Self-directed learning and critical thinking show weak links to nursing internship scores

Every nursing student learns the same lesson early: the internship is where classroom knowledge collides with real patients, real time pressure, and real consequences. Two learner attributes are widely assumed to carry trainees through that transition. Self-directed learning, the capacity to diagnose one’s own learning needs, set goals, find resources, and evaluate outcomes, is often described as a prerequisite for lifelong clinical development. Critical thinking disposition, the tendency to appraise evidence, weigh alternatives, and reason systematically under uncertainty, is treated as the dispositional backbone of safe, evidence-informed care. But a new study from a urology nursing rotation in China suggests that these celebrated attributes may be far less visible in the scores that actually define internship performance than educators have long assumed.

The research, published in Frontiers in Public Health by a team led by Caixia Dai and Shuang Peng of The Second Xiangya Hospital of Central South University, took an unusually direct approach to the question. Rather than designing special research assessments, the investigators mined the department’s routine evaluation system for eighty nursing interns who completed a standardized urology rotation between April 2022 and March 2024. Five routinely recorded performance indicators were analyzed: a written theory examination, a practical skills average, weekly assignment scores, case presentation scores, and a comprehensive end-of-rotation instructor evaluation. The interns, whose mean age was 21.36 years and who were predominantly female, completed validated Chinese instruments measuring self-directed learning ability and critical thinking disposition during the teaching period. The study was registered as a clinical trial, ChiCTR2400085591, and approved by the hospital’s ethics committee.

The technical design mattered. Because many previous studies in this field rely on a single endpoint and unadjusted bivariate comparisons, the team built multivariable linear regression models that adjusted for age, sex, and the final surgical nursing course score, a proxy for prior academic performance. Sensitivity analyses included Spearman rank correlations to test robustness to distributional assumptions, standardized regression models, and refitted models excluding individual covariates. At the domain level, the investigators examined four self-directed learning domains and seven critical thinking domains against the five performance indicators, producing fifty-five exploratory comparisons that were then screened with Benjamini-Hochberg false-discovery-rate adjustment to control for chance findings.

The headline result is stark. Critical thinking disposition total score was essentially uncorrelated with the theory examination score in the Pearson analysis, with a correlation coefficient of 0.003 and a p value of 0.980, about as close to zero as statistics allows. After covariate adjustment, the association remained null, with an unstandardized coefficient of 0.014 and a 95 percent confidence interval spanning -0.103 to 0.131. Self-directed learning total score also failed to show an independent association with theory performance, with a coefficient of -0.110 and a p value of 0.065 that fell just short of conventional significance. Neither attribute was independently associated with the comprehensive instructor rating either, with p values of 0.469 and 0.361 respectively. The overall regression models were not statistically significant for either outcome.

Yet the picture was not uniformly null. In exploratory domain-level analyses, two nominal positive associations emerged. Critical thinking open-mindedness showed a positive correlation with weekly assignment performance, with r = 0.244 and p = 0.032, a finding that also held in the Spearman rank analysis with ρ = 0.250 and p = 0.028. Critical thinking inquisitiveness showed a positive correlation with case presentation performance, with r = 0.292 and p = 0.010, and the Spearman analysis was even stronger at ρ = 0.328 and p = 0.004. Both associations are theoretically coherent: weekly assignments reward continuous engagement and self-regulation, while case presentations require organizing clinical information, prioritizing problems, and communicating reasoning in a structured format, processes that plausibly engage dispositional curiosity and openness.

But here the study delivers its most sobering statistical lesson. Across all fifty-five exploratory Pearson comparisons, not a single domain-level association remained statistically significant after false-discovery-rate adjustment, with all q values at or above 0.315. The same held for the fifty-five corresponding Spearman comparisons, with all q values at or above 0.197. The authors are explicit that these domain-level findings should be considered exploratory and hypothesis-generating rather than confirmatory. With fifty-five comparisons, the probability of at least a few unadjusted p values falling below 0.05 by chance alone is substantial, and the FDR screen is designed precisely to expose that risk.

Why would critical thinking disposition fail to predict a written theory examination? The authors point to a construct-coverage gap that has been recognized in the assessment literature for decades. The CTDI-CV, the Chinese adaptation of the California Critical Thinking Disposition Inventory, measures a dispositional tendency toward critical thinking rather than demonstrated critical-thinking competence. The theory examination, in turn, primarily assessed knowledge retrieval and application within a test environment, and was never designed as a direct measure of reasoning ability. Biggs’ model of learning approaches provides a theoretical basis for expecting that learner attributes and performance will align differently across assessment formats. Empirically, Fero and colleagues demonstrated this method-dependency directly in nursing students: critical thinking disposition scores were not associated with performance on paper-based vignettes but were significantly associated with high-fidelity simulation performance, an assessment modality that more authentically samples clinical reasoning processes.

The comprehensive instructor evaluation presents a different measurement problem. Workplace-based assessment research has shown that instructor ratings in clinical settings are heavily influenced by rater cognition, idiosyncratic performance theories, and contextual cues, often reflecting the rater’s subjective impression of the trainee rather than the specific learner attributes captured by self-report instruments. Williams and colleagues further identified that cognitive, social, and environmental factors introduce systematic bias into clinical performance ratings, reducing their ability to discriminate between learners who differ primarily in cognitive dispositions. In this cohort, the mean comprehensive evaluation score was 92.48 out of 100, consistent with the well-documented leniency tendency in summative workplace-based assessments, where social dynamics between supervisors and trainees compress the upper range of ratings and reduce their discriminative power. Ceiling effects of this kind can attenuate observable correlations even when true associations exist.

The self-directed learning null result also warrants careful interpretation. Previous studies have linked self-directed learning to positive learning outcomes and clinical competence development, but others have found that this relationship is substantially mediated by contextual factors in the clinical learning environment. In a single specialty ward, all interns were subject to the same supervision structure, workload, and available learning resources, which may have constrained the variance in self-directed learning behaviors and thereby attenuated any detectable association with assessment scores. Restricted variability in some performance indicators, particularly the practical skills and case presentation scores, which were concentrated within narrow ranges, may have further reduced the ability to detect associations. The self-report format of both instruments also introduces the possibility of social desirability bias.

The study has clear strengths. It links validated learner ability measures to multiple routinely collected internship outcomes, enabling outcome-specific interpretation rather than reliance on a single summary score. It incorporates covariate adjustment and multiple sensitivity checks. And because it uses rotation records generated during usual educational practice rather than a specially designed research instrument, the findings have direct relevance to existing clinical education systems and can inform practical decisions about assessment design without requiring additional data infrastructure. The predominance of null and non-robust associations is itself informative, suggesting that routinely used internship assessments may not closely align with the learner attributes captured by self-directed learning and critical thinking disposition measures.

Limitations temper the conclusions. The study was conducted within a single specialty rotation at one institution, and the findings may not generalize to other clinical settings or student populations. The observational cross-sectional design precludes causal inference, and covariate adjustment was limited to age, sex, and prior academic performance. Age varied only from 20 to 23 years, and the sample was predominantly female, so these demographic variables provided limited information for adjustment. Unmeasured educational and clinical factors, including broader measures of prior academic achievement, prior clinical exposure, supervisor characteristics, workload, and the clinical learning environment, were not available and may have introduced residual confounding. Formal documentation regarding theory-examination blueprinting and psychometric consistency across cohorts was not available, nor were detailed scoring anchors or inter-rater reliability records for the comprehensive instructor evaluation.

The confidence intervals around the principal null associations indicate that the study provides reasonable precision for excluding large associations but does not exclude small relationships of potential educational relevance. The critical thinking total to theory score correlation was close to zero but had a 95 percent confidence interval of -0.221 to 0.227, while the self-directed learning total to comprehensive evaluation correlation had a 95 percent confidence interval of -0.103 to 0.347. The non-significant findings should therefore be interpreted as an absence of clearly demonstrated associations in this cohort rather than definitive evidence of no association.

For educators, the practical implication is a call for more construct-aligned assessment approaches. If self-regulated learning and clinical reasoning are the competencies that matter for safe practice, then assessment systems should include components that directly capture them, whether through portfolio dimensions, simulation-based evaluations, or process-oriented measures that reward self-initiated information seeking and independent reflection. Written examinations and global instructor ratings, the workhorses of routine clinical teaching, may simply be the wrong instruments for the job. The authors emphasize that these findings are preliminary and context-specific, requiring replication across other clinical rotations and institutions before broader conclusions can be drawn. But the message for clinical nursing education is already clear: the attributes we celebrate in our trainees may be invisible in the scores we collect, and closing that gap demands assessment design that looks as carefully at how students learn as at what they know.

Subject of Research: Associations of self−directed learning ability and critical thinking disposition with internship performance indicators in nursing interns

Article Title: Associations of self−directed learning ability and critical thinking disposition with internship performance indicators in nursing interns

Article References: Dai, C., Wei, Y., Liang, Y., Huang, L., & Peng, S. (2026). Associations of self−directed learning ability and critical thinking disposition with internship performance indicators in nursing interns. Frontiers in Public Health, 14, Article 1945090. https://doi.org/10.3389/fpubh.2026.1945090

Image Credits: AI Generated

DOI: 10.3389/fpubh.2026.1945090

Keywords: Associations, self, directed, learning, ability, critical, thinking, disposition, internship, performance, indicators, nursing