DLBCL Disease Module

[mickohanlon23]

Module Describing Clinical Treatment of Diffuse Large B-Cell Lymphoma.
Starts from initial, and covers DLBCL symptoms, diagnosis, staging, and treatment. Chance of survival is assigned based upon diagnosed cancer stage, and alligns with 5-year DLBCL survival rates, as reported by NCI SEER. Module aligns with mCODE data elements where possible.

Primary sources for information gathering:

1. NCI SEER: https://seer.cancer.gov/statfacts/html/dlbcl.html
2. Lymphoma Research Foundation: https://lymphoma.org/
3. American Cancer Society: https://www.cancer.org/cancer/non-hodgkin-lymphoma/
4. Official Journal of the National Comprehensive Cancer Network: https://jnccn.org/
5. Conversation with Dr. William A. Wood, UNC hematologist and oncologist with a clinical focus in blood cancers
6. Conversation with Dr. Miguel-Angel Perales, MSKCC Chief, Adult Bone Marrow Transplant Service

The primary author of this module is Jenn Holmes ([email protected]).

[awatson1978]

1. Dual Treatment Outcome Logic (Design Question)

In dlbcl_stage_i_ii_treatment_round_one.json:

The module sets Treatment_1_Outcome twice:

After immediate treatment response (90% remission, 10% worsened)

• End PET-CT Scan Encounter → Patient is in remission after treatment → Set Remission Outcome
• Sets Treatment_1_Outcome = "Remission"

After long-term follow-up (months later)

• Check-ups → Check-Up → Blood tests → Patient stays in remission/Patient relapses
• Remission or Worsened states set Treatment_1_Outcome again

This overwrites the first outcome. The main module expects Treatment_1_Outcome to determine the flow, but it will only see the final value.

Questions:

Is this intentional (modeling late relapse)?
Should there be separate attributes (Treatment_1_Immediate vs Treatment_1_Final)?
Should we remove the dual setting?

2. Potential Logic Issue in Blood Tests

The complex_transition in Blood tests depends on DLBCL_Survival:

• If DLBCL_Survival = "No" → 100% relapse
• If DLBCL_Survival = "Yes" → 60% remission, 40% relapse

This seems backwards medically: patients with better survival prognosis should have lower relapse rates, not 40% relapse.

[awatson1978]

The Stage III/IV treatment module now properly models:

✅ Initial encounter and treatment selection
✅ R-CHOP medication cycles with proper start/stop
✅ Interim restaging after 2-4 cycles
✅ Treatment continuation decisions
✅ Final restaging and outcome assessment
✅ Long-term follow-up with relapse monitoring
✅ Clinical trial pathway for 10% of patients

Major Issues Resolved

✅ Clinical trial connection fixed - "Clinical Trial" now properly transitions to "Length of clinical trial"
✅ Blood test logic corrected - Good prognosis patients now have 70% remission / 30% relapse (medically realistic)
✅ Clean state flow - All encounters properly managed
✅ No duplicate states - All state names are unique
✅ No dead code - All states are reachable

Medical Accuracy

• Stage-appropriate treatment intensity - 6-8 cycles for advanced disease vs 3-6 for early stage
• Realistic clinical trial enrollment - 10% for high-stage patients
• Proper interim restaging - PET-CT after 2-4 cycles for treatment modification
• Appropriate outcome distributions - 90% initial response, then risk-stratified long-term outcomes

Technical Review

• Proper encounter management - All encounters opened and closed correctly
• Consistent attribute setting - Treatment_1_Outcome set appropriately
• Clean transitions - No orphaned or unreachable states
• Proper reason codes - All encounters reference "DLBCL_Condition"

Clinical Pathway Modeling

• Realistic treatment flow - Matches NCCN guidelines for advanced DLBCL
• Risk-stratified outcomes - Uses DLBCL_Survival for appropriate prognostic stratification
• Proper medication management - R-CHOP started and stopped correctly
• Long-term follow-up - Models late relapse patterns accurately

[awatson1978]

The Stage I/II treatment module now properly models:

✅ Blood test logic corrected - Good prognosis patients now have 80% remission / 20% relapse (even better than Stage III/IV's 70/30, which is medically appropriate for early-stage disease)
✅ Missing encounter reason added - Check-Up now properly references "DLBCL_Condition"
✅ Clean state flow - All encounters properly managed
✅ Proper encounter management - All encounters opened and closed correctly

[awatson1978]

Round Two Treatments

🔴 Distribution Validation Failures

"Patients_with_intention_to_proceed_to_transplant":

• Distributions sum to 1.005 (100.5%) instead of 1.0 (100%)
• R-DHAP: 0.25, R-ESHAP: 0.05, R-GDP: 0.095, R-GemOx: 0.2, R-ICE: 0.4, R-MINE: 0.005
• Fix: Adjust R-GDP to 0.090 to achieve exactly 1.0

🔴 Invalid Clinical Codes

"Check-Up After Second Line Treatment":

• Uses SNOMED code "1234" which is not a valid clinical code
• Display: "Encounter for check up (procedure)"
• Fix: Replace with valid SNOMED code (e.g., 410620009 for "Well child visit")

🔴 Duplicate State Definition

"PET-CT_Scan" appears twice in the module:

• Once with direct_transition to "Delay after Second-Line Treatment"
• Once with direct_transition to "End PET-CT Scan Encounter"
• Fix: Remove duplicate and ensure single, correct definition

Moderate Issues
🟡 State Flow Issues

1. Unreachable States: Several states appear unreachable due to flow logic:

• "Wait for PET-CT Scan"
• "Encounter for PET-CT Scan"
• "End PET-CT Scan Encounter"

2. Missing Encounter Ends: Some encounter states may be missing proper EncounterEnd states

🟡 Clinical Logic Concerns

1. Complex Transition Logic: The "Blood Tests" state has complex conditional logic based on "DLBCL_Survival" attribute that may not be properly set in all cases
2. Inconsistent Categorization: Observation states use different categories:

• "Patient Relapses": category "vital-signs"
• "Second-Line Therapy Shown to be Effective": category "laboratory"
• "Second-Line Treatment Halted Spread of Disease": category "procedure"
• Recommendation: Standardize to "survey" or "procedure" for treatment responses

[awatson1978]

dlbcl_treatment_round_two

✅ Probability Distribution: Fixed - all distributions sum to exactly 1.0
✅ Encounter Classes: Fixed - all consistent as "ambulatory"
✅ SNOMED-CT Code: Fixed - now uses valid "185349003"
✅ Typos: Fixed - "Ineffective" spelling corrected
✅ Spacing: Fixed - proper spacing in state names

[awatson1978]

New bug found: All of the dlbcl files rely on SNOMED code 271299001: "Patient's condition worsened (finding)"

Apostrophe was blowing up the .github/workflows/ci-build-test.yml / Java 11 (pull_request) CI runner, so added a new method to handles apostrophes properly.

private String escapeFieldNameForJsonPath(String fieldName) {
    // If field name contains single quotes, use double quotes
    if (fieldName.contains("'")) {
        return "\"" + fieldName.replace("\"", "\\\"") + "\"";
    } else {
        return "'" + fieldName + "'";
    }
}

[codecov]

Codecov Report

Attention: Patch coverage is 71.50854% with 1018 lines in your changes missing coverage. Please review.

Project coverage is 77%. Comparing base (05d2a08) to head (95f4189).
Report is 1646 commits behind head on master.

Files with missing lines	Patch %	Lines
...a/org/mitre/synthea/export/flexporter/Actions.java	73%	64 Missing and 44 partials ⚠️
src/main/java/RunFlexporter.java	0%	90 Missing ⚠️
...c/main/java/org/mitre/synthea/export/Exporter.java	51%	67 Missing and 21 partials ⚠️
src/main/java/App.java	0%	83 Missing ⚠️
...ain/java/org/mitre/synthea/export/CSVExporter.java	84%	43 Missing and 37 partials ⚠️
src/main/java/org/mitre/synthea/engine/State.java	80%	35 Missing and 36 partials ⚠️
.../mitre/synthea/export/flexporter/FieldWrapper.java	62%	43 Missing and 27 partials ⚠️
...ava/org/mitre/synthea/export/MetadataExporter.java	0%	66 Missing ⚠️
src/main/java/Graphviz.java	0%	54 Missing ⚠️
...a/org/mitre/synthea/export/rif/BB2RIFExporter.java	74%	45 Missing and 8 partials ⚠️
... and 25 more

Additional details and impacted files

@@            Coverage Diff             @@
##             master   #1356     +/-   ##
==========================================
- Coverage        81%     77%     -4%     
- Complexity     2687    4017   +1330     
==========================================
  Files           101     178     +77     
  Lines         16847   25210   +8363     
  Branches       2323    3601   +1278     
==========================================
+ Hits          13684   19543   +5859     
- Misses         2453    4526   +2073     
- Partials        710    1141    +431     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[awatson1978]