Advancing Problem-Based Learning with Clinical Reasoning for Improved Differential Diagnosis in Medical Education

Research Background and Issues

• What problems or challenges did the authors identify?
  Current medical education, particularly Problem-Based Learning (PBL), effectively promotes student learning but has limitations in practical application. The main issues include:

  1. Cases in PBL are typically selected by experienced instructors, leaving students unable to tailor their learning to individual needs.
  2. The fragmented presentation of information (e.g., paper-based data or slides) hinders students' ability to conduct comprehensive analyses.
  3. Insufficient mechanisms for recording and reviewing make it difficult for students to systematically revisit and reflect on their learning.
  4. While existing research enhances the immersive aspects of PBL, it neglects the construction of evidence-based logical chains and clinical reasoning methods.

• Why is this issue important?
  The primary goal of medical education is to train students to translate theoretical knowledge into practical skills. While PBL has shown promise in simulating real-world clinical diagnostic scenarios, its shortcomings limit students' self-directed learning and diagnostic reasoning development. Clinical diagnosis, particularly differential diagnosis, is a complex reasoning process where strong logical chains and evidence-based analytical skills are crucial for medical students.

• Research Motivation and Related Work
  The authors aim to address these gaps by strengthening PBL's logical reasoning and evidence-based diagnostic processes. Related work includes the use of Virtual Reality (VR) and AI to enhance PBL's immersive experience, as well as Clinical Decision Support Systems (CDSS) for diagnostic assistance. However, these tools primarily focus on data modeling or prediction accuracy rather than the diagnostic reasoning process in educational contexts.

Proposed Solution

• What methods or solutions did the authors propose?
  The authors designed and developed a PBL system called "e-MedLearn," which aims to support more efficient evidence-based applications and clinical reasoning training. The system integrates the following three core functionalities:

  1. Data Construction: Helps students select target cases and categorize them to support systematic learning pathways.
  2. Information Analysis: Provides AI-based interactive Question-Answering (QA) functionality to encourage students to engage in logical reasoning around case evidence.
  3. Record and Review: Organizes and records the analysis process using logical templates (e.g., diagnostic checklists and mind maps) to facilitate subsequent reflection and adjustment.

• What are the innovative aspects of this solution?

  1. Simulating real clinical reasoning with AI-driven QA, enabling students to uncover negative symptoms not explicitly mentioned in cases.
  2. Providing clear diagnostic logic support (e.g., FILA framework-based guiding questions) to emphasize evidence-based analysis.
  3. Systematic review functionality ensures students can clearly revisit their analytical processes rather than focusing solely on the correctness of answers.

• What are the implementation steps and key technologies used?
  The implementation plan unfolds in three stages:

  1. Data Construction Stage: A multi-level filtering mechanism helps students select appropriate cases based on their learning objectives.
  2. Information Analysis Stage: GPT-3.5 language model is used to generate interactive QA, supporting fact analysis and reasoning supplementation. Students update diagnostic checklists and mind maps to capture new evidence and hypotheses.
  3. Record and Review Stage: After completing the analysis, students review and summarize their diagnostic logic, identify errors, and develop continuous learning plans.

Research Outcomes

• What specific outcomes were achieved?

  1. Identified the primary needs of instructors and students in current PBL practices, including students' desire for more systematic analysis tools and review mechanisms.
  2. Developed and launched the e-MedLearn system, providing evidence-based and logical chain support.
  3. Conducted a controlled experiment with 19 medical students and in-depth interviews with 13 participants, demonstrating the system's advantages in clinical reasoning and learning efficiency.

• What are its advantages compared to existing solutions?
  Compared to traditional PBL educational methods or existing diagnostic support tools, e-MedLearn offers:

  1. A targeted case selection function.
  2. Enhanced ability for students to construct logical chains from facts to reasoning.
  3. A systematic review and reflection mechanism to promote long-term student improvement.

• What were the experimental or evaluation results?

  • Quantitative Results: Students using e-MedLearn outperformed the control group in multiple dimensions, such as system usability (5.33 vs. 4.78) and structured support for problem analysis (significant p-value < 0.01).
  • Task Performance: Students' case analysis reports under e-MedLearn scored higher in dimensions such as logical consistency, problem identification, and action planning, reflecting significant improvements in analytical skills.
  • Interview Summary: Users found the system's logical reasoning and evidence recording features helpful for developing diagnostic thinking but suggested adding more flexible interactive prompts in the future.

• Limitations and Future Directions

  1. Limitations:
     • The study involved a relatively small sample size, primarily from orthopedics and neurosurgery, without covering other medical disciplines.
     • The long-term learning effects on students require further tracking and validation.
  2. Future Directions:
     • Expand to more medical specialties to verify the system's generalizability.
     • Incorporate humanistic factors and treatment decision-making into the system design.
     • Build a medical education ecosystem that supports cross-institutional collaboration and enhances human-computer interaction feedback mechanisms.