 items), Truth-seeking (4 items), and Reflection (1 item). All items are rated on a 5-point Likert scale ranging from 0 (strongly disagree) to 4 (strongly agree). The total score ranges from 0 to 104, with higher scores indicating a stronger self-assessed clinical reasoning competency. The questionnaire demonstrates good reliability and validity, with a content validity index of 0.98 and a Cronbach’s alpha of 0.754. In the present study, the overall Cronbach’s alpha for the scale was 0.967.

Generalized self-efficacy scale

The Generalized Self-Efficacy Scale (GSE) was originally developed by Schwarzer and Jerusalem [26] to assess an individual’s overall belief in their ability to cope with various challenging situations, and was later adapted into Chinese by Zhang and Schwarzer [27]. The scale consists of a single dimension with 10 items, scored on a 4-point Likert scale ranging from 1 (strongly disagree) to 4 (strongly agree). Total scores range from 10 to 40, with higher scores indicating greater self-efficacy. The Chinese version has demonstrated high reliability, with a previously reported Cronbach’s alpha of 0.91 [28]. In this study, the scale showed excellent internal consistency, achieving a Cronbach’s alpha of 0.934.

The health professional education in patient safety survey

The Health Professional Education in Patient Safety Survey (H-PEPSS) was originally developed by Ginsburg et al. [29] based on the patient safety framework of WHO and related international professional guidelines, aiming to assess clinical nursing interns’ perceptions of their patient safety knowledge and competence. The Chinese version consists of 17 items across two dimensions: self-perceptions (7 items) and organizational environment perceptions (10 items). Each item is rated on a 5-point Likert scale from 1 (strongly disagree) to 5 (strongly agree). The total score ranges from 17 to 85, with higher scores indicating a stronger perceived level and need for patient safety knowledge and competence. The scale demonstrates excellent reliability, with an overall Cronbach’s alpha of 0.94 and a content validity index of 0.903, indicating strong content validity [30]. In this study, the internal consistency reliability of the scale was tested, yielding a Cronbach’s alpha of 0.983.

Data collection

Data were collected using “wenjuanxing” (www.wenjuanxing.cn), a widely used online survey platform in China. With the assistance of participating nursing school internship tutors, liaison personnel were appointed and trained to distribute the questionnaire via WeChat to eligible students. Prior to the survey, an informed consent page was provided, and the survey was conducted anonymously. It required approximately 10 min to complete and did not involve any sensitive information. To prevent duplicate submissions, each IP address, WeChat account, and device was restricted to one response. All items were mandatory, and incomplete questionnaires could not be submitted. This helped ensure data integrity. Among the 1,380 recruited participants, 1,185 valid questionnaires were recovered after excluding responses with a response time of less than 120 s and those that failed the attention check, yielding a response rate of 85.87% (Figure S1). No significant differences in demographic characteristics were found between the exclusion group and the inclusion group (all p > 0.05), indicating that the exclusion process did not introduce systematic bias.

Statistical analyses

Given that the scale scores were continuous variables, LPA was conducted using Mplus 8.3 software to classify nursing interns into distinct subgroups based on their clinical reasoning competency. Model fit was evaluated to determine the optimal number of classes, employing the following three types of fit indices: (1) Information criteria—Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), and adjusted Bayesian Information Criterion (aBIC). Lower values indicate a better model fit. (2) Classification accuracy, assessed by the entropy value. A value exceeding 0.80 reflects a high classification accuracy [31]. (3) Likelihood ratio tests—the Lo-Mendell-Rubin adjusted likelihood ratio test (LMR) and the Bootstrap Likelihood Ratio Test (BLRT). A significant p-value (p < 0.05) suggests that the model with k classes fits significantly better than the model with k-1 classes [32]; (4) Average posterior probabilities for most likely class membership (AvePP), which reflect the mean posterior probability of individuals being assigned to their most likely class; higher values indicate clearer class separation and greater certainty of classification. AvePP values ≥ 0.70 for each class are commonly considered indicative of acceptable classification quality [33]. To ensure the robustness of the results, we re-estimated various models using different starting values and random seeds, and compared model fit using the Bootstrap likelihood ratio test.

Data entry and analysis were performed using SPSS 28.0. No variable was missing in the final data included in the analysis. Normality was verified by examining skewness and kurtosis values. Skewness < 2 and kurtosis < 7 are considered indicative of a normal distribution. All key variables in this study met this criterion [34]. Descriptive statistics are presented as counts and percentages for categorical variables, and as mean and standard deviation (SD) for continuous variables. Between-group comparisons were performed using the chi-square test, ANOVA, or the Kruskal-Wallis H test. Multinomial logistic regression analysis was conducted to explore the associations between clinical reasoning competency and independent variables. To account for potential within-institution correlation, cluster-robust standard errors were calculated at the institutional level using Stata 18.0, as conventional standard errors may yield invalid statistical inference under clustering [35]. A p-value < 0.05 was considered statistically significant.

Quality control

To ensure the quality and completeness of the questionnaire data, this study implemented several strict quality control measures. First, data entry was conducted independently by two researchers, followed by a cross-checking procedure to verify accuracy. Second, IP address and device restrictions were set to prevent the same respondent from submitting repeatedly. Additionally, two attention-check questions were embedded within the questionnaire, explicitly instructing participants to select the options for “Self-efficacy” (near the midpoint) and “Green” (the final item) respectively. This was designed to identify and exclude samples from inattentive or careless respondents [36]. A total of 183 respondents were excluded based on failure to pass these attention checks. Demographic comparisons showed no statistically significant differences between these excluded participants and the retained sample (p > 0.05).

Results

Participant characteristics

The majority of respondents were female (N = 1,028, 86.8%). Most interns had participated in patient safety training (N = 988, 83.4%) and clinical reasoning training (N = 1,100, 92.8%). However, fewer interns participated in gamification training (N = 220, 18.6%). As detailed in Table 1.

Table 1.

General characteristics of nursing students (N = 1185)

Variables	N	%
Ⅰ. Sociodemographic Characteristics
Gender
Male	157	13.2
Female	1028	86.8
Family Location
Urban	506	42.7
Rural	679	57.3
Only Child
No	992	83.7
Yes	193	16.3
Student Leader Experience
No	739	62.4
Yes	446	37.6
Education Level
Junior degree or below	588	49.6
Bachelor degree or above	597	50.4
Choosing the field out of interest
No	267	22.5
Yes	918	77.5
Ⅱ. Training Experience
Participation in Patient Safety Training
No	197	16.6
Yes	988	83.4
Participation in Clinical Reasoning Training
No	85	7.2
Yes	1100	92.8
Participation in Gamification Training
No	965	81.4
Yes	220	18.6
Ⅲ. Attitudes and Preferences towards Gamified Patient Safety Training
Willingness to Participate in Future Training
No	248	20.9
Yes	937	79.1
Preferred Training Frequency (if willing)
Twice a weel	98	8.3
Once a week	226	19.1
Once every two weeks	131	11.1
Once a month	521	43.9
Flexible timing	209	17.6
Preferred Session Duration (if willing)
<30 min	443	37.4
30–45 min	486	39.9
45–60 min	124	10.5
>60 min	41	3.5
Flexible timing	91	7.7
Preferred Internship Phase for Training
0-2months	450	38.1
3-4months	343	28.9
5-6months	273	23.0
7-8months	119	10.0

Characteristics of the different classes

The clinical reasoning competency score among the nursing interns was 99.83 (SD: 14.54), while the self-efficacy score was 24.86 (SD: 5.98). Scores for each dimension are presented in Table S1. Due to differing scoring criteria between the two scales, dimension scores were standardized using z-score normalization. Using the standardized dimension scores as observed indicators, five models were fitted (Table 2). As the number of classes increased, the AIC, BIC, and aBIC decreased monotonically. The three-class model yielded the highest entropy (0.944), with average posterior probabilities ranging from 0.990 to 0.994 (all > 0.90), indicating high classification certainty and good class separation. Additionally, both the LMR and BLRT were significant for the three-class model (p < 0.001), supporting its superior fit over the two-class model. In contrast, the four-class model showed lower entropy (0.862) and a wider AvePP range (0.839–0.982), reflecting increased classification uncertainty. The five-class model was not supported by LMR or BLRT (p > 0.05) and included a very small class (0.3%), potentially compromising model stability and interpretability. To ensure robustness, models were re-estimated using different starting values and random seeds; the fit indices and classification probabilities for the three-class solution remained consistent across specifications, indicating a stable cluster structure. Bootstrap tests based on 1,000 samples further confirmed statistical stability, with 95% confidence intervals for all parameter estimates excluding zero. Considering model fit, classification accuracy, class size, and interpretability, the three-class solution was identified as the optimal model.

Table 2.

Model fitting information for the latent profiles (N = 1185)

Model	AIC	BIC	aBIC	Entropy	LMR	BLRT	Category probability (%)	Avepp
					P value	P value
1-class	16829.420	16880.194	16848.431	-	-	-	100	-
2-class	15139.822	15221.062	15170.240	0.927	<0.001	<0.001	81.8/18.2	0.988–0.991
3-class	13498.290	13609.995	13540.115	0.944	<0.001	<0.001	60.5/24.6/14.9	0.990–0.994
4-class	13263.224	13405.394	13316.456	0.862	0.0457	0.0483	43.3/22.5/20.0/14.2	0.839–0.982
5-class	13056.717	13229.352	13121.356	0.890	0.1786	0.1830	42.9/22.4/20.2/14.2/0.3	0.951–0.992

Note. Bolded values indicate the“best”fit for each respective statistic

Abbreviations: AIC Akaike Information Criterion; BIC Bayesian Information Criterion; aBIC Adjusted BIC; LMR LoMendell-Rubin Test; BLRT Bootstrap Likelihood Ratio Test; Avepp Average posterior probabilities. -Not applicable

Profile 1, labeled the “foundational competency group” (N = 291, 24.6%), scored lower than the other two groups across core dimensions such as information systematization and problem analysis, and also exhibited lower self-efficacy. Profile 2, designated the “transitional competency group” (N = 717, 60.5%), showed dimension scores close to the overall sample average, with no pronounced strengths or weaknesses. Profile 3, identified as the “advanced competency group” (N = 177, 14.9%), demonstrated significantly higher scores than the other two groups across all four core dimensions of clinical reasoning as well as in self-efficacy (Fig. 1).

Fig. 1.

Profiles of clinical reasoning competency among nursing interns based on latent profile analysis (N = 1185). Nursing interns had different levels of SACRR and GSE. Category 1: Foundational competency group; Category 2: Transitional competency group; Category 3: Advanced competency group

Demographic and related characteristics of each profile

Multivariate analysis (Table 3) indicated that the distribution of participants across the latent profiles differed significantly with respect to several characteristics, including gender, educational level, student leadership experience, whether nursing was chosen out of interest, participation in patient safety training, participation in clinical reasoning training, willingness to participate in future training, and patient safety perception (p < 0.05).

Table 3.

Participants’ demographic characteristics in different profile types (N = 1185)

Variables	Class 1	Class 2	Class 3	χ²/H	P
n(%)	n(%)	n(%)	n(%)
Gender				10.9021)	0.004**
Male	32	88	37
	(11.0)	(12.3)	(20.9)
Female	259	629	140
	(89.0)	(87.7)	(79.1)
Student Leader Experience				9.2831)	0.010 *
No	203	433	103
	(69.8)	(60.4)	(58.2)
Yes	88	284	74
	(30.2)	(39.6)	(41.8)
Education Level				16.6841)	<0.001***
Junior degree or below	174	327	87
	(59.8)	(45.6)	(49.2)
Bachelor degree or above	117	390	90
	(40.2)	(54.4)	(50.8)
Choosing the field out of interest				19.9011)	<0.001***
No	93	137	37
	(32.0)	(19.1)	(20.9)
Yes	198	580	140
	(68.0)	(80.9)	(79.1)
Participation in Patient Safety Training				7.3521)	0.025*
No	63	110	24
	(21.6)	(15.3)	(13.6)
Yes	228	607	153
	(78.4)	(84.7)	(86.4)
Participation in Clinical Reasoning Training				15.9591)	<0.001***
No	36	41	8
	(12.4)	(5.7)	(4.5)
Yes	255	676	169
	(87.6)	(94.3)	(95.5)
Willingness to Participate in Future Training				42.8661)	<0.001***
No	100	123	25
	(34.4)	(17.2)	(14.1)
Yes	191	594	152
	(65.6)	(82.8)	(85.9)
H-PEPSS	62.44 ± 9.84	70.02 ± 7.23	80.99 ± 6.42	393.7692)	<0.001***

Note. Given the ordinal nature of the scale data, all comparisons were confirmed using non-parametric Kruskal-Wallis tests, which yielded the same pattern of statistical significance

*p < 0.05, ** p < 0.01, *** p < 0.001;

(1) χ² value; (2) H value

Logistic regression analysis results for potential profiles

Prior to conducting multivariable logistic regression, multicollinearity among independent variables was assessed using linear regression. The variance inflation factor for all variables ranged from 1.008 to 1.205, with tolerance values exceeding 0.1, indicating no severe multicollinearity [37]. Using the three latent profiles as the dependent variable, variables significant in univariate analyses were entered into a multinomial logistic regression model with cluster-robust standard errors at the institution level (Table 4). The model showed good explanatory power (log pseudolikelihood = − 817.61; pseudo R² = 0.2603) [38]. Results were largely consistent with the unclustered model after applying cluster-robust standard errors. The association between willingness for future training and membership in the advanced group decreased from significant (p = 0.048) to marginally significant (p < 0.06), while its association with the transitional group remained significant (p < 0.05). The direction and statistical significance of effects for other variables were essentially unchanged. The final model indicated that gender, education level, interest in choosing nursing as a major, willingness for future training, and patient safety perception were significant factors associated with class membership (p < 0.05).

Table 4.

Predictors of latent profile membership

	Class 11) VS Class 2				Class 11) VS Class 3				Class 32) VS Class 2
	β	P	OR	95%CI	β	P	OR	95%CI	β	P	OR	95%CI
Gender
Female	-0.067	0.690	0.935	(0.672, 1.300)	-0.863	<0.001	0.422***	(0.278, 0.640)	0.796	0.001	2.216**	(1.385, 3.548)
Male(Refer)
Education level
Bachelor degree or above	0.629	<0.001	1.876***	(1.356, 2.594)	0.439	0.172	1.550	(0.826, 2.908)	0.191	0.365	1.210	(0.801, 1.827)
Junior degree or below (Refer)
Student experience
Yes	0.298	0.140	1.348	(0.907, 2.002)	0.365	0.418	1.441	(0.596, 3.484)	-0.067	0.808	0.935	(0.545, 1.604)
No (Refer)
Choosing the field out of interest
Yes	0.656	<0.001	1.927***	(1.479, 2.511)	0.224	0.321	1.251	(0.804, 1.945)	0.432	0.103	1.541	(0.916, 2.592)
No (Refer)
Participation in patient safety training
Yes	0.118	0.594	1.125	(0.730, 1.734)	0.108	0.808	1.114	(0.466, 2.661)	0.010	0.976	1.010	(0.522, 1.956)
No(Refer)
Participation in clinical reasoning training
Yes	0.468	0.162	1.597	(0.829, 3.075)	0.489	0.392	1.631	(0.531, 5.007)	-0.021	0.970	0.979	(0.320, 2.996)
No (Refer)
Willingness to participate in future training
Yes	0.456	<0.001	1.577***	(1.239, 2.008)	0.665	0.060	1.944	(0.972, 3.887)	-0.209	0.440	0.811	(0.477, 1.379)
No (Refer)
H-PEPSS	0.116	<0.001	1.123***	(1.100, 1.147)	0.311	<0.001	1.364***	(1.321, 1.409)	-0.195	<0.001	0.823***	(0.808, 0.839)

Note. Std. err. adjusted for 15 clusters in institution

Refer Reference group; OR Odds ratio; 95% CI 95% Confidence Interval;

¹⁾ foundational competency group as the reference category; ²⁾ advanced competency group as the reference category

*p < 0.05, ** p < 0.01, *** p < 0.001;

Specifically, using the basic competency group as the reference, higher perceived patient safety perception was positively associated with belonging to the advanced competency group (adjusted OR per 5-point increase = 4.72, 95% CI: 4.023–5.553, p < 0.001). Meanwhile, a bachelor’s degree or higher (OR = 1.876, 95% CI: 1.356–2.594), voluntarily choosing nursing (OR = 1.927, 95% CI: 1.479–2.511), and willingness for future training (OR = 1.577, 95% CI: 1.239–2.008) were associated with higher relative odds of being in the transitional competency group (all p < 0.05). Conversely, with the advanced competency group as the reference, females had higher relative odds than males of being in the transitional competency group (OR = 2.21, 95% CI: 1.385–3.548, p = 0.001). Adjusted predicted probabilities were calculated based on the model stratified by gender and education level (Table S2). Female interns with a bachelor’s degree or higher had a 13.4% probability of belonging to the advanced competency group, lower than the 21.5% probability for male interns with a diploma or lower. Furthermore, within the same gender, those with a bachelor’s degree or higher consistently showed higher predicted probabilities of belonging to the transitional competency group compared to those with lower education levels (Males: 60.6% vs. 51.4%; Females: 66.2% vs. 56.4%).

Discussion

Heterogeneity in clinical reasoning competency among nursing interns

This study employed latent profile analysis, identifying three distinct subgroups with significant differences. These groups were categorized as the foundational competency group (24.6%), the transitional competency group (60.5%), and the advanced competency group (14.9%). This finding reveals the heterogeneity in the composition of clinical reasoning ability within the intern population, contrasting with previous approaches that often treated this group as a homogeneous whole [39, 40]. Importantly, these profiles represent distinct competency patterns observed at a single point in time, rather than sequential stages in a linear developmental pathway. Only a minority of interns (14.9%) were classified into the advanced profile, whereas the majority (60.5%) were classified into the transitional profile, characterized by balanced but not yet proficient competency. This underscores the necessity of systematically enhancing clinical reasoning competency within this population.

The identified profiles align with Benner’s “novice to expert” framework [41], which describes skill acquisition through qualitative performance changes rather than linear knowledge accumulation. For the foundational competency group (low competency–low self-efficacy), scores were lowest across core dimensions including information systematization, problem analysis, and self-efficacy. This pattern aligns with Bandura’s self-efficacy theory and existing evidence linking self-efficacy positively with clinical reasoning. Individuals with low self-efficacy tend to avoid challenges in complex situations, a tendency associated with hindered competency development [19, 42]. From a Bennerian perspective, this profile reflects characteristics of the “novice” stage, where performance relies on objective attributes and context-free rules [41]. An encouraging finding is the emerging upward trend in their reflective ability, indicating potential for reflective learning. These interns demonstrate capacity to examine shortcomings in their own clinical behaviors, gradually building a cognitive loop from practice to reflection and back to practice [43]. Thus, educational efforts for this group should prioritize consolidating foundational knowledge and building confidence.

In contrast, the advanced competency group (high competency–high self-efficacy) scored highest across all dimensions, yet showed comparatively lower reflective ability. This pattern may indicate that as clinical tasks become more familiar, deliberate reflection diminishes—a phenomenon consistent with Benner’s “proficient” stage, where pattern recognition and intuition guide performance but may occasionally be accompanied by overconfidence [41, 44]. Overconfidence has been linked to hasty conclusions without considering contradictory evidence or alternative possibilities [45]. The intense demands of the internship environment, including time pressures and cognitive load, may further limit opportunities for systematic reflection among these students [46, 47]. Given that reflective practice is crucial for integrating new knowledge [47], managing complex cases, and preventing errors [48, 49], targeted interventions are warranted. For these students, structured reflection—such as guided debriefings or reflective journals—should be embedded within specialized training or complex case discussions [50].

Regarding the transitional profile (moderate competency–balanced development), this group demonstrated mid-range, stable scores across dimensions. Their pattern most closely resembles Benner’s “advanced beginner”, demonstrating marginally acceptable performance and beginning to recognize recurrent patterns, yet lacking the holistic grasp of more proficient practitioners [41]. While their competency development appears relatively balanced, they may lack impetus or direction for improvement. Research suggests that effective teaching interactions and targeted questioning during clinical guidance are associated with enhanced clinical reasoning development [51]. Furthermore, explicitly linking teaching methods to target skills helps stimulate critical thinking, clinical judgment, and decision-making [52]. Therefore, for this largest subgroup, core teaching strategies should focus on further stimulating students’ critical thinking and guiding the integration of knowledge and skills.

Analysis of factors influencing clinical reasoning competency profiles

Gender

Gender was a significant predictor of profile membership. With the foundational competency group as the reference, females had a 57.8% lower odds of being in the advanced competency group (OR = 0.422, 95% CI: 0.278–0.640, p < 0.001). With the advanced group as the reference, females were significantly more likely than males to be classified into the transitional group (OR = 2.21, 95% CI: 1.385–3.548, p = 0.001). This finding contrasts with Hege et al. [53], who reported generally better clinical reasoning performance among female medical students. This discrepancy may be related to gender role socialization within the nursing education context. Research suggests that in nursing, a traditionally female-dominated profession, gender stereotypes may influence how students perceive and present their competencies [54] Other studies have found that while biological sex itself shows no significant association with critical thinking ability, certain traits related to gender role orientation do, suggesting that factors such as socialized roles, behavioral characteristics, and learning styles may be the operative elements [55]. Furthermore, gender differences exist in perceived stress during clinical placements, with female nursing students often reporting higher stress levels, which may be accompanied by instability in complex clinical reasoning performance [56, 57]. Meanwhile, as a minority group within the profession, male interns may face specific role pressures and societal expectations, potentially contributing to differences in self-efficacy and career trajectories [58].

Given that males comprised only 13.2% of the sample, the identified latent characteristics and their associations with gender likely reflect the response patterns and competency distribution of the female population. These findings are best interpreted as statistical associations within this specific sample structure. Furthermore, as the sample was drawn from South China, variations in clinical teaching resources and supervision models across regions may limit generalizability. Future multicenter studies with increased male participation are recommended to further validate the stability and external validity of these results.

Education level

There was a significant correlation between education level and profile membership. With the foundational competency group as the reference, interns holding a bachelor’s degree or above had 87.6% higher odds of being classified into the transitional competency group (OR = 1.876, 95%CI: 1.356–2.594, p < 0.001). This finding is consistent with existing research indicating that higher education levels serve as a positive predictor of clinical reasoning competency among nurses [59, 60]. A possible explanation is that more systematic and in-depth training within higher-level curricula enables students to establish a more robust theoretical framework, develop critical thinking skills, and form a more comprehensive understanding of nursing practice [61]. Furthermore, educational background may also strengthen interns’ academic self-concept and learning confidence, making them more persistent and adept at integrating complex reasoning skills when facing clinical challenges [62, 63]. This finding clearly suggests the necessity of designing differentiated clinical education support plans tailored to the educational backgrounds of nursing interns.

Choosing the field out of interest

This study found that professional interest was also closely related to the profile membership. With the foundational competency group as the reference, interns who chose nursing out of interest had 92.7% higher odds of being classified into the transitional competency group (OR = 1.927, 95%CI: 1.479–2.511, p < 0.001). This finding supports the positive role of professional identity and intrinsic motivation in competency development [64]. Within the study sample, only 22.5% of interns reported choosing nursing out of interest. This proportion is lower than the 55.4% reported in some international studies [65], a discrepancy potentially attributable to cultural differences, educational background, and influences from the work environment, particularly the university setting [65, 66]. This early, self-initiated sense of identification may serve as an intrinsic engine, motivating students to actively engage in clinical learning, integrate knowledge proactively, and construct a stable competency framework [67]. This finding suggests that educators should focus more on the development of students’ early professional values and identity. Research indicates that role modeling within the clinical learning environment and early clinical experiences can influence the professional identity development of nursing students [66]. Therefore, more targeted support could be provided for students with a weaker initial professional identity.

Patient safety perception

One of the most significant findings of this study was the strong association between the patient safety perception and membership in the advanced clinical reasoning competency group. With the foundational competency group as the reference, nursing interns with a higher level of patient safety perception had significantly increased odds of belonging to the advanced group (OR = 1.364, 95%CI: 1.321–1.409, p < 0.001). This result aligns with existing literature indicating that nursing students’ level of patient safety perception correlates with stronger clinical decision-making, risk judgment, and comprehensive ability [68–70]. Interns with a high level of patient safety perception typically possess sharper risk consciousness and more standardized operational practices. These characteristics enable them to accumulate and integrate clinical experience more effectively, exhibiting a more prudent and systematic approach to clinical reasoning [39].Of note, this study assessed interns’ perceived rather than actual safety behaviors. The cross-sectional design precludes causal inference. The findings indicate only an association between patient safety perception and clinical reasoning ability, with possible reverse or bidirectional effects. Unmeasured educational factors, such as clinical learning environment and mentorship quality, may also confound this relationship.

Furthermore, willingness to participate in gamified patient safety training was also associated with profile membership. With the foundational competency group as the reference, students willing to undergo such training had 57.7% higher odds of belonging to the transitional group (OR = 1.577, 95% CI: 1.239–2.008, p < 0.05). This finding is consistent with self-directed learning theory, which posits that learning motivation and autonomy are intrinsic drivers of professional competency development [71, 72]. Although various forms of gamified teaching show potential in enhancing clinical reasoning competency, their effectiveness remains highly dependent on appropriate alignment with learning objectives and the instructional context [73]. Gamification in nursing education holds the potential to enhance engagement and learning performance [74]. Therefore, future efforts could introduce progressive, feedback-rich simulations or gamified tasks in safety-critical situations, with prospective interventional studies further assessing its impact on clinical reasoning competency.

Limitations

This study has several limitations. First, the cross-sectional design precludes causal inference and does not allow examination of the dynamic developmental trajectory of clinical reasoning over time. Longitudinal studies are needed to clarify how clinical reasoning profiles evolve during internship training.

Second, the sample was limited to nursing interns from four provinces in South China, which may reduce the generalizability of the findings. Regional differences in culture, educational resources, and clinical training practices may have influenced the observed profile structure. In addition, the use of convenience sampling and online questionnaire distribution via WeChat may have introduced selection bias, and duplicate submissions from shared devices or networks cannot be fully ruled out.

In addition, several measurement and methodological issues should be considered. The latent profiles identified in this study are model-based classifications rather than directly observed groups. Moreover, all key variables were assessed using self-report instruments, which may be subject to response bias. The reflection dimension was measured using a single item, which may have limited measurement precision compared with multi-item subscales.

Although gender and education level were identified as significant factors, other potentially important variables were not included, such as individual cognitive styles, family support, and learning environment characteristics. These unmeasured factors may exert mediating or moderating effects on clinical reasoning ability. Moreover, measurement invariance and profile stability across institutions were not formally tested, which limits confidence in the comparability and robustness of the identified profiles.

Future research should address these limitations by using longitudinal designs, expanding sampling to more diverse geographic regions and institutions, and incorporating a broader range of psychosocial and environmental variables. Including objective competency measures would also provide a more comprehensive assessment of clinical reasoning. Finally, future studies should develop and evaluate targeted curricular interventions tailored to specific profile subgroups.

Conclusions

This study identified three latent profiles of clinical reasoning competency among nursing interns: foundational, transitional, and advanced. Each profile demonstrated unique characteristics: the foundational group requires development in systematic reasoning and self-efficacy; the transitional group needs support to bypass developmental plateaus; and the advanced group must address deficits in reflective practice. Furthermore, gender, educational level, professional interest, patient safety perception, and training willingness were associated with profile membership. These findings may assist educators in designing targeted strategies to foster clinical reasoning related to patient safety, based on students’ latent profiles and associated factors.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Y.T. (Corresponding Author): Conceptualization, Methodology, Supervision, Project administration, Writing – Review & Editing. Y.Z. (Co-first Author): Investigation, Resources, Data Curation, Writing – Original Draft. H.C. (Co-first Author): Formal analysis, Data Curation, Visualization, Writing – Original Draft. G.Z.: Investigation, Validation, Writing – Review & Editing. S.X.: Methodology, Software, Validation. Y.F.: Resources, Investigation. All authors read and approved the final manuscript.

Funding

This work was supported by the Teaching Reform Project of Central South University [grant numbers 2025jy078].

Data availability

The dataset supporting the conclusions of this article is included within the articles. Specifically, the de-identified analytic dataset, the corresponding Mplus syntax for the latent profile analysis, and the codebook are provided as Supplementary Material 4–6.

Declarations

Ethical approval

This study obtained ethical approval from the the Clinical Medical Ethics Committee of Xiangya Hospital, Central South University (Approval No. 2025091587). The research was conducted in accordance with the Declaration of Helsinki. Participation in the survey was voluntary, and implied consent was obtained when participants completed the questionnaire. Anonymity and confidentiality were ensured, as all data were analyzsed using numerical codes.

Disclosure

All authors have approved the manuscript for publication, and it is an original study that has never been published previously and it is not being considered for publication elsewhere. Yuting Zeng and Huiqiong Chen are co-first authors. Yinying Tang is corresponding author.

Competing interests

The authors declare no competing interests.