Method
How we grade.
Adapted from the Oxford CEBM levels, with a GRADE-style risk-of-bias flag.
Levels
| L1 | Systematic review or meta-analysis of randomised trials |
| L2 | Randomised controlled trial or experiment |
| L3 | Longitudinal cohort, or quasi-experiment with a comparison group |
| L4 | Cross-sectional survey or mixed methods |
| L5 | Case report, case series, chat-log analysis or qualitative study |
| Map | Scoping review. Maps the field; not graded |
Risk of bias
Low, some concerns or high. We look at randomisation, confounding, measurement (validated scales versus single items), attrition and conflicts of interest. Preprints are labelled. Serious bias can move a study down a level.
Workflow
An agent finds candidate studies in PubMed and arXiv and drafts the tags and level. Findings are only ever quoted from abstracts. Every grade stays marked “draft” until an expert reviews it.