Methodology

The corpus

25,037 papers sourced from OpenAlex and PubMed with citation chaining; 18,082 triaged for relevance; 13,222 judged relevant; 9,257 fully extracted into structured records (findings, treatments, methodology, safety, populations). Community signal comes from 1,139,360 posts across eczema-focused communities, triaged into 761,985 intervention mentions and 133,215 trigger reports. Paper snapshot: February 2026, the date is shown on every dossier.

Two axes, never one score

Every treatment shows an evidence grade (computed from papers only) and a community interest tier (computed from posts only), separately. Blending them into one number buries the most useful information: a treatment with 18,000 mentions and 19 papers is not the same kind of knowledge as one with 5,000 mentions and 28 trials. The gap between the axes is presented as a finding.

Evidence grading

Grades derive from study design (systematic reviews and meta-analyses, then RCTs, then observational designs), cross-checked against the evidence levels recorded at extraction. Where the stated level and the study design disagree, the study design wins. The rubric is sanity-gated: it must grade topical corticosteroids and dupilumab as strong evidence and must not grade elimination diets as strong, or nothing publishes until it is fixed.

Verification gates

Every citation on every dossier is machine-resolved against OpenAlex and Crossref at build time; anything that fails resolution is removed and logged. Every number is computed from the database by code, never written by a language model. Models phrase sentences from structured rows and must reference the rows behind every sentence; drafts that assert anything unsupported are rejected mechanically.

Honest limitations

• Patient-level outcomes are nearly absent in the literature we extracted; dossiers describe population evidence, not predictions for your child.
• Community themes are categorized at scale and not clinically validated; they are presented as experience volume, never as efficacy.
• Demographic slicing (age, severity, allergy profile) is limited by what papers report; where pediatric data is missing, the dossier says so explicitly.
• The paper corpus is a snapshot; treatments approved or studied after February 2026 may be underrepresented.

What this is not

Not medical advice, not a substitute for a dermatologist or pediatrician, and not a store. There are no affiliate links and nothing for sale. It is a research map built by a parent who needed it, published so other parents can check the sources themselves.