Skip to contents

Quick Guide: Diagnostic Style Clustering in jamovi

ClinicoPath Team

2025-06-30

Source: vignettes/meddecide-14-diagnostic-style-quick-guide.Rmd
meddecide-14-diagnostic-style-quick-guide.Rmd

🚀 Quick Start: 5-Minute Diagnostic Style Analysis

Step 1: Prepare Your Data

Required columns: - Rater columns: Each pathologist’s diagnoses (e.g., Path_01, Path_02, etc.) - Case ID: Unique identifier for each case

Optional columns: - Experience: Years of experience or level (Junior/Senior) - Training: Training institution - Institution: Current practice location - Specialty: Subspecialty focus - True diagnosis: Gold standard (for accuracy analysis)

Step 2: Load Data in jamovi

  1. Open jamovi
  2. Import your CSV file
  3. Ensure rater columns are set as Factor (nominal/ordinal)

Step 3: Run Basic Analysis

  1. Go to meddecide → Agreement → Interrater Reliability
  2. Move rater columns to Raters/Observers box
  3. Click Run

✅ Check: Overall agreement should be 40-90% for meaningful style analysis

Step 4: Enable Diagnostic Style Clustering

  1. Check ✓ Diagnostic Style Clustering (Usubutun Method)
  2. Accept defaults:
    • Method: Ward’s Linkage
    • Distance: Percentage Agreement
    • Groups: 3

✅ Expected: See Style Groups (1, 2, 3) assigned to each pathologist

Step 5: Add Pathologist Characteristics

  1. Check ✓ Include Rater Characteristics
  2. Select your characteristic variables:
    • Experience Variable → your experience column
    • Training Institution → your training column
    • Current Institution → your institution column
    • Specialty → your specialty column

✅ Expected: Style groups correlate with characteristics

Step 6: Identify Problem Cases

  1. Check ✓ Identify Discordant Cases

✅ Expected: List of cases causing style-based disagreement

📊 Interpreting Results

Diagnostic Style Table

Rater Style Group Within-Group Agreement Experience Training
Path_01 Style 1 85% Senior Academic
Path_02 Style 1 83% Senior Academic
Path_03 Style 2 78% Junior Community

🔍 What to look for: - High within-group agreement (>80%): Strong style membership - Similar characteristics within groups: Validates clustering - Clear separation between groups: Distinct diagnostic approaches

Style Summary Table

Style Group Members Avg Agreement Predominant Training
Style 1 6 84% Academic_Eastern
Style 2 5 79% Community
Style 3 4 87% Academic_Western

🔍 Interpretation: - Style 1: Conservative academics - Style 2: Moderate community pathologists
- Style 3: Aggressive specialists

Discordant Cases

Case Discord Score Style 1 Diagnosis Style 2 Diagnosis Style 3 Diagnosis
Case_042 0.85 Benign Atypical Atypical
Case_067 0.78 Atypical DCIS DCIS

🔍 Meaning: - High discord (>0.7): Cases revealing diagnostic philosophy differences - Pattern: Conservative vs. aggressive diagnostic tendencies

🎯 Common Applications

Quality Assurance

Goal: Identify diagnostic outliers

Setup: - Include all department pathologists - Use routine sign-out cases - Focus on challenging diagnoses

Action Items: - Pathologists with unusual styles → additional training - Cases with high discord → consensus review - Style patterns → standardized protocols

Training Evaluation

Goal: Assess resident diagnostic development

Setup: - Compare residents vs. attendings - Track individual residents over time - Include mentor information

Action Items: - Residents clustering with wrong style → mentor reassignment - Persistent style differences → targeted education - Mentor influence patterns → faculty development

Consensus Development

Goal: Improve diagnostic guidelines

Setup: - Select expert panel representing different styles - Include cases showing style disagreements - Document reasoning for style differences

Action Items: - Style-specific guidelines → comprehensive protocols - High-discord cases → educational materials - Expert disagreements → research priorities

⚙️ Advanced Settings

Distance Metrics

Percentage Agreement (Default) - ✅ Use for: Most categorical diagnoses - 📊 Measures: Direct diagnostic concordance

Correlation - ✅ Use for: Ordinal scales (grades, stages) - 📊 Measures: Linear relationship patterns

Euclidean - ✅ Use for: Quantitative scores - 📊 Measures: Geometric distance

Clustering Methods

Ward’s Linkage (Default) - ✅ Use for: Compact, similar-sized groups - 🎯 Creates: Tight, meaningful clusters

Complete Linkage - ✅ Use for: Well-separated distinct groups - 🎯 Creates: Very tight clusters

Average Linkage - ✅ Use for: Moderate separation - 🎯 Creates: Balanced clusters

Number of Groups

2 Groups: Conservative vs. Aggressive 3 Groups: Conservative, Moderate, Aggressive (Usubutun standard) 4+ Groups: Subspecialty-specific approaches

🔧 Troubleshooting

Problem: All pathologists in one group

Causes: - Too few pathologists (<6) - Very high agreement - Cases too easy/difficult

Solutions: - Add more pathologists - Include challenging cases - Reduce groups to 2

Problem: Groups don’t match characteristics

Causes: - Characteristics don’t influence style - Small sample size - Hidden confounders

Solutions: - Add more characteristic variables - Increase case numbers - Consider interaction effects

Problem: Poor within-group agreement

Causes: - Wrong number of groups - High random error - Inappropriate cases

Solutions: - Adjust group number - Focus on moderate difficulty cases - Exclude technical failures

📚 Test Datasets

ClinicoPath includes three test datasets:

Endometrial Pathology

  • File: endometrial_diagnostic_styles.csv
  • Cases: 80 endometrial biopsies
  • Pathologists: 15 (various backgrounds)
  • Categories: Benign, EIN, Carcinoma
  • Expected Styles: 3 groups correlating with training

Breast Pathology

  • File: breast_diagnostic_styles.csv
  • Cases: 60 breast biopsies
  • Pathologists: 12 (breast specialists vs. generalists)
  • Categories: Benign, Atypical, DCIS, Invasive
  • Expected Styles: Conservative, Moderate, Aggressive

Lymphoma Classification

  • File: lymphoma_diagnostic_styles.csv
  • Cases: 45 lymphoid lesions
  • Pathologists: 10 hematopathologists
  • Categories: Reactive, DLBCL, Follicular, Marginal Zone, Mantle Cell
  • Expected Styles: WHO-strict, Molecular-heavy, Morphology-first

📖 Key References

Primary Diagnostic Style Literature

  1. Usubutun A, Mutter GL, Saglam A, Dolgun A, Ozkan EA, Ince T, Akyol A, Bulbul HD, Calay Z, Eren F, Gumurdulu D, Haberal AN, Ilvan S, Karaveli S, Koyuncuoglu M, Muezzinoglu B, Muftuoglu KH, Ozdemir N, Ozen O, Baykara S, Pestereli E, Ulukus EC, Zekioglu O. (2012). Reproducibility of endometrial intraepithelial neoplasia diagnosis is good, but influenced by the diagnostic style of pathologists. Modern Pathology, 25(6), 877-884. doi: 10.1038/modpathol.2011.220. PMID: 22301705.

  2. Elmore JG, et al. (2015). Diagnostic concordance among pathologists interpreting breast biopsy specimens. JAMA, 313(11), 1122-1132.

  1. Sterlacci W, Fiegl M, Juskevicius D, Tzankov A. (2020). Cluster Analysis According to Immunohistochemistry is a Robust Tool for Non-Small Cell Lung Cancer and Reveals a Distinct, Immune Signature-defined Subgroup. Applied Immunohistochemistry & Molecular Morphology, 28(4), 274-283. PMID: 31058655.

  2. Olsen SH, Thomas DG, Lucas DR. (2006). Cluster analysis of immunohistochemical profiles in synovial sarcoma, malignant peripheral nerve sheath tumor, and Ewing sarcoma. Modern Pathology, 19(5), 659-668. PMID: 16528378.

  3. Matsuoka T, Mitomi H, Fukui N, Kanazawa H, Saito T, Hayashi T, Yao T. (2011). Cluster analysis of claudin-1 and -4, E-cadherin, and β-catenin expression in colorectal cancers. Journal of Surgical Oncology, 103(7), 674-686. PMID: 21360533.

  4. Carvalho JC, Wasco MJ, Kunju LP, Thomas DG, Shah RB. (2011). Cluster analysis of immunohistochemical profiles delineates CK7, vimentin, S100A1 and C-kit (CD117) as an optimal panel in the differential diagnosis of renal oncocytoma from its mimics. Histopathology, 58(2), 169-179. PMID: 21323945.

💡 Tip: Start with the test datasets to learn the interface, then apply to your own data. The diagnostic style clustering reveals hidden patterns in pathologist decision-making that traditional agreement statistics miss.