| name | description |
|---|---|
Medical Paper Analysis |
Comprehensive framework for analyzing medical research papers with critical appraisal, statistical interpretation, and clinical application assessment |
This skill provides a structured approach to analyzing medical research papers, from quick summaries to in-depth critical appraisal. It helps extract key findings, evaluate study quality, and assess clinical applicability.
Use this skill when you need to:
- Quickly summarize medical research papers
- Perform critical appraisal of study methodology
- Interpret complex statistical analyses
- Compare multiple studies
- Evaluate clinical applicability of research findings
- Analyze systematic reviews and meta-analyses
For rapid assessment of any medical paper, provide:
Study Purpose: What clinical question does this research address? (1 sentence)
Study Design: Research type and sample size (1-2 sentences)
Key Findings: Main results with specific data (p-values, 95% CI, effect sizes)
Clinical Significance: Impact on clinical practice and current treatment guidelines
Evidence Level: Strength of evidence (High/Moderate/Low)
Output should be:
- Concise (under 300 words)
- Bullet-point format
- Each point maximum 2 sentences
- Include specific numerical data
For comprehensive analysis, evaluate:
Study Design
- Type: RCT / Cohort / Case-Control / Cross-sectional
- Appropriateness for research question
- Evidence pyramid level
Methodological Quality (Use CONSORT checklist for RCTs)
- Randomization method and execution
- Allocation concealment adequacy
- Blinding (single/double/triple)
- Sample size calculation and statistical power
Subject Characteristics
- Inclusion/exclusion criteria appropriateness
- Baseline characteristics balance
- Loss to follow-up rate
- ITT vs PP analysis
Method Selection
- Appropriateness of statistical tests
- Multiple comparison handling
- Missing data management
Results Reporting
- Primary endpoint statistical significance
- Clinical significance of effect size
- Confidence interval interpretation
- Secondary endpoint interpretation
Assess risk in five domains:
- Randomization process
- Deviations from intended interventions
- Missing outcome data
- Outcome measurement
- Selective reporting
Label each: Low risk / Some concerns / High risk
Evaluate and grade from:
- Study design (starting score)
- Risk of bias (downgrade factor)
- Inconsistency (downgrade factor)
- Indirectness (downgrade factor)
- Imprecision (downgrade factor)
- Publication bias (downgrade factor)
Final grade: High / Moderate / Low / Very Low
- Main strengths (3 points)
- Main limitations (3 points)
- External validity: Generalizability to your patients
- Conflicts of interest or sponsorship influence
- Overall recommendation: Should clinical practice change?
For complex statistical sections:
Method Identification
- List all statistical methods used
- Provide Chinese and English full names
Method Selection Rationale
- Why was this method chosen?
- What data types does it suit? (continuous/categorical/time-to-event)
- How did research question influence method selection?
Statistical Assumptions
- Key assumptions for each method
- Were assumptions tested?
- Consequences of assumption violations
Results Interpretation Guide
- p-value meaning: Statistical vs clinical significance
- 95% CI interpretation and relationship to minimal clinically important difference
- Effect size clinical meaning (Cohen's d / OR / HR)
- Interaction term interpretation if present
Alternative Methods Discussion
- Other applicable statistical approaches
- Pros and cons of alternatives
- Why author's choice may (or may not) be optimal
Common Pitfalls
- Typical interpretation errors clinicians make
- Critical details to watch for
For any figure/table in the paper:
Basic Information
- Figure/table type identification
- Why this visualization was chosen
- What information type it best conveys
Axis and Legend Interpretation
- X-axis meaning and units
- Y-axis meaning and units
- Legend interpretation
- Color/symbol/line meanings
- Error bar meaning (SD/SE/95% CI)
Data Interpretation
- Observable trends
- Between-group differences and statistical significance
- p-value markings and their meaning
- Effect size magnitude and clinical importance
Special Figure Types
For Forest Plots:
- Diamond meaning
- Horizontal line meaning
- Vertical dashed line (typically at 1.0) meaning
- Study weight determination
- Pooled effect size
- Heterogeneity index (I²) and interpretation
For Kaplan-Meier Curves:
- Survival curve trajectory
- When curves separate
- Log-rank test p-value
- Median survival times
- Survival rates at specific timepoints (e.g., 5-year)
- Censoring marks and frequency
For ROC Curves:
- AUC (area under curve) value
- AUC interpretation for diagnostic value
- Optimal cut-off point location
- Sensitivity and specificity at that cut-off
- Distance from diagonal line meaning
Potential Issues
- Misleading presentation
- Y-axis manipulation exaggerating differences
- Selective data presentation concerns
- Missing important information (sample sizes, confidence intervals)
- Statistical method appropriateness
Evaluate clinical practice impact:
A) Best Evidence
- Evidence level: 1a/1b/2a/2b/3a/3b/4/5
- Study quality: High/Moderate/Low
- Recommendation strength: Strong/Weak/Against
B) Clinical Expertise
- Intervention usage in your specialty:
- Routine use
- Occasional use
- Rare use
- Never used
- Required special skills or equipment
- Learning curve considerations
C) Patient Values
- Patient benefits
- Patient risks
- Potential patient concerns
- How to discuss this treatment option with patients
Population Compare study population vs your patients:
- Age
- Gender
- Disease severity
- Comorbidities
- Race/geographic region
Conclusion: Similarity (High/Moderate/Low)
Intervention
- Feasibility in your setting
- Drug/device regulatory approval and insurance coverage
- Dose/frequency appropriateness for your patients
- Needed adjustments
Comparison
- Is study control group current standard of care?
- What is current standard of care in your practice?
- How do differences affect result interpretation?
Outcome
- Are measured outcomes important to your patients?
- Missing important outcome measures?
- Adequate follow-up duration?
If paper provides sufficient information:
- NNT = [number]
- Interpretation: Need to treat X patients for 1 to benefit
- Number Needed to Harm (NNH) = [number] (if adverse event data available)
- Benefit-harm ratio: NNH/NNT
Scenario 1: Typical Patient Patient meeting study inclusion criteria:
- Should new treatment be used?
- Rationale:
Scenario 2: Marginal Patient Patient partially meeting criteria (older, comorbidities):
- Should new treatment be used?
- Additional considerations?
- Rationale:
Scenario 3: Patient Not Meeting Inclusion Criteria
- Can results be extrapolated?
- What requires special monitoring?
- Risk assessment:
- Current guideline recommendations
- Consistency between study findings and guidelines
- Is this study sufficient to change guidelines?
- When to wait for more evidence vs early adoption?
If adopting this new treatment:
- Required resources (personnel/equipment/budget)
- Needed training
- How to monitor efficacy and safety
- Documentation and tracking
- When to reassess
Provide:
- Plain language explanation of new treatment
- How to explain potential benefits (use absolute risk, not relative risk)
- How to explain potential risks
- How to address common patient questions
- Shared decision-making dialogue examples
For systematic reviews, use PRISMA checklist:
Search Strategy
- Databases searched - comprehensive?
- Search strategy appropriateness (sensitivity and specificity)
- Search timeframe
- Grey literature searched?
- Hand-searching of references?
- Language restrictions and potential bias
Study Selection
- Clear inclusion/exclusion criteria
- PICO framework present
- Two independent reviewers for screening
- Disagreement resolution method
- PRISMA flow diagram analysis:
- Initial search results
- After duplicate removal
- After title/abstract screening
- After full-text assessment
- Finally included
- Exclusion reasons at each stage
Data Extraction
- Extracted data elements
- Standardized form used
- Two independent extractors
- Missing data handling
- Contact with original authors for data
Quality Assessment
- Tool used for bias risk assessment:
- RCTs: Cochrane RoB 2.0
- Observational studies: Newcastle-Ottawa Scale
- Other tools
- Assessment results and bias risk per study
- Sensitivity analysis excluding low-quality studies
Included Studies Characteristics Create table of study characteristics: | Study | Year | Design | Sample Size | Intervention | Control | Follow-up | Bias Risk |
Heterogeneity Assessment
- I² statistic value:
- I² <25%: Low heterogeneity
- I² 25-50%: Moderate heterogeneity
- I² >50%: High heterogeneity
- Cochran's Q test p-value
- Possible sources of heterogeneity (explore via subgroup analysis or meta-regression)
Pooled Effect
- Fixed-effect or random-effects model? Why?
- Pooled effect size:
- Odds Ratio / Risk Ratio / Hazard Ratio
- 95% CI
- p-value
- Forest plot interpretation:
- Overall effect direction
- Consistency across studies
- Does diamond cross line of no effect?
Publication Bias
- Publication bias assessed?
- Methods used:
- Funnel plot (visual)
- Egger's test
- Begg's test
- Results - evidence of publication bias?
- If yes, trim-and-fill analysis performed?
Assess certainty, starting from high quality, consider downgrading factors:
| GRADE Factor | Assessment | Downgrade? | Reason |
|---|---|---|---|
| 1. Study design | |||
| 2. Risk of bias | |||
| 3. Inconsistency | |||
| 4. Indirectness | |||
| 5. Imprecision | |||
| 6. Publication bias |
Final evidence level: ⊕⊕⊕⊕ High / ⊕⊕⊕◯ Moderate / ⊕⊕◯◯ Low / ⊕◯◯◯ Very Low
For meta-analyses:
- Fixed-effect vs random-effects model
- Selection rationale (based on I² or predetermined)
- If random-effects, which method:
- DerSimonian-Laird
- Restricted maximum likelihood (REML)
- Hartung-Knapp adjustment
Statistical Heterogeneity
- I² = [value] %
- Cochran's Q test:
- Q statistic
- p-value
- df
- τ² (tau-squared, heterogeneity variance)
Heterogeneity Source Investigation
-
Subgroup analyses
- Defined subgroups
- Test for subgroup differences results
-
Meta-regression
- Covariates used
- Can results explain heterogeneity?
- R² (proportion of heterogeneity explained)
-
Sensitivity analysis
- Leave-one-out analysis
- Does result change after excluding specific studies?
- Which are influential studies?
Visual Inspection
- Point estimate distribution across studies
- Confidence interval overlap
- Obvious outliers
- Weight distribution (square sizes)
Pooled Effect
- Pooled estimate = [value]
- 95% CI = [lower, upper]
- p-value
- Diamond position relative to line of no effect (RR/OR=1 or MD=0)
Clinical Significance
- Effect size clinical importance
- Reaches minimal clinically important difference (MCID)?
- Does CI include clinically important difference threshold?
- Absolute risk reduction (ARR)
- Number Needed to Treat (NNT)
Funnel Plot
- Plot symmetry
- Small-study effect present?
- Which part is missing (potential unpublished studies)
Statistical Tests
- Egger's test: p = ?
- Begg's test: p = ?
- Conclusion: [Evidence/No evidence] of publication bias
Adjustment Analysis If publication bias exists:
- Trim and fill method
- How many studies added?
- Adjusted pooled estimate = ?
- Other methods (selection models)
Cumulative Meta-Analysis
- Performed?
- When did evidence reach statistical significance?
- Do recent studies change conclusions?
Trial Sequential Analysis (TSA)
- Performed?
- Required information size (RIS) = ?
- Sufficient information reached?
- Does Z-curve cross monitoring boundary?
- Conclusion: More studies needed?
All analyses should:
- Use clear, structured formatting
- Include tables for comparative data
- Highlight key statistical values
- Provide both statistical and clinical interpretations
- Use bullet points and numbered lists for readability
- Include visual descriptions when discussing figures
- Mark statistical significance: * for p<0.05, ** for p<0.01
- Use symbols for results: ↑ positive, ↓ negative, → no difference
Adjust complexity based on user background:
- Medical students: Focus on basic concepts, use more analogies
- Residents: Balance between learning and application
- Attending physicians: Emphasize clinical application and practice changes
- Researchers: Include methodological details and statistical nuances
Default to English for medical papers, but can provide:
- Chinese (Traditional) for Taiwanese medical terminology
- Bilingual summaries when requested
- Culturally appropriate clinical examples
- Evidence-based: All assessments grounded in established frameworks (CONSORT, PRISMA, GRADE, etc.)
- Clinical focus: Always connect findings to clinical practice
- Critical thinking: Identify both strengths and limitations
- Clear communication: Translate complex statistics into clinically meaningful language
- Actionable: Provide specific recommendations for practice changes
- Comprehensive yet concise: Thorough analysis without unnecessary verbosity
When analyzing a paper, specify:
- Analysis type needed (quick summary, full critical appraisal, statistical focus, etc.)
- Your background/specialty
- Specific questions or concerns
- Target patient population for applicability assessment
The skill will then provide structured analysis using the appropriate framework from above.