Proposed CRAFT model organizes clinical-reasoning instruction into five functions

A theory-informed instructional design model proposes a coordinated approach to teaching clinical reasoning through five recurring functions: Conceptualize and Frame, Reveal and Model Clinical Reasoning, Apply Through Scaffolded Practice, Feedback and Reflect, and Transfer Toward Independence. Known as CRAFT, the framework is intended to connect decisions about cases, instructional support, feedback, progression, and transfer across a curriculum. It is a hypothesis-and-theory article rather than a learner intervention: no cohort data were collected, and the authors report no evidence that the model improves clinical-reasoning performance. Its usability, feasibility, mechanisms, incremental value, and boundary conditions remain to be tested empirically.

Clinical reasoning is described in the article as iterative, context-sensitive work: clinicians gather and interpret information, construct problem representations, generate and prioritize diagnostic and management possibilities, act, and revise decisions as new evidence emerges. The demands vary with the clinical task, organized knowledge, case features, prior experience, uncertainty, team and system conditions, and a learner’s ability to monitor and adapt. Teaching this process is a separate design problem involving case selection and sequencing, making expert performance visible, distributing and withdrawing support, timing feedback, and creating opportunities for increasingly independent performance. Assessment is a third level: it samples performance to guide feedback, progression, or curriculum evaluation but does not by itself specify an instructional sequence.

The proposal addresses an uneven and fragmented implementation landscape for clinical-reasoning education. A national survey of internal medicine clerkship directors found that 57% of responding institutions lacked sessions dedicated to clinical-reasoning concepts; limited curricular time, reported by 88% of respondents, and limited faculty expertise, reported by 69%, were the most frequent barriers. A recent scoping review described wide variation in curricular theories, teaching methods, and assessments and found no all-encompassing design across the included health-professions curricula. A 2025 systematic review and meta-analysis found that explicit instruction was generally beneficial, especially during clinical years, while also identifying few multiyear studies that deliberately scaffolded the same cohort across a full program. These findings describe gaps and variation in the available literature, not tests of CRAFT.

CRAFT integrates several theoretical traditions for particular design decisions. Elaboration theory contributes the idea of an “epitome,” an authentic but simplified case that preserves the structure of the whole reasoning task. The Four-Component Instructional Design model contributes task classes, supportive information, just-in-time procedural guidance, varied practice, and systematic fading. Cognitive apprenticeship adds modeling, coaching, articulation, reflection, and exploration; situated cognition supports planned variation in patient, team, resource, emotional, and workflow conditions. Self-regulated learning and deliberate practice emphasize goals, monitoring, repeated performance, feedback, and revision, while cognitive load theory guides decisions about complexity and support. The article stresses that no single source supplies the entire model and that the synthesis itself does not establish effectiveness.

The proposed model resolves tensions among these approaches through four precedence rules. Safety and prerequisite knowledge take precedence over realism. Evidence of performance takes precedence over a calendar-based schedule for withdrawing support. Task-relevant complexity takes precedence over realism pursued for its own sake, and human oversight takes precedence over automated instructional convenience. A realistic case may therefore need simplification if its interacting details prevent a novice from constructing an interpretable representation. Support should remain when withdrawing it could lead to a safety-critical omission, even if the learner has completed the scheduled number of cases. The model also distinguishes intrinsic task complexity, extraneous demands, and learning-relevant processing; because a scaffold can become redundant as expertise grows, fading is tied to observed performance rather than to the passage of time alone.

The five functions are distinguished by their dominant purpose and by who performs the principal cognitive work, not by a rigid chronology. During Conceptualize and Frame, learners organize knowledge and build a concise representation of an interpretable task, often beginning with a simplified whole case. During Reveal and Model Clinical Reasoning, an experienced clinician or calibrated near-peer makes consequential cues, uncertainty, probability shifts, priorities, and revisions visible through think-alouds and contrastive examples. The learner is asked to annotate, explain, predict, or partially complete the reasoning rather than merely watch. In Apply Through Scaffolded Practice, the learner commits to a representation, prioritized differential, information requests, and plan before receiving definitive feedback, using targeted worksheets, completion tasks, cueing questions, or coaching when needed.

Feedback and Reflect shifts the emphasis from performing the task to analyzing and revising reasoning. Feedback compares performance with explicit criteria or a reference path and may come from faculty, peers, or locally validated digital tools. Learners are expected to distinguish knowledge gaps from process or contextual failures and specify a concrete adjustment for the next attempt. For example, identifying premature closure is not sufficient unless the learner states which evidence threshold or comparison strategy will change. Reflection should remain brief and criterion-referenced; the next case provides a test of whether the planned adjustment is used. If the same error pattern persists, educators can return to targeted modeling or supported practice rather than treating reflection itself as evidence of learning.

Transfer Toward Independence requires reasoning with substantially reduced support across unfamiliar, variable, or workplace-like cases while supervisors preserve patient safety. Complexity may involve atypical presentations, competing diagnoses, multiple problems, time pressure, interrupted information, resource constraints, and team coordination. Independence does not mean working without oversight or ordinary clinical resources: it includes knowing when to slow down, seek help, escalate a concern, or consult a trusted reference. Proposed evidence for this stage is triangulated across cases, occasions, contexts, and assessors, with attention to delayed transfer, consistency, calibration, critical omissions, and appropriate help-seeking. Failure in one contextual dimension may warrant a targeted return to an earlier CRAFT function without implying that all previously demonstrated competence has disappeared.

An illustrative sepsis case family shows how the functions might recur as learners progress. During an early cycle, learners interpret a prototypical deterioration with a static schema, staged clinician modeling, a worksheet, and immediate low-stakes debriefing. An intermediate cycle introduces atypical presentation, comorbidity, incomplete data, sepsis mimics, fewer prompts, and moderate-stakes case assessment. In an advanced cycle, learners lead assessment and management during dynamic shock, interruptions, handoffs, and resource constraints, with minimal instructional support, safety supervision, and post-event feedback. Progression would depend on stable performance across multiple cases, reduced prompt dependence, absence of critical omissions, contextual adaptation, and appropriate escalation. The authors describe this as an instructional design demonstration informed by Sepsis-3 and the Surviving Sepsis Campaign, not as a report of learner outcomes. Repeated exposure may also produce case familiarity mistaken for transfer, so educators would need additional diagnoses and cases with similar reasoning demands but different surface features.

Five falsifiable propositions set out possible causal sequences for future testing. Readiness matching is proposed to improve problem representations by aligning case complexity and support with prior knowledge. A visible-to-active transition from modeling to learner commitment is hypothesized to strengthen explanations of changing probabilities and priorities. Adaptive fading is proposed to reduce prompt dependence without increasing safety-critical errors, while a feedback-to-action loop is expected to make targeted reasoning changes visible in a subsequent case. Progressive recursion, which increases selected dimensions of complexity and responsibility across successive cycles while reducing unnecessary support, is hypothesized to improve performance on increasingly dissimilar cases. These are predictions, not measured effects. Each mechanism could fail if readiness criteria are poorly matched, modeling produces passive imitation, evidence for fading is untrustworthy, reflection becomes ritualized, or increases in difficulty exceed learners’ capacity.

Implementation would require shared terminology, a small set of common criteria, cross-course governance, faculty development, and reliable evidence about performance. The authors suggest first mapping existing activities to identify where each function is strong or absent, then piloting one case family or course sequence while documenting faculty time and learner workload, and scaling only after a sustainable case-production, feedback, faculty-development, and data-review process has been established. Assessment should sample problem representation, differential prioritization, use of evidence, management reasoning, response to uncertainty, critical omissions, and metacognitive calibration across multiple methods and cases rather than relying on one diagnosis or global impression. A feasible pilot could compare baseline, immediate, and delayed performance, but a contemporaneous comparator or other credible design would be required to distinguish CRAFT’s contribution from maturation, clinical exposure, repeated testing, or concurrent curriculum changes. Until such comparative studies establish added value, CRAFT remains an auditable proposal for coordinating established clinical-reasoning instruction rather than a proven curriculum model.

Subject of Research: A theory-informed instructional design model for coordinating clinical-reasoning education

Article Title: A five-function theory-informed instructional design model for developing clinical reasoning

Article References: Jung, E., Reigeluth, C. M., Zheng, B., Durning, S. J., & Merkebu, J. (2026). A five-function theory-informed instructional design model for developing clinical reasoning. Frontiers in Education, 11, Article 1861062. https://doi.org/10.3389/feduc.2026.1861062

Image Credits: AI Generated

DOI: 10.3389/feduc.2026.1861062

Keywords: clinical reasoning, medical education, instructional design, CRAFT model, scaffolding, cognitive apprenticeship, curriculum design, learner assessment, transfer, feedback, five-function, theory-informed