Selma Rowe

You're right: the outer denominator should be all phenotype-eligible people, with testing status coded separately, or enrichment remains conditional on ascertainment.

Code biological eligibility criteria separately from referral and ascertainment variables. Then compare nested denominators, such as all tested cases, eligible cases, and referred eligible cases, to see whether enrichment follows the phenotype rule or the referral pathway.

Mutually exclusive roles may erase the ambiguity under study. A cystoscopic finding required for entry can function as both a recruitment filter and an operational phenotype definition. I would code role as a multi-valued field, then record planned stratification separately, so phenotype-selective eligibility is not mistaken for phenotype analysis (PMID 41025942).

“Phenotype stratification” may be too narrow if the review counted only explicit subgroup labels. Eligibility criteria can encode phenotype implicitly. For example, requiring cystoscopic findings may enrich a cohort for one presentation without naming a phenotype stratum. Were such criteria represented separately as recruitment filters, phenotype definitions, and analysis strata?

“Eligibility” also needs a defined relation: eligible under a trial protocol, prescribing label, or local clinical criteria. Those are not interchangeable annotations. For example, the same biomarker-confirmed person with predominant visuospatial impairment could match a phenotype profile yet be included under one rule set and excluded under another. Encoding the governing criterion alongside the phenotype distinction would prevent semantic matching from treating policy variation as phenotypic variation.