domain-research-health-science_skill

Overview

This skill structures clinical and health science research using evidence-based medicine frameworks to produce answerable questions, rigorous appraisals, and decision-ready evidence summaries. It guides PICOT question formulation, study design selection, risk-of-bias assessment, outcome prioritization, GRADE certainty rating, and synthesis for guidelines or regulatory dossiers. Use it to translate vague clinical questions into focused, actionable research tasks.

How this skill works

The skill inspects your clinical question and converts it into a PICOT template, recommends appropriate study designs, and provides checklists for systematic appraisal (Cochrane RoB 2, ROBINS-I, QUADAS-2, PROBAST). It guides outcome prioritization (patient-important vs surrogate), performs evidence synthesis steps (search strategy, study selection, meta-analysis criteria), and produces GRADE-based certainty ratings and summary-of-findings formats. Outputs are practical artifacts: PICOT, risk-of-bias notes, evidence profile, and decision-ready interpretation.

When to use it

• Formulating clinical research questions or trial protocols (use PICOT)
• Planning or conducting systematic reviews and meta-analyses
• Critically appraising study quality for guidelines, policy, or clinical decisions
• Prioritizing patient-important outcomes and defining MCIDs
• Assessing regulatory or reimbursement evidence for interventions
• Evaluating diagnostic accuracy, prognosis models, or harm/safety signals

Best practices

• Always frame clinical questions with PICOT before searching or designing studies
• Match study design to question type (RCT for therapy, cohort for prognosis, cross-sectional for diagnosis)
• Use validated bias tools (RoB 2, ROBINS-I, QUADAS-2, PROBAST) rather than informal judgment
• Prioritize patient-important outcomes and report MCIDs, absolute effects, and NNT where relevant
• Apply GRADE to rate certainty and state limitations and applicability explicitly
• Explore heterogeneity before pooling studies; prespecify subgroup analyses and interaction tests

Example use cases

• Convert a vague effectiveness query into a PICOT for an RCT protocol (population, dose, comparator, primary outcome, follow-up)
• Plan a systematic review: define eligibility, search, duplicate screening, extract outcomes, assess RoB, calculate pooled estimates and I²
• Appraise a published RCT: identify randomization issues, blinding, attrition, selective reporting and adjust certainty with GRADE
• Design safety surveillance: choose cohort or case-control approaches for rare harms, define time windows and confounding control
• Create a guideline evidence profile: summary of findings table with effect sizes, certainty, applicability notes