Effects of a closed-loop “Subspecialty-mind Mapping-Rolling Quiz” model based on cognitive load theory on the clinical decision-making ability of ICU nurses: a mixed-methods study

Abstract

Background

In intensive care unit (ICU), traditional training models fail to meet the integrated cognitive demands. Additionally, subspecialty training and mind mapping have rarely been applied with a unifying theoretical framework for knowledge retention.

Aims

Based on the Cognitive Load Theory and experiential learning cycle, this study aimed to construct and validate a closed-loop training model (Subspecialty-Mind Mapping-Rolling Quiz) to enhance ICU nurses’ clinical decision-making, knowledge retention, self-efficacy and critical thinking.

Design

An explanatory sequential mixed-methods design.

Methods

The quantitative phase was a cluster randomized controlled trial conducted in a tertiary teaching hospital in China. A total of 196 nurses from 4 ICU units were randomly assigned by unit to either an intervention group (2 units, n = 102) receiving the closed-loop training or a control group (2 units, n = 94) receiving traditional training from January to September 2025. Participants’ Lasater Clinical Judgment Rubric (LCJR), theoretical and practical examination, General Self-Efficacy Scale (GSES), Critical Thinking Disposition Inventory-Chinese Version (CTDI-CV) scores, and knowledge retention rates were compared at pre-intervention, post-intervention (6 months), and at a 3-month follow-up. All 196 nurses completed the study. In the qualitative phase, 16 participants were sampled for semi-structured interviews.

Results

Quantitative results showed that the intervention group had significantly higher LCJR scores (t = 12.917, p < 0.01), post-intervention and delayed post-test theoretical and practical, GSES, and CTDI-CV scores (all p < 0.05). Qualitative analysis identified four themes: reconstruction of cognitive schema, shift in learning identity, sustained activation of metacognition and incentives coexisting with challenges in implementation.These themes provided insights into the mechanisms behind the quantitative improvements.

Conclusion

The Subspecialty-Mind Mapping-Rolling Quiz enhanced ICU nurses’ clinical decision-making, knowledge retention, self-efficacy, and critical thinking ability. By fostering knowledge internalization and contextualized retrieval, this model serves as an effective and theoretically grounded training solution.

Clinical trial number

Not applicable.

Introduction

The intensive care unit (ICU) is one of the most complex and dynamic environments in healthcare, requiring nurses to possess advanced abilities in making rapid and accurate clinical decisions under uncertainty [1, 2]. Traditional ICU nurse training often focuses on imparting discrete knowledge and single skills, leading to fragmented knowledge, a theory-practice gap, and simplistic evaluation methods [3, 4]. These limitations hinder nurses’ ability to translate learned knowledge into practical decision-making skills in complex clinical situations, creating a bottleneck in improving nursing quality. Although new approaches such as subspecialty development, mind mapping, and mobile learning have been sporadically attempted, most remain at the level of tool application, lacking systematic theoretical integration and mechanistic exploration [5–7]. The Cognitive Load Theory (CLT) suggests that learning effectiveness is limited by working memory capacity, and that instructional design should optimize intrinsic, extraneous, and germane cognitive load [8, 9]. Meanwhile, Kolb’s Experiential Learning Cycle emphasizes learning as a continuous process of experience, reflection, conceptualization, and practice. These studies innovatively integrated these theories into a top-level design, constructing a closed-loop training model that reorganized content through critical care subspecialties (reducing intrinsic load) [10], used mind maps as visual knowledge tools (reducing extraneous load) [11], employed rolling quizzes as retrieval practice and feedback mechanisms (increasing germane load) [12], and linked to performance management (providing motivation) [13]. This model aimed to simulate real clinical thinking pathways, promoted structured knowledge internalization and contextualized application, and ultimately enhanced nurses’ clinical decision-making abilities.

This study used a mixed-methods approach combining a cluster randomized controlled trial with qualitative interviews [14] to comprehensively address the following questions:

The specific research questions were:

1. Could this closed-loop model effectively improve ICU nurses’ clinical decision-making ability, knowledge retention, and critical thinking?

2. What were the key components and internal mechanisms through which the model operated?

3. What opportunities and challenges existed for the sustainability and scalability of this model in practice?

Methods

Study design

This study employed an explanatory sequential mixed-methods design [15, 16]. This design involves first collecting and analyzing quantitative data (the cluster randomized controlled trial) to evaluate the intervention effects, followed by qualitative data (semi-structured interviews) to help explain, deepen, and extend the quantitative findings. The theoretical framework and technical route were detailed in Fig. 1.

Fig. 1.

Research theoretical framework and technical route

Subjects and blinding

Four ICU units (Internal Medicine ICU, Surgical ICU, Respiratory ICU, and Emergency ICU) in a tertiary teaching hospital in China were selected as clusters using cluster random sampling. All registered ICU nurses at the hospital from January to September 2025 were invited to participate. Inclusion criteria: (1) holding a valid nurse practitioner certificate and being on duty; (2) having ≥ 1 year of ICU work experience; (3) voluntary participation with signed informed consent. Exclusion criteria: (1) interns, standardized training nurses, or rotating nurses; (2) those on leave or out for training for more than one month during the study period. A total of 196 nurses were included (15 male, 181 female), with an average age of 28.52 ± 4.22 years and work experience of 6.84 ± 3.51 years. The four units were randomized into an intervention group (2 units) and a control group (2 units) using a random number table. To avoid contamination, randomization was performed at the unit level. Due to the nature of the intervention, blinding of participants and trainers was not possible, but outcome assessors (simulated case examiners and data analysts) were blinded to group allocation.

Intervention protocol

Intervention design: building the “Subspecialty-mind mapping-rolling Quiz” closed-loop model based on cognitive load Theory

The intervention was designed based on Cognitive Load Theory (CLT) to optimize ICU nurses’ cognitive resource allocation during specialized knowledge acquisition. The model was built on the theoretical premise that content reorganization would reduce intrinsic cognitive load, optimized representation would reduce extraneous cognitive load, and well-designed learning tasks would increase germane cognitive load, ultimately promoting deep processing and clinical transfer of knowledge.The entire model formed a continuous “teaching-assessment-feedback-incentive” closed loop, which was detailed below:

1. Content reorganization: reducing intrinsic cognitive load through critical care subspecialty grouping

• Procedure: The complex body of critical care knowledge was modularized into specialized directions (e.g.,respiratory therapy, circulatory support, Continuous Renal Replacement Therapy (CRRT), critical care nutrition) to form multiple subspecialty learning groups.

• Theoretical Rationale: Intrinsic cognitive load is determined by the inherent complexity of the learning material and the learner’s prior knowledge. Critical care knowledge is inherently high in element interactivity, leading to cognitive overload. Subspecialty grouping divides this complex domain into smaller, and more focused modules with lower element interactivity, allowing nurses to concentrate on processing a simplified and focused knowledge module at a time. This significantly reduces the intrinsic cognitive load of the learning task, freeing up working memory capacity for processing core concepts and their interrelationships.

• (2)Knowledge visualization: reducing extraneous cognitive load with mind maps

• Procedure: After each subspecialty training session, core members created standardized mind maps using a radial, hierarchical structure to represent key concepts, decision-making processes, critical evidence, and precautions, which were then uploaded to a nursing app.

• Theoretical Rationale: Extraneous cognitive load arises from suboptimal instructional design and information presentation unrelated to learning goals. Traditional linear text-based PPTs or handouts require learners to self-organize knowledge structures, consuming significant cognitive resources. As a multimedia learning tool, mind maps visually present the internal logical structure of knowledge, making implicit knowledge explicit and systematic. This was hypothesized to save cognitive effort otherwise spent on organizing information or understanding disjointed points, greatly reducing extraneous cognitive load caused by disorganized information presentation.

• (3)Reinforcement and Feedback: Increasing Germane Cognitive Load through Rolling Quizzes as Retrieval Practice

• Procedure: Each month, an assessment team designed scenario-based questions based on the mind map content and conducted random, rolling quizzes during morning meetings, etc., providing immediate feedback.

• Theoretical Rationale: Germane cognitive load refers to mental activities related to schema processing, construction, and automation—the “good” load that facilitates learning. The essence of rolling quizzes is “retrieval practice,” actively recalling information from memory, one of the most effective learning strategies known. Scenario-based questions compel nurses to activate and apply learned knowledge in simulated clinical situations, engaging in pattern recognition, reasoning, and decision-making. This deep processing consumes considerable cognitive resources to build and strengthen schemas. This was hypothesized to increase beneficial germane cognitive load, channeling the freed-up resources (from the previous two stages) into deep learning activities.

• (4)Motivational support: providing external motivation through performance management

Procedure: Nurses’ performance in the rolling quizzes was incorporated into the monthly performance evaluation system, linked to bonuses, awards, etc., and recorded in personal digital files.

Theoretical Rationale: Motivation is a prerequisite for mobilizing cognitive resources for learning. The performance management mechanism acts as an external incentive, providing clear goals and timely rewards, which was hypothesized to enhance nurses’ engagement and persistence throughout the training process. It ensures their willingness to actively invest the necessary cognitive effort into the deep learning tasks aimed at “reducing intrinsic load,” “reducing extraneous load,” and “increasing germane load,” thereby guaranteeing the long-term effectiveness of the closed-loop model [17].

• (4)Closed-Loop formation: feedback and process optimization

• Monthly analysis of quiz evaluation data identified common knowledge weaknesses and training gaps. These results were sent to the respective subspecialty groups to inform their next training plan and mind map revisions, enabling continuous quality improvement (PDCA cycle) and forming a self-optimizing closed-loop system.

Intervention implementation

Phase 1: Establishing Critical Care Subspecialty Groups (Months 1–2)

• Survey and Grouping: Based on patient types and technical challenges, six subspecialty groups were established: Respiratory Therapy, Circulatory Support, CRRT, ECMO, Early Rehabilitation, and Nutritional Support.

• Recruitment and Roles: Using a combination of voluntary sign-up and head nurse coordination, there were 3–5 core members (senior, interested nurses as leaders) and several rotating members in each group. Core members were responsible for collecting, organizing, and updating knowledge (guidelines, literature, protocols) within their subspecialty and delivering primary training.

• Planning: Each group developed an annual learning plan and training objectives.

Phase 2: Subspecialty Training and Mind Map Development (Months 3–6)

• Systematic Training: Each subspecialty group was organized to have at least one thematic training session per month (e.g., mini-lectures, workshops, case discussions, bedside teaching) focused on clinical practice (e.g., ventilator waveform interpretation, vasoactive drug titration, CRRT complication monitoring).

• Knowledge Mapping: Within one week after each training, core members created clear, focused mind maps covering key knowledge points, critical procedures, and precautions.

• Review and Upload: Mind maps were reviewed by the head nurse and department experts before being uploaded to the “Nursing Assistant” app knowledge base, which was categorized by subspecialty for anytime access.

Phase 3: Rolling Quizzes and Performance Evaluation (Starting from Month 4 onwards)

• Quiz Mechanism: In late each month, an assessment team (head nurse, team leaders, core members from the training subspecialty) conducted quizzes based on the previous month’s training content and corresponding mind map.

• Quiz Format: Utilizing fragmented time (e.g., morning handovers) or dedicated monthly sessions, nurses were randomly selected or named for quizzes. Questions focused not only on recall but also on scenario simulation and case analysis.

• Recording and Feedback: A dedicated person recorded responses (correct, partially correct, incorrect). Incorrect or incomplete answers received immediate correction and explanation, achieving “instant questioning, learning, and feedback.”

• Performance Linkage: Participation and accuracy rates were incorporated into the monthly performance evaluation system, linked to bonus distribution, awards, external training opportunities, and personal training records.

Control Group: Received traditional training and participated in 1–2 routine theoretical lectures and large-group skill demonstrations monthly, without structured mind maps or regular rolling quizzes.

Evaluation indicators

Quantitative indicators (collected at T0-pre-intervention, T1-post 6 months intervention, T2-3 months post-intervention)

1. Primary Outcome: The Lasater Clinical Judgment Rubric (LCJR) scores. The LCJR is a validated instrument for assessing clinical judgment across four dimensions: noticing, interpreting, responding, and reflecting. The Chinese version contains 11 items rated on a 4-point Likert scale (total score range: 11-44). This instrument was selected as the primary outcome because it is the gold standard for evaluating clinical judgment in nursing education. The Chinese version has demonstrated good reliability (Cronbach’s α = 0.89) and validity in previous studies [18]. At T1, participants were assessed via standardized simulated cases using high-fidelity simulators, and were scored blindly by two trained assessors who were unaware of group allocation.

2. Secondary Outcomes:

• Specialized Knowledge and Skill Scores. Self-designed standardized tests (100 points) assessed theoretical knowledge, while operational checklists evaluated clinical skills. Content validity was established through review by 5 senior ICU experts (content validity index = 0.92). Internal consistency for the theoretical test was Cronbach’s α = 0.87, and inter-rater reliability for the practical examination was κ = 0.84 based on 20 pilot participants. Scores were assessed at T0, T1, and T2.

• General Self-Efficacy Scale (GSES). This 10-item scale measures general self-efficacy with a total score range of 10–40. Higher scores indicate stronger self-efficacy. The Chinese version has demonstrated good reliability (Cronbach’s α = 0.87) in Chinese nurse populations [19]. • Critical Thinking Disposition Inventory-Chinese Version (CTDI-CV). This 70-item instrument assesses critical thinking across 7 dimensions, with a total score range of 70–420. The Chinese version has established validity and reliability (Cronbach’s α = 0.90) [6].

• Patient Safety Indicators (Exploratory).Unplanned extubation rate, CLABSI incidence, etc., collected for 6 months pre-T0 and 6 months post-T1.

Qualitative data

Sampling and Participants: At post-T2, 16 nurses from the intervention group were purposively sampled using maximum variation sampling based on LCJR score change (high gain: n = 8; low gain: n = 8), seniority (junior: n = 8; senior: n = 8), and subspecialty group membership. This approach ensured diverse perspectives. Individual semi-structured interviews were conducted.

Interview Guide Development: The semi-structured interview guide was developed by the research team based on the study’s theoretical framework (CLT and experiential learning) and the quantitative findings. The guide was pilot-tested with two ICU nurses not involved in the main study, with minor wording adjustments made for clarity. The full interview guide is provided as Supplementary Material.

The interview guide included the following details:

• How did you think of the new model?

• How did subspecialty grouping and mind maps help build your knowledge system?

• How did the app and rolling quizzes affect your learning habits and effectiveness?

• Was the performance linkage a motivator or pressure? Please elaborate.

• Strengths and areas for improvement of the model?

Data Collection: Individual interviews were conducted in a private room within the hospital by a trained researcher not involved in the intervention delivery. Each interview lasted 30–45 minutes, was audio-recorded, and transcribed verbatim. Data saturation was determined by consensus among the research team that no new themes emerged after 14 interviews; two additional interviews were conducted to confirm saturation.

Data Management and Anonymization: All transcripts were anonymized using participant codes (N1–N16). Identifying information was removed. Data were stored on a password-protected university server accessible only to the research team.

Translation: Interviews were conducted in Chinese. For publication, quotes were translated into English using a forward-backward translation process: a bilingual researcher performed initial translation, followed by back-translation by a second independent translator, with discrepancies resolved through team discussion. Minor adjustments were made to enhance readability while preserving meaning.

Data Analysis: Colaizzi’s 7-step phenomenological method was applied independently by two researchers:

1. Reading all transcripts to gain a sense of the whole

2. Extracting significant statements

3. Formulating meanings from significant statements

4. Clustering meanings into themes

5. Developing an exhaustive description

6. Identifying the fundamental structure

7. Returning to participants for member checking (conducted with 5 participants, who confirmed the themes)

Disagreements between the two researchers were resolved through discussion; when consensus could not be reached, a third senior researcher was consulted.

Statistical analysis

Data were analysed using IBM SPSS statistical software version 25.0 and R software (version 4.2.0) for mixed-effects models. Descriptive statistics were used to summarise the demographic data, presenting categorical variables as frequencies (n) and proportions (%), while continuous variables were described using means with standard deviations. The t-test or a Mann-Whitney U test or chi-square test was used to compare the difference between two groups. The intra-group comparisons were conducted with a repeated measures ANOVA. p < 0.05 was considered statistically significant. Qualitative data were analyzed using Colaizzi’s 7-step method by two researchers independently, with cross-checking of themes.

Accounting for Clustering: Given the cluster randomized design, we accounted for clustering in the following ways:

1. Intracluster Correlation Coefficient (ICC): The ICC for the primary outcome (LCJR total score) was calculated as 0.016, indicating low correlation within clusters.

2. Design Effect: The design effect was calculated as 1 + (average cluster size − 1)× ICC = 1+ (49 - 1)×0.016 = 1.77.

3. Effective Sample Size: The effective sample size was calculated as total sample size / design effect = 196 /1.77≈ 111.

4. Post-hoc Power Analysis: Using the effective sample size, a post-hoc power analysis for the primary outcome (LCJR) with a two-sample t-test, α = 0.05, and the observed effect size (Cohen’s d = 1.71) showed achieved power > 0.99.

5. Primary Analysis: In addition to t-tests, we re-analyzed the primary outcome using linear mixed-effects models with group as a fixed effect and unit as a random effect, which yielded consistent results.

Qualitative data were analyzed using Colaizzi’s 7-step method by two researchers independently, with cross-checking of themes as described in Section “Evaluation indicators”.

Results

Quantitative results

All 196 nurses completed the study (Intervention: n = 102; Control: n = 94).

Baseline data comparison

No significant differences were found in age, gender, education, or professional title between groups at baseline (p > 0.05).

Primary outcome comparison

As shown in Table 1 No significant difference in LCJR scores existed at T0 (t = −0.732, p = 0.465). At T1, the intervention group showed significantly higher total and dimensional LCJR scores than the control group (Total: t = 12.917, p < 0.001).

Table 1.

Comparison of Lasater clinical judgment Rubric (LCJR) scores between the two groups before and after intervention

Dimension	Time Point	Intervention Group (n = 102)	Control Group (n = 94)	t-value	p-value	Cohen’s d
Noticing	T0	7.21 ± 1.45	7.38 ± 1.52	−0.823	0.412	−0.12
	T1	9.05±0.89	7.65±1.41	8.672	<0.001	1.18
Interpreting	T0	6.92 ± 1.61	7.11 ± 1.57	−0.891	0.374	−0.12
	T1	9.22±0.76	7.42±1.50	10.752	<0.001	1.51
Responding	T0	6.85 ± 1.70	6.97 ± 1.63	−0.514	0.608	−0.07
	T1	9.18±0.81	7.28±1.59	10.787	<0.001	1.52
Reflecting	T0	7.05 ± 1.58	6.88 ± 1.66	0.761	0.448	0.11
	T1	8.95±0.92	7.05±1.60	10.202	<0.001	1.44
Total Score	T0	28.03 ± 5.12	28.34 ± 5.45	−0.423	0.673	−0.06
	T1	36.40±2.58	29.40±5.21	12.917	<0.001	1.71

Note: T0: Pre-intervention; T1: Post-intervention (6 months). Bold text highlights the intervention group’s post-intervention scores

Secondary outcomes comparison

1. Theoretical and Skill Scores: Repeated measures ANOVA showed a significant time*group interaction effect (F = 185.632, p < 0.001). Simple effects analysis (Table 2) indicated: no significant difference at T0 (p > 0.05); significantly higher scores for the intervention group at T1 and T2 (both p < 0.001); no significant decline from T1 to T2 within the intervention group (p = 0.102), but a significant decline in the control group (p < 0.001), indicating higher knowledge retention in the intervention group.

2. Scale Scores and Knowledge Retention Rate: As shown in Table 3, no significant differences in GSES and CTDI-CV scores at T0 (both p > 0.05). At T1, the intervention group had significantly higher scores on both scales (both p < 0.001). The knowledge retention rate ((T2 score / T1 score) * 100%) was significantly higher in the intervention group (98.57% vs. 93.08%, p < 0.001).

Table 2.

Comparison of theoretical and skill scores at different time points between the two groups

Group	n	T0 (Pre-intervention)	T1 (Immediate Post-test)	T2 (Delayed Post-test)	Intra-group Comparison (T1 vs T2) p-value
Intervention Group	102	72.35 ± 8.41	90.26±5.37	88.97±5.89	0.102
Control Group	94	71.89 ± 9.12	75.44 ± 8.15	70.21 ± 8.94	<0.001
t-value (at T1)	-	0.371	14.917	19.245	-
p-value (at T1)	-	0.711	<0.001	<0.001	-

Note: T0: Pre-intervention; T1: Post-intervention (6 months); T2: 3 months after intervention (Delayed post-test). Bold text highlights the intervention group’s post-intervention scores

Table 3.

Comparison of Scale scores and knowledge retention rate between the two groups before and after intervention

Indicator	Time Point	Intervention Group (n = 102)	Control Group (n = 94)	t-value	p-value
GSES (points)	T0	24.56 ± 5.13	25.01 ± 4.87	−0.634	0.527
	T1	30.45±4.22	25.88 ± 4.95	7.183	<0.001
CTDI-CV (points)	T0	278.32 ± 24.15	281.05 ± 22.76	−0.845	0.399
	T1	315.67±18.43	285.12 ± 23.51	10.214	<0.001
Knowledge Retention Rate (%) (T2/T1)	(T2/T1)	98.57±3.25	93.08 ± 5.71	8.432	<0.001

Note: T0: Pre-intervention; T1: Post-intervention (6 months); GSES: General Self-Efficacy Scale; CTDI-CV: Critical Thinking Disposition Inventory-Chinese Version. Knowledge Retention Rate = (T2 score / T1 score) × 100%. Bold text highlights the intervention group’s post-intervention scores and retention rate

Qualitative results

To deeply explore how the closed-loop model enhanced nurses’ abilities, 16 nurses from the intervention group were interviewed. Colaizzi’s analysis identified four core themes (Table 4), revealing the internal mechanisms and drivers, which resonated with and explained the quantitative findings.

Table 4.

Analysis of qualitative themes from interviews on the ‘Closed-loop training model’ experience among intervention group nurses (n = 16)

Theme	Sub-theme/Connotation	Frequency (n = 16)	Supporting Quotations	Link to Quantitative Results/Mechanistic Explanation
1. Reconstruction of Cognitive Schemata: From Fragments to Navigation	Knowledge Structuring: Mind maps connect scattered knowledge points into a systematic network. Mental Simulation: Rehearsing clinical decision-making paths mentally enhances confidence. Deep Understanding: Progressing from “knowing what” to “understanding why and how”.	15 (93.75%)	N11 (Senior Nurse): “Before, managing shock was a tangled mess in my mind. Now the mind map is like a clear ‘navigation map’, showing me step one: check volume, step two: assess vascular tone, step three: consider cardiac function … I feel confident when facing real patients.” N3 (Junior Nurse): “Creating the blood gas analysis mind map forced me to understand the meaning behind every arrow (compensation), not just memorize results. This allows me to reason possible causes when seeing abnormal reports.”	Explains LCJR improvement: This theme directly corresponds to the enhancement in the “Interpreting” and “Responding” dimensions of clinical judgment. The structured knowledge schema reduces working memory load, allowing nurses to allocate cognitive resources quickly to analysis and decision-making, thereby demonstrating more systematic and confident management in simulated cases.
2. Transformation of Learning Identity: From Passive Recipient to Active Co-creator	Active Inquiry: Proactively consulting literature and guidelines to prepare for quizzes and create maps. Learning through Teaching: Core members deepen reflection and integration of their own knowledge through the processof “teaching”. Team Collaboration: Collaborative mind mapping within groups promotes knowledge exchange and collective wisdom.	14 (87.5%)	N7 (Subspecialty Leader): “To clearly explain CRRT’s coagulation mechanism to everyone, I had to reorganize my years of experience and consult much new literature. Teaching others is the best way to learn; I broke through many knowledge gaps myself.” N14 (Rotating Member): “I’m not a core member, but to avoid being unable to answer during quizzes, I actively look at the maps on the APP and discuss cases with senior nurses in the group. It feels like learning has changed from ‘have to learn’ to ‘I want to learn’.”	Explains GSES/CTDI-CV improvement: This active, co-creative learning mode significantly enhanced nurses’ self-efficacy (GSES) and the “truth-seeking” and “analytical ability” dimensions of critical thinking (CTDI-CV). The increased sense of control is a core source of intrinsic motivation.
3. Activation of Metacognition: Questioning as a ‘Whetstone’ for Thinking	Self-Monitoring: The pressure of questioning forces continuous self-assessment: “Do I truly understand this point?” Immediate Feedback: On-the-spot correction after incorrect or incomplete answers provides the most effective learning moments. Contextual Application: Scenario-based questions bridge the “last mile” between theoretical knowledge and clinical practice.	16 (100%)	N9 (Charge Nurse): “The rolling quizzes act like a metacognition activator. It forces you to constantly self-examine, leaving no room for ‘vague understanding’; you must be very clear. That time I was asked about managing a rare complication and didn’t answer fully, but after the examiner’s explanation, I will never forget that case.” N2 (Head Nurse): “We design questions like ‘What would you do if the patient presented with XX situation?’. The purpose is not to fail them, but to force them to ‘think’ and ‘use’, not just ‘memorize’.”	Explains Knowledge Retention Rate: This theme perfectly embodies the “Testing Effect” or “Retrieval Practice”, the most effective method for strengthening long-term memory. Immediate feedback accelerates the correction of misconceptions and the consolidation of correct schemas, directly explaining the excellent performance of the intervention group in the delayed post-test.
4.The Double-Edged Sword of Sustainability: Performance Incentives and Implementation Challenges	Positive Pressure: Performance linkage is widely recognized as an effective “driver”, increasing importance attached and engagement). Implementation Cost: The significant time and energy investment by core members is a major implementation barrier. Need for Optimization: Highlights the demand for standardization and rotation systems regarding the fairness of questioning and quality control of mind maps.	16 (100%)	N15 (Nurse): “Linking to performance definitely makes everyone more engaged; no one wants to lose points for not answering. It’s a positive kind of tension that drives learning.” N5 (Core Member): “The biggest challenge is time. Besides my own clinical work, preparing lessons, creating maps, and preparing questions add a very heavy burden. It would be more sustainable if this was counted as teaching hours or had rewards.”	Informs Limitations and Practical Implications: This theme does not directly explain quantitative results but provides crucial insights into the real-world scalability and sustainability of the model. It affirms the effectiveness of performance incentives (extrinsic motivation) but also candidly points out core issues (time burden, standardization) that need addressing for long-term operation, offering vital “lessons learned” for other hospitals seeking to replicate this model.

Theme 1: reconstruction of cognitive schemata—from “fragmented memory” to “decision-making navigation”

Participants reported that the model significantly changed their knowledge organization and retrieval. Traditional training left “memory fragments,” whereas mind maps transformed complex guidelines into clear “mental maps” or “decision-making navigation.” A senior nurse (N11) stated: “Previously, managing shock felt overwhelming. Now the mind map acts like a ‘navigation map,’ visually outlining the path from assessment (volume, vascular tone, cardiac function) to intervention, allowing for methodical action even under pressure.” A junior nurse (N3) added: “Creating the blood gas analysis mind map forced me to understand the meaning behind every arrow (compensation), not just memorize results. This allows me to reason about possible causes when seeing abnormal reports.”

Theme 2: transformation of learning identity—from “passive recipient” to “active Co-creator”

Many participants experienced a shift in learning identity. The subspecialty group model fostered ownership. A core member (N7) explained: “To clearly explain CRRT’s coagulation mechanism to everyone, I had to reorganize my years of experience and consult new literature. Teaching others is the best way to learn; I broke through many knowledge gaps myself.” A rotating member (N14) noted: “I’m not a core member, but to avoid being unable to answer during quizzes, I actively look at the maps on the app and discuss cases with senior nurses in the group. It feels like learning has changed from ‘have to learn’ to ‘I want to learn.’”

Theme 3: sustained activation of metacognition—questioning as a “whetstone” for thinking

All participants found the rolling quizzes challenging yet effective. A charge nurse (N9) described: “The rolling quizzes act like a metacognition activator. It forces you to constantly self-examine, leaving no room for ‘vague understanding’; you must be very clear. That time I was asked about managing a rare complication and didn’t answer fully, but after the examiner’s explanation, I will never forget that case.” A head nurse (N2) explained the design intent: “We design questions like ‘What would you do if the patient presented with XX situation?’ The purpose is not to fail them, but to force them to ‘think’ and ‘use,’ not just ‘memorize.’”

Theme 4: The Double-Edged Sword of Sustainability—Performance Incentives and Implementation Challenges

Participants acknowledged the motivational effect of performance linkage while noting challenges. A nurse (N15) stated: “Linking to performance definitely makes everyone more engaged; no one wants to lose points for not answering. It’s a positive kind of tension that drives learning.” However, a core member (N5) pointed out: “The biggest challenge is time. Besides my own clinical work, preparing lessons, creating maps, and preparing questions add a very heavy burden. It would be more sustainable if this was counted as teaching hours or had rewards.”

Discussion

This study systematically evaluated the effects of a CLT-based closed-loop training model on ICU nurses’ specialized training. Quantitative results showed there were significant improvements in the intervention group’s clinical judgment (LCJR), theoretical/practical scores (immediate and delayed), self-efficacy (GSES), and critical thinking (CTDI-CV) compared to traditional training. Qualitative findings further revealed the mechanisms: reconstructing cognitive schemata, transforming learning identity, activating metacognition, and providing performance-driven motivation.

Significant enhancement of clinical decision-making ability by the closed-loop Model

Optimizing Cognitive Load is Key. The core finding was the significant improvement in LCJR scores, reflecting advanced abilities in insight, understanding, decision-making, and reflection within complex clinical situations [18]. This effectiveness stems from the systematic application of CLT, optimizing cognitive resource allocation. Subspecialty grouping reduced intrinsic load by modularizing complex knowledge [20], aligning with findings that specialized training reduces cognitive overload in novices [21]. Mind maps reduced extraneous load by visually presenting knowledge structures [5], supporting Sweller’s findings [22] that optimizing information presentation reduces extraneous load. Rolling quizzes increased germane load through retrieval practice and application [12], forcing knowledge activation and schema building. Thus, the combined “reduce intrinsic-reduce extraneous-increase germane” strategy laid a solid cognitive foundation for enhancing clinical decision-making.

Efficient knowledge retention and transfer

The Triumph of Retrieval Practice and Spacing Effect. The significantly higher knowledge retention rate at the delayed post-test demonstrates the model’s advantage in promoting long-term memory. This is attributable to retrieval practice (testing effect) and the spacing effect [23]. Unlike one-off traditional training, rolling quizzes created a continuous, spaced retrieval process. Monthly quizzes based on previous content forced active recall from long-term memory, a process proven more effective than rereading [24]. The qualitative theme “activation of metacognition” exemplifies that uiz-induced pressure prompted self-monitoring and review and strengthened memory traces greatly, which were highly consistent with Karpicke and Blunt’s findings [25].

Concurrent enhancement of self-efficacy and critical thinking

Driven by identity shift and success experiences, improvements in GSES and CTDI-CV indicate the model fostered not only knowledge but also confidence and thinking habits. Self-efficacy gains stem from mastery experiences (successfully handling challenging quizzes, making correct decisions in simulations) and vicarious experiences (observing peers and discussing within groups), emphasized by Bandura’s social cognitive theory [19]. Similar findings on the shift from passive to active learning identity have been reported in studies of junior doctors undergoing evidence-based medicine training [26]. “The interview quote “learning changed from ‘being told to learn’ to ‘I want to learn’” signals this active identity shift. Critical thinking improvement is tied to the model’s design: case-based teaching in subspecialties and scenario-based quizzes inherently train high-order skills like analysis, reasoning, evaluation, and decision-making [6], internalizing critical thinking from an ability into a disposition, reflecting in increased CTDI-CV dimension scores.

Research Implications, limitations, and Future directions

Theoretical Implications: This study translates CLT from abstract principles into concrete and operable nursing training practices, validating its effectiveness in complex medical vocational training and providing empirical evidence from the Chinese nursing context. Practical Implications: It provides nursing managers a validated and replicable new training model. Particularly, directly linking performance management to training effects provides an excellent solution to the pitfalls of formalized and hard-to-sustain in-service training, greatly enhancing sustainability and impact.

Several limitations should be acknowledged. First, the cluster randomized design with only four units (two per group) limits statistical power and increases vulnerability to unit-level confounding. Although we accounted for clustering in our analysis, future multi-center studies with more clusters are needed to confirm these findings. Second, due to the nature of the intervention, blinding of participants and trainers was not possible, potentially introducing performance and detection bias. However, outcome assessors were blinded to group allocation to mitigate this. Third, the performance linkage (bonuses, awards) may have introduced extrinsic motivation that could have inflated engagement beyond the intrinsic educational effects of the model. Future studies should evaluate the intervention with and without performance incentives to isolate the educational effects. Fourth, the self-designed knowledge and skill tests, while validated, may not fully capture real-world clinical performance. Fifth, the study was conducted at a single tertiary teaching hospital in China, which may limit generalizability to other settings. Sixth, the Hawthorne effect—where participants may have changed behavior due to awareness of being observed—cannot be ruled out.

Future Directions: Future research should: 1) Conduct multi-center studies to verify scalability; 2) Perform longer-term follow-up to assess impact on ultimate patient outcomes (e.g., mortality, complication rates); 3) Further optimize the digital implementation of rolling quizzes (e.g., daily app pushes) to reduce organizational burden.

Integration of quantitative and qualitative findings

The explanatory sequential mixed-methods design allowed us to use qualitative findings to explain and deepen the quantitative results.

First, the significant improvement in LCJR scores (t = 12.917, p < 0.001) was illuminated by Theme 1 (Reconstruction of Cognitive Schemata). Participants described how mind maps transformed fragmented knowledge into structured “navigation maps,” reducing cognitive load during decision-making. This qualitative insight provides a mechanistic explanation for the quantitative improvement, supporting the CLT-based hypothesis that reducing extraneous load enhances clinical judgment.

Second, the improvements in GSES (t = 7.183, p < 0.001) and CTDI-CV (t = 10.214, p < 0.001) scores were explained by Theme 2 (Transformation of Learning Identity). The shift from passive recipient to active co-creator, described by core and rotating members alike, suggests that increased ownership and intrinsic motivation drove self-efficacy and critical thinking dispositions.

Third, the superior knowledge retention rate in the intervention group (98.6% vs. 93.1%, p < 0.001) was directly explained by Theme 3 (Activation of Metacognition). Participants described how the pressure of rolling quizzes prompted continuous self-monitoring and reinforced memory traces—a perfect illustration of the retrieval practice effect [25].

Fourth, Theme 4 (The Double-Edged Sword of Sustainability) did not directly explain a quantitative outcome but provided crucial context for interpreting the results. While performance incentives were acknowledged as effective motivators, the time burden highlighted by core members suggests that the observed effects may be partially attributable to extrinsic motivation, and long-term sustainability requires addressing implementation costs.

Conclusion

This study successfully developed and validated a CLT-framed closed-loop training model for ICU nurses. By systematically reorganizing content (subspecialty), visually presenting knowledge (mind maps), contextually testing application (rolling quizzes), and linking to incentive mechanisms,the model demonstrated significant improvements in clinical decision-making, knowledge retention, self-efficacy, and critical thinking compared to traditional training. Qualitative findings revealed the underlying mechanisms: reconstruction of cognitive schemata, transformation of learning identity, and activation of metacognition. While these findings are promising, the model should be further tested in larger, multi-center trials with more clusters to establish its generalizability and to isolate the specific contributions of each component.This model serves as a powerful tool for transforming ICU nurses from “skill operators” to “thinking decision-makers,” holding significant value for clinical promotion.

Author contributions

Yao Ning Zhuang: Writing – review & editing, Writing – original draft, Visualization, Resources, Project administration, Methodology Funding acquisition, Formal analysis, Data curation, Conceptualization. Zhen Mei Xu and Dan Dan Lin: Writing – review & editing, Writing – original draft, Investigation. Qing Qing Zhan: Resources, Methodology, Investigation, Funding acquisition, Conceptualization.YI Yuan Chen: Writing – review & editing, Funding acquisition, Conceptualization. Bin Qin Lin: Writing – review & editing, Investigation, Funding acquisition, Conceptualization. Shu Fen Ke: Writing – review & editing, Writing – original draft, Methodology, Investigation, Funding acquisition, Formal analysis, Conceptualization.

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Data availability

The data that support the findings of this study are available within the article.

Declarations

Ethical review

This study was approved by the Ethics Committee of the Affiliated Hospital of Putian University (Approval No: 2,025,313). Data were anonymized and all participants provided informed consent.

Ethics approval and consent to participate

This study was conducted in accordance with the Declaration of Helsinki and was approved by the Ethics Committee of Affiliated Hospital of Putian University (Approval No: 2,025,313). Written informed consent was obtained from all participants.The participants provided their consent to participate in the study.

Consent for publication

Not applicable. No identifying images or personal details of participants are presented in this manuscript.

Competing interests

The authors declare no competing interests.