Methodology

A rare disease is researched differently from a common one. The trial record is thin, patient communities carry real signal years before the literature does, and the diagnostic criteria themselves have changed three times in forty years. This is the methodology the index runs on — enforced in its schema, not just described in prose.

1. Five source strata

Every source enters under exactly one stratum, and every record attests its evidence per stratum. The strata are never collapsed into a single evidence score.

• Clinical — peer-reviewed articles, trials, guidelines, consensus statements, and mechanistic studies.
• Community — venue-level patterns from forums, support groups, and patient organizations. Indexed at venue level only; no individual poster, handle, or verbatim post enters the corpus.
• Historical — pre-nosology case reports, contemporaneous accounts, archival records, and documented folk practice.
• Registry — patient registries, trial registrations, and rare-disease reference rails (Orphanet, GARD, ClinicalTrials.gov).
• Gray literature — preprints, theses, conference abstracts, and working papers ahead of peer review.

2. Per-stratum evidence tiers

A randomized trial and a recurring forum pattern are both real evidence; they are not the same kind of real. Each stratum defines its own tier ladder — systematic review down to case report in the clinical stratum, structured patient survey down to individual account in the community stratum, primary historical document down to folk tradition in the historical stratum. The schema rejects attestations whose tier does not belong to their stratum.

3. Cross-stratum corroboration

When a record's evidence spans more than one stratum, it must declare a corroboration state: convergent (independent strata agree), contested (strata disagree), or refuted. Single-stratum records cannot claim corroboration. Disagreement between strata is publishable signal — the lidocaine record is the model: decades of community reports, then structured surveys, then a randomized trial.

4. Diagnostic-era tagging

The meaning of “EDS” changed with each nosology — Berlin 1988 expanded it to eleven types, Villefranche 1997 consolidated to six, and the 2017 International Classification defines thirteen and created HSD as the residual category. Every record carries the criteria era its sources worked under, so pre-2017 cohorts are never silently read as modern hEDS.

5. Subtype scoping

Every record declares its subtype scope. hEDS findings are never silently generalized to vEDS or other monogenic types — the conditions share a name and differ in mechanism, risk, and management. Records scoped to all types say so explicitly.

6. Epistemic status

• Established — attested at consensus strength (guideline, consensus statement, review, trial, or cohort) in the clinical stratum, or by registry authority for program records.
• Probable — supported by sub-consensus clinical or registry evidence.
• Emerging — new and not yet replicated; always carries a reassessment date.
• Contested — strata or studies disagree; always carries a reassessment date.
• Community signal — real as community evidence, not yet clinically earned; always carries a reassessment date.
• Historical record — documented history; presence records provenance, never efficacy.
• Refuted — contradicted by stronger evidence; kept visible rather than deleted.

7. Review lifecycle

Every record carries a reviewed_at date. Emerging, contested, and community-signal records are inadmissible without a reassess_by date — undated emerging claims do not enter the corpus. Reassessment cadences are set by the publication policy: emerging 60 days, contested 90 days, community signal 90 days, established 365 days.

8. Publication policy

What may enter the corpus and under what review, verbatim from the published policy file:

• The index describes evidence and provenance; it never recommends, prescribes, or discourages a specific course of care for an individual.
• Community evidence is indexed at venue level. No individual poster, handle, or verbatim post enters the corpus.
• Folk and historical practices are admitted as documented history. Admission records existence and provenance, never efficacy.
• Every record declares its subtype scope. hEDS findings are never silently generalized to vEDS or other monogenic types.
• Records carry the diagnostic-criteria era of their sources so pre-2017 claims are not read as modern nosology.
• Multi-stratum records must declare convergent, contested, or refuted; disagreement between strata is publishable signal, not noise.
• Practices with documented harm potential carry a visible risk note rather than silent omission.
• Research runs, decisions, and publication attestations are append-only; corrections add a new record rather than rewriting history.
• Emerging, contested, and community-signal records must carry a reassessment date; undated emerging claims are inadmissible.

Auto-publishable admissions are bibliographic only (registry trial-record; clinical journal-article). Everything else requires review: community, historical, clinical, gray, registry-stratum material at claim level, and all community and historical material without exception.

9. Discovery monitors and the run ledger

The index is a living program, not a one-time article. Declared monitors watch PubMed, ClinicalTrials.gov, society feeds, preprint servers, guideline bodies, community venues, and historical archives on fixed cadences. Every intake pass is logged in an append-only run ledger — admitted sources, admitted records, and rejected candidates with reasons. Corrections append; nothing rewrites history.

10. What this index does not do

It does not diagnose. It does not recommend or discourage any course of care for any individual. It does not treat forum reports as clinical evidence, and it does not treat documented folk practice as treatment. When evidence is weak or contradictory, the record says so and says why.