paper-writer
VerifiedMedical/scientific paper writing workflow skill. Manages the full pipeline from literature search to submission-ready manuscript. Creates and manages a project directory with IMRAD-format section files, literature matrix, reference management, and quality checklists. Supports both English and Japanese papers. Triggers: 'write paper', 'paper-write', 'start manuscript', '論文を書く', '論文執筆', '論文プロジェクト', 'manuscript', 'research paper', '原稿作成'.
templates/sr-grade.md
View on GitHub# GRADE Evidence Assessment Template
> The Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach
> rates the certainty of evidence for each outcome across studies.
> Reference: Guyatt et al., BMJ 2008;336:924-926.
---
## Overview
GRADE assesses certainty of evidence on a four-level scale:
| Level | Definition | Symbol |
|-------|-----------|--------|
| **High** | Very confident that the true effect lies close to the estimate | ++++ |
| **Moderate** | Moderately confident; the true effect is likely close but may be substantially different | +++O |
| **Low** | Limited confidence; the true effect may be substantially different | ++OO |
| **Very Low** | Very little confidence; the true effect is likely substantially different | +OOO |
**Starting point:**
- Randomized trials start at HIGH
- Observational studies start at LOW
---
## 1. Downgrade Factors
### 1.1 Risk of Bias (Study Limitations)
**Assessment question:** Are there serious limitations in study design or execution that reduce confidence?
| Severity | Action | Criteria |
|----------|--------|----------|
| No serious | No downgrade | Most studies at low RoB; no study at high RoB contributes >50% of the weight |
| Serious | Downgrade 1 level | Several studies with some concerns; or one influential study at high RoB |
| Very serious | Downgrade 2 levels | Most studies at high RoB; or all studies contributing to the estimate have serious limitations |
**Decision rules:**
- Focus on studies that contribute most weight to the pooled estimate
- Consider whether limitations likely bias toward or away from the intervention effect
- Use individual study RoB assessments (from RoB 2, NOS, etc.) as input
**Rating for this review:**
| Outcome | Key Concerns | Severity | Downgrade |
|---------|-------------|----------|-----------|
| {{OUTCOME_1}} | {{CONCERNS}} | Not serious / Serious / Very serious | 0 / -1 / -2 |
| {{OUTCOME_2}} | {{CONCERNS}} | Not serious / Serious / Very serious | 0 / -1 / -2 |
---
### 1.2 Inconsistency
**Assessment question:** Are results consistent across studies?
| Severity | Action | Criteria |
|----------|--------|----------|
| No serious | No downgrade | I2 < 40% and overlapping CIs; consistent direction of effect |
| Serious | Downgrade 1 level | I2 = 40-75%; some variation in effect size but consistent direction |
| Very serious | Downgrade 2 levels | I2 > 75%; inconsistent direction of effects; unexplained heterogeneity |
**Decision rules:**
- Consider I2 statistic, but do not rely on it alone
- Examine overlap of confidence intervals across studies
- Check for consistent direction of effect
- Consider whether subgroup analyses explain heterogeneity
- If heterogeneity is fully explained by pre-specified subgroups, may not downgrade
- With only 2 studies, I2 is unreliable; focus on effect direction and magnitude
**Additional considerations:**
- Point estimates vary widely across studies: suggests inconsistency
- Prediction interval crosses null or clinically important threshold: suggests inconsistency
- I2 can be misleading with few studies or when all studies are large
**Rating for this review:**
| Outcome | I2 | Prediction Interval | Direction Consistent? | Severity | Downgrade |
|---------|-----|--------------------|-----------------------|----------|-----------|
| {{OUTCOME_1}} | {{I2}}% | {{PI}} | Yes / No | Not serious / Serious / Very serious | 0 / -1 / -2 |
| {{OUTCOME_2}} | {{I2}}% | {{PI}} | Yes / No | Not serious / Serious / Very serious | 0 / -1 / -2 |
---
### 1.3 Indirectness
**Assessment question:** Is the evidence directly applicable to the review question?
| Severity | Action | Criteria |
|----------|--------|----------|
| No serious | No downgrade | PICO elements in studies closely match the review question |
| Serious | Downgrade 1 level | One element of PICO substantially differs; or surrogate outcome used |
| Very serious | Downgrade 2 levels | Multiple elements differ; or the evidence addresses a fundamentally different question |
**Types of indirectness:**
| Type | Example |
|------|---------|
| **Population** | Studies in adults when question is about elderly; different severity |
| **Intervention** | Different dose, formulation, or delivery from the intervention of interest |
| **Comparator** | Active comparator when question asks about placebo comparison |
| **Outcome** | Surrogate outcome (e.g., LDL levels instead of cardiovascular events) |
| **Setting** | Hospital-based when question is about primary care |
| **Indirect comparison** | No head-to-head trials; evidence from network of trials |
**Decision rules:**
- Assess each PICO element separately
- Surrogate outcomes always warrant consideration for downgrade
- Head-to-head comparisons are preferable to indirect comparisons
- Consider whether differences are large enough to meaningfully alter the effect
**Rating for this review:**
| Outcome | Indirectness Type | Details | Severity | Downgrade |
|---------|-------------------|---------|----------|-----------|
| {{OUTCOME_1}} | {{TYPE}} | {{DETAILS}} | Not serious / Serious / Very serious | 0 / -1 / -2 |
| {{OUTCOME_2}} | {{TYPE}} | {{DETAILS}} | Not serious / Serious / Very serious | 0 / -1 / -2 |
---
### 1.4 Imprecision
**Assessment question:** Is the effect estimate precise enough to support a decision?
| Severity | Action | Criteria |
|----------|--------|----------|
| No serious | No downgrade | Narrow CI; adequate sample size; CI does not cross decision thresholds |
| Serious | Downgrade 1 level | Wide CI crossing one decision threshold; or OIS not met |
| Very serious | Downgrade 2 levels | Very wide CI crossing both sides of the decision threshold; very small sample |
**Decision rules:**
For **dichotomous outcomes:**
- Calculate the Optimal Information Size (OIS): the sample size a single adequately powered trial would need
- If total N < OIS, consider downgrading for imprecision
- If 95% CI includes both appreciable benefit (RR < 0.75) and appreciable harm (RR > 1.25), downgrade
- If 95% CI crosses the null (RR = 1.0), consider clinical significance of included range
For **continuous outcomes:**
- If 95% CI includes both clinically important and clinically unimportant effects, downgrade
- Define the minimal clinically important difference (MCID) a priori
- If total N < 400 (rule of thumb), consider downgrading
**Thresholds (customize per outcome):**
| Outcome | MCID or Decision Threshold | OIS |
|---------|---------------------------|-----|
| {{OUTCOME_1}} | {{THRESHOLD}} | {{OIS}} |
| {{OUTCOME_2}} | {{THRESHOLD}} | {{OIS}} |
**Rating for this review:**
| Outcome | Total N | 95% CI | Crosses Threshold? | OIS Met? | Severity | Downgrade |
|---------|---------|--------|-------------------|----------|----------|-----------|
| {{OUTCOME_1}} | {{N}} | {{CI}} | Yes / No | Yes / No | Not serious / Serious / Very serious | 0 / -1 / -2 |
| {{OUTCOME_2}} | {{N}} | {{CI}} | Yes / No | Yes / No | Not serious / Serious / Very serious | 0 / -1 / -2 |
---
### 1.5 Publication Bias
**Assessment question:** Is the body of evidence likely affected by selective publication or reporting?
| Severity | Action | Criteria |
|----------|--------|----------|
| Undetected | No downgrade | Comprehensive search; funnel plot symmetric; no evidence of selective reporting |
| Suspected | Downgrade 1 level | Funnel plot asymmetry; Egger test significant; predominantly small positive studies; industry funding concerns |
| Strongly suspected | Downgrade 2 levels | Clear evidence of missing studies; registered protocols without published results |
**Decision rules:**
- Funnel plots require 10 or more studies to be informative
- Egger test p < 0.10 suggests asymmetry
- Check trial registries for registered but unpublished studies
- Consider whether commercial interest could drive selective reporting
- Small study effects (small studies showing larger effects) suggest publication bias
- If unable to assess (too few studies), note this but do not automatically downgrade
**Rating for this review:**
| Outcome | N Studies | Funnel Plot | Egger p | Registry Check | Severity | Downgrade |
|---------|----------|------------|---------|---------------|----------|-----------|
| {{OUTCOME_1}} | {{N}} | Symmetric / Asymmetric / N/A | {{P}} | {{RESULT}} | Undetected / Suspected | 0 / -1 |
| {{OUTCOME_2}} | {{N}} | Symmetric / Asymmetric / N/A | {{P}} | {{RESULT}} | Undetected / Suspected | 0 / -1 |
---
## 2. Upgrade Factors (Observational Studies Only)
These factors may upgrade certainty of evidence from observational studies. Apply only when no downgrade factors are present (or after accounting for downgrades).
### 2.1 Large Magnitude of Effect
| Magnitude | Action | Criteria |
|-----------|--------|----------|
| Large | Upgrade 1 level | RR > 2 or < 0.5 (consistent evidence from 2+ studies, no plausible confounders) |
| Very large | Upgrade 2 levels | RR > 5 or < 0.2 (direct evidence with no major threats to validity) |
**Decision rules:**
- Effect must be consistent across studies
- No plausible confounding that could fully explain the effect
- Based on unadjusted estimates (adjusted estimates may already account for confounding)
### 2.2 Dose-Response Gradient
| Presence | Action | Criteria |
|----------|--------|----------|
| Present | Upgrade 1 level | Clear dose-response relationship observed across or within studies |
**Decision rules:**
- The gradient should be biologically plausible
- Must be evident in the data, not assumed
- Consider whether a threshold effect might exist
### 2.3 Plausible Confounding Would Reduce Effect
| Situation | Action | Criteria |
|-----------|--------|----------|
| All plausible confounders would reduce the observed effect | Upgrade 1 level | If all residual confounding acts in the opposite direction, the true effect may be even larger |
**Decision rules:**
- List all plausible confounders and their expected direction of bias
- If all bias toward the null, this supports upgrading
- Rarely applied; requires strong understanding of confounding structure
**Rating for this review:**
| Outcome | Large Effect? | Dose-Response? | Confounding Direction? | Total Upgrade |
|---------|--------------|---------------|----------------------|---------------|
| {{OUTCOME_1}} | No / +1 / +2 | No / +1 | No / +1 | {{TOTAL}} |
| {{OUTCOME_2}} | No / +1 / +2 | No / +1 | No / +1 | {{TOTAL}} |
---
## 3. GRADE Evidence Profile Table
Complete one row per outcome.
| Outcome | N Studies (Design) | Risk of Bias | Inconsistency | Indirectness | Imprecision | Pub Bias | Upgrade | Certainty |
|---------|-------------------|-------------|---------------|-------------|------------|----------|---------|-----------|
| {{OUTCOME_1}} | {{N}} ({{DESIGN}}) | 0 / -1 / -2 | 0 / -1 / -2 | 0 / -1 / -2 | 0 / -1 / -2 | 0 / -1 | 0/+1/+2 | HIGH/MOD/LOW/VERY LOW |
| {{OUTCOME_2}} | | | | | | | | |
| {{OUTCOME_3}} | | | | | | | | |
---
## 4. Summary of Findings (SoF) Table
This is the primary output for readers. Present in the manuscript or as a supplementary table.
### Summary of Findings: {{INTERVENTION}} compared to {{COMPARATOR}} for {{CONDITION}}
| Outcome | Timeframe | N Studies (Participants) | Certainty (GRADE) | Relative Effect (95% CI) | Risk with Control | Risk Difference |
|---------|-----------|------------------------|-------------------|-------------------------|-------------------|-----------------|
| {{OUTCOME_1}} | {{TIME}} | {{N_STUDIES}} ({{N_PART}}) | {{GRADE_LEVEL}} | {{RR_OR_HR}} ({{CI}}) | {{BASELINE_RISK}} per 1,000 | {{RISK_DIFF}} per 1,000 ({{CI_DIFF}}) |
| {{OUTCOME_2}} | {{TIME}} | {{N_STUDIES}} ({{N_PART}}) | {{GRADE_LEVEL}} | -- | {{CONTROL_MEAN}} | MD {{MEAN_DIFF}} ({{CI_DIFF}}) |
| {{OUTCOME_3}} | {{TIME}} | {{N_STUDIES}} ({{N_PART}}) | {{GRADE_LEVEL}} | -- | -- | {{NARRATIVE_EFFECT}} |
**Footnotes:**
1. {{FOOTNOTE_EXPLAINING_DOWNGRADE_REASON}}
2. {{FOOTNOTE_EXPLAINING_DOWNGRADE_REASON}}
**Explanations:**
- Baseline risk sourced from: {{SOURCE}}
- GRADE Working Group grades: High = very confident; Moderate = moderately confident; Low = limited confidence; Very low = very little confidence.
---
## 5. GRADE Certainty Calculation Worksheet
Use this worksheet for each outcome to systematically calculate the final certainty rating.
### Outcome: {{OUTCOME_NAME}}
| Step | Item | Value |
|------|------|-------|
| **Starting level** | Study design | RCT = HIGH (4) / Observational = LOW (2) |
| **Downgrade 1** | Risk of bias | 0 / -1 / -2 |
| **Downgrade 2** | Inconsistency | 0 / -1 / -2 |
| **Downgrade 3** | Indirectness | 0 / -1 / -2 |
| **Downgrade 4** | Imprecision | 0 / -1 / -2 |
| **Downgrade 5** | Publication bias | 0 / -1 |
| **Upgrade 1** | Large effect | 0 / +1 / +2 |
| **Upgrade 2** | Dose-response | 0 / +1 |
| **Upgrade 3** | Plausible confounding | 0 / +1 |
| **Final score** | Sum | {{SCORE}} |
| **Final certainty** | 4=High, 3=Moderate, 2=Low, 1 or less=Very low | {{LEVEL}} |
**Justification narrative:**
> {{NARRATIVE_EXPLAINING_ALL_RATING_DECISIONS}}
---
## 6. Common Pitfalls
| Pitfall | Guidance |
|---------|----------|
| Double-counting RoB and inconsistency | If all studies are biased in the same direction, this affects RoB, not inconsistency |
| Downgrading for imprecision when effect is significant | Statistical significance alone does not mean the estimate is precise; check CI width against clinical thresholds |
| Not downgrading observational studies for RoB | Even though they start at LOW, further downgrade is appropriate if studies have serious limitations |
| Upgrading observational studies that have been downgraded | Upgrading is generally only applied when there are no serious downgrades |
| Using GRADE for individual studies | GRADE assesses a body of evidence for each outcome, not individual studies |
| Conflating certainty with recommendation strength | GRADE certainty is about evidence; recommendation strength also considers values, preferences, and resources |
---
## 7. Reporting GRADE in the Manuscript
### In Methods
> The certainty of evidence for each outcome was assessed using the GRADE approach. Evidence was classified as high, moderate, low, or very low based on five domains: risk of bias, inconsistency, indirectness, imprecision, and publication bias. For observational studies, upgrading was considered for large magnitude of effect, dose-response gradient, and plausible confounding reducing the effect. GRADE assessments were performed using [GRADEpro GDT / manual assessment] by two reviewers with disagreements resolved by discussion.
### In Results
> The certainty of evidence was [high/moderate/low/very low] for [outcome] (Table X), downgraded [one level/two levels] due to [specific reasons].
### In Discussion
> The [high/moderate/low/very low] certainty evidence for [outcome] means [interpretation from GRADE definitions]. This [supports/limits] confidence in [specific clinical recommendation or conclusion].
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| writing-skillsobra/superpowersUse when creating new skills, editing existing skills, or verifying skills work before deployment | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| xlsxanthropics/skillsUse this skill any time a spreadsheet file is the primary input or output. This means any task where the user wants to: open, read, edit, or fix an existing .xlsx, .xlsm, .xltx, .csv, or .tsv file (e.g., adding columns, computing formulas, formatting, charting, cleaning messy data); create a new spreadsheet from scratch or from other data sources; or convert between tabular file formats. Trigger especially when the user references a spreadsheet file by name or path — even casually (like "the xlsx in my downloads") — and wants something done to it or produced from it. Also trigger for cleaning or restructuring messy tabular data files (malformed rows, misplaced headers, junk data) into proper spreadsheets. The deliverable must be a spreadsheet file. Do NOT trigger when the primary deliverable is a Word document, HTML report, standalone Python script, database pipeline, or Google Sheets API integration, even if tabular data is involved. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| shipping-and-launchaddyosmani/agent-skillsPrepares production launches. Use when preparing to deploy to production, or when asking what needs to be in place before shipping. Use when you need a pre-launch checklist, when setting up monitoring, when planning a staged rollout, or when you need a rollback strategy. | Stars: 119k | Updated | Claude CodeCodex | Verified | Install |
| childhoodthedotmack/claude-memBabysit a shipped release after merge, publication, or deployment. Verify published packages, production behavior, installed plugins, and late CI or user reports until the observation window is clean; use babysit for PR readiness before release. | Stars: 118k | Updated | Claude CodeCodex | Verified | Install |