研究假說驗證與可行性評估技能

純指令技能。LLM 醫學知識 + compare_with_literature, list_saved_references, search_local_references, validate_concept。

觸發：hypothesis、假說、PICO、feasibility、可行性、testable、研究計畫評估

---

Framework 1: Hypothesis Validation

Step 1: PICO 分析

Component	偵測重點
Population	patients, subjects, ICU, surgical...
Intervention	treatment, therapy, drug, procedure...
Comparison	versus, vs, placebo, standard, control...
Outcome	mortality, survival, recovery, score, rate...

Step 2: Testability

1. Directional — higher/lower/better/worse/reduce? ✅ vs 只有描述性 ❌
2. Comparison — significantly/compared/than? ✅ vs 缺基準 ❌
3. Type — Null (no difference) vs Alternative (有方向)
4. Specificity — <10 words 太模糊, 10-50 適當, >50 過複雜

Step 3: Study Type Recommendation

假說特徵	建議設計
effect/efficacy/treatment	RCT or Cohort
risk factor/association	Cohort or Case-Control
prevalence/frequency	Cross-Sectional

---

Framework 2: Feasibility Assessment

Sample Size Quick Check

類型	注意
N ≤ 10	Case report/series
N < 30 RCT	Pilot or crossover
N ≥ 100	大多設計合理

Timeline by Study Type

類型	IRB	主要工作	分析寫作
RCT	2 mo	招募+收案 ÷3	÷4
Retrospective	1 mo	資料萃取 2 mo	÷2
SR	1 mo (protocol)	搜尋+篩選 ÷3	÷3

Feasibility Checklist (17 items, 5 categories)

Category	Critical Items	High Items
Data & Sample	資料可及、樣本可達	納排標準
Ethics	IRB plan、informed consent	privacy
Resources	—	funding, timeline, team, equipment
Methodology	primary outcome	stat plan, validated tools, confounders
Publication	—	target journal, reporting guideline, novelty

All Critical items must be ✅ before proceeding.

---

使用原則

1. PICO 分析用語意理解，非關鍵字匹配
2. 可行性評估寧可保守
3. 文獻比較用 compare_with_literature MCP tool