What research methods produce reliable population estimates of differences in sex development (DSDs) and what are their limitations?
Executive summary
Reliable population estimates of differences (or variations) in sex development (DSD/intersex) depend on three complementary methods: clinical registry and administrative-data ascertainment, population surveys with careful question design, and targeted field or capture–recapture approaches that adjust for detection bias; each yields useful but different pictures because DSDs are rare, variably defined, and often under- or mis‑recorded [1] [2] [3]. Major limitations include changing diagnostic nomenclature and care pathways, low statistical power to detect rare conditions, sampling and survivorship biases in clinical data, and misclassification or overrepresentation in self‑report surveys unless questions and follow‑up are validated [4] [1] [2] [3].
1. Clinical registries and administrative-case ascertainment: depth with blind spots
Population-based registries and clinical-case ascertainment—like the Swiss retrospective I‑DSD registry analysis—produce detailed, clinically verified prevalence estimates and allow classification by karyotype and diagnosis, yielding per‑100,000 birth rates that are directly actionable for health services planning [1]. These data are limited by who reaches care (referral, regional specialty centers), changes in diagnostic criteria over time, variable participation of private practices and hospitals, and the impossibility of capturing undiagnosed or non‑medicalized individuals; registries therefore provide high specificity but uncertain completeness [1] [2].
2. Population surveys and self‑report methods: scale and inclusivity versus validity concerns
Large population surveys asking direct items—e.g., “Were you born intersex, or with a variation of sex characteristics or sex development?”—can capture people who never entered clinical systems and can produce surprisingly large estimates (up to ~1.7% in one community‑sample study), but they trade clinical verification for reach and therefore risk false positives, overrepresentation of LGBT samples, and ambiguity when respondents lack specific medical diagnoses [2]. Improving survey validity requires iterative item development, community engagement, and mixed‑mode follow‑up to verify and disambiguate self‑reports, a point stressed in systematic reviews of sex, sexual orientation, and gender identity measurement [5] [6].
3. Capture‑recapture, mark–recapture and detection‑adjusted methods: correcting for hidden cases
Methods borrowed from ecology—mark–recapture, N‑mixture, and other detection‑based models—offer tools to estimate unobserved cases by modeling detectability differences between groups and repeated sampling occasions; in wildlife research these reduce bias in adult sex ratio estimates and, analogously, could help account for differential clinical visibility of DSD presentations [7] [8]. Practical limitations include the need for repeated, independent ascertainment events, assumptions about closed populations or known migration, and complexity of implementation in human health data where ethical constraints and mobility make the assumptions harder to satisfy [8] [7].
4. Targeted field methods and hard‑to‑reach sampling: street intercepts, venue‑based and respondent‑driven designs
For marginalized or community‑defined groups, targeted approaches—street‑intercept, venue‑based sampling, and respondent‑driven sampling—have been used successfully to size hidden populations (e.g., MSM studies) and might identify individuals not present in clinical registries [9]. These yield insight into behavior and identity but are vulnerable to selection biases, variable network structures, and ethical risks of inadvertent disclosure; they must be paired with ethical safeguards and validation strategies to avoid inflating prevalence through convenience sampling [9] [2].
5. Analytical design, conceptual clarity, and statistical power: the final mile
Designing studies that validly estimate DSD prevalence requires precise operational definitions separating biological sex characteristics from gender identity, pre‑specifying analytic strategies (sex‑stratified analyses, interaction tests, causal frameworks) and acknowledging the power limits when conditions are rare—detecting interactions or subgroup differences demands much larger samples than detecting main effects [3] [4] [10]. No single method suffices: the most reliable picture will triangulate registry verification, validated survey screening, and detection‑adjusted statistical models while being transparent about coverage gaps and the evolving taxonomy of DSD diagnoses [1] [2] [3].