/clinical-decision-support-documents

name: clinical-decision-support-documents
description: "Guidelines for clinical decision support (CDS) documents: biomarker-stratified cohort analyses and GRADE-graded treatment reports. Covers structure, executive summaries, evidence grading (1A–2C), stats (HR, CI, survival), and biomarker integration. Use for pharma research docs, clinical guidelines, regulatory submissions."
license: CC-BY-4.0

Clinical Decision Support Documents

Overview

Clinical decision support (CDS) documents are analytical reports for pharmaceutical research, guideline development, and regulatory submissions. This knowhow covers two main document types: Patient Cohort Analyses (biomarker-stratified group outcomes) and Treatment Recommendation Reports (evidence-graded clinical guidelines). For individual patient-level treatment plans, use the `treatment-plans` skill instead.

Key Concepts

1. Document Types

**Patient Cohort Analysis** — Group-level statistical comparison of patient subgroups stratified by biomarkers, molecular subtypes, or clinical characteristics.

• Typical content: demographics, biomarker stratification, outcome metrics (OS, PFS, ORR), Kaplan-Meier curves, forest plots
• Audience: pharmaceutical companies, clinical researchers, regulatory bodies
• Length: 5–15 pages (1-page executive summary + detailed sections)

**Treatment Recommendation Report** — Evidence-based clinical guidelines with GRADE-graded recommendations for disease management.

• Typical content: evidence review, recommendations by line of therapy, decision algorithm flowcharts, monitoring protocols
• Audience: guideline committees, medical affairs, KOLs
• Length: 5–20 pages

2. GRADE Evidence Grading System

The Grading of Recommendations, Assessment, Development and Evaluations (GRADE) system classifies recommendations by strength and evidence quality:

| Grade | Strength | Evidence Quality | Meaning | |-------|----------|-----------------|---------| | **1A** | Strong | High | Benefits clearly outweigh risks; consistent RCT data | | **1B** | Strong | Moderate | Benefits likely outweigh risks; limited RCT data | | **2A** | Weak | High | Trade-offs exist; high-quality evidence but patient values matter | | **2B** | Weak | Moderate | Uncertain trade-offs; limited evidence | | **2C** | Weak | Low | Very uncertain; expert opinion or observational data only |

3. Outcome Metrics

| Metric | Abbreviation | Definition | |--------|-------------|------------| | Overall Survival | OS | Time from treatment start to death from any cause | | Progression-Free Survival | PFS | Time to disease progression or death | | Objective Response Rate | ORR | Proportion with CR + PR per RECIST 1.1 | | Duration of Response | DOR | Time from first response to progression | | Disease Control Rate | DCR | Proportion with CR + PR + SD |

4. Statistical Reporting Standards

• **Hazard ratios**: Report with 95% CI (e.g., HR 0.65, 95% CI 0.48–0.89, p=0.007)
• **Survival data**: Median OS/PFS with 95% CI + landmark rates (6-mo, 12-mo, 24-mo)
• **Response rates**: Point estimate with 95% CI
• **Kaplan-Meier curves**: Include number-at-risk tables below, censoring markers, log-rank p-value
• **Subgroup analyses**: Forest plots with interaction p-values; clearly label pre-specified vs exploratory

Decision Framework

Use this framework to select the appropriate document type:

Is this about a POPULATION or an INDIVIDUAL patient?
├── POPULATION (group-level analysis)
│   ├── Comparing outcomes between subgroups? → Patient Cohort Analysis
│   ├── Developing treatment guidelines? → Treatment Recommendation Report
│   └── Both analysis and recommendations? → Combined (cohort analysis + recommendations chapter)
└── INDIVIDUAL (single patient)
    └── Use treatment-plans skill instead

| Scenario | Document Type | Key Sections | |----------|--------------|-------------| | Phase 2/3 trial subgroup analysis | Cohort Analysis | Biomarker stratification, survival curves, forest plots | | Clinical practice guideline | Treatment Recommendations | GRADE-graded recs, decision algorithm, evidence tables | | Companion diagnostic development | Cohort Analysis | Biomarker-response correlation, sensitivity/specificity | | Medical affairs strategy | Treatment Recommendations | Competitive landscape, positioning, KOL education | | Real-world evidence study | Cohort Analysis | EMR cohort definition, outcomes by treatment arm |

Best Practices

1. **Always start with a full-page executive summary**: Page 1 should contain 3–5 colored summary boxes (findings, biomarkers, implications, statistics, safety) that are scannable in 60 seconds. No table of contents on page 1. This is the single most impactful formatting decision for CDS documents.

2. **Use GRADE consistently**: Every treatment recommendation must have a GRADE rating (1A–2C) with documented rationale. Do not mix GRADE with other rating systems within the same document.

3. **Report effect sizes, not just p-values**: Always include hazard ratios or odds ratios with 95% confidence intervals. A p-value alone does not convey clinical significance or effect magnitude.

4. **Specify biomarker assay details**: Name the platform (e.g., FoundationOne CDx, Ventana PD-L1 SP263), cut-points, and validation status. Biomarker results are only actionable when the assay is known.

5. **Use RECIST 1.1 for response assessment**: For immunotherapy cohorts, note iRECIST criteria and pseudoprogression handling. Clearly state which criteria were used.

6. **Include number-at-risk tables**: Below every Kaplan-Meier curve, show the number of patients at risk at each time point. This is mandatory for credible survival analysis.

7. **Declare data completeness and follow-up**: Report median follow-up time, data maturity (% events), and how missing data was handled (complete case, imputation method).

8. **De-identify per HIPAA Safe Harbor**: Remove all 18 HIPAA identifiers before including any patient-level da