---
name: evidence-strength-assessment
description: Use when weighing whether a collection of studies supports a specific research claim or decision.
---

# Assess the strength of evidence

Define the exact claim or decision, population or system, outcome, comparison and time horizon. Inventory the studies or source material actually available and distinguish a user-supplied set from a completed search. A confident verdict requires evidence relevant to the stated question; do not infer consensus from a count of abstracts or from inaccessible full texts.

For each relevant source, record its design, setting, methods, estimates or qualitative findings, uncertainty, limitations and directness to the question. Consider selection, measurement, confounding, missing data, modeling assumptions and conflicts of interest where applicable. Determine whether apparently separate reports share participants or datasets so dependent evidence is not counted twice. For theoretical or numerical research, inspect assumptions, derivations, boundary cases and validation rather than forcing an intervention-study hierarchy onto the question.

Compare the evidence across studies. Explain the direction and magnitude of findings, consistency, precision, plausible alternative explanations and any important contrary evidence. Distinguish absence of evidence from evidence of absence and statistical uncertainty from systematic bias. If there are too few, conflicting, indirect or inaccessible studies, state the resulting uncertainty rather than inventing a numeric score.

Deliver a claim-to-evidence table, a cautious conclusion, material limitations and the next verification step. Cite only sources actually inspected, distinguishing full text from metadata or abstracts. If a field-specific grading framework is requested, verify its current rules and identify the actual judgments made; do not label an improvised verdict as a formal assessment.