Ethnicity-resolved rare disease prevalence exposes hidden diagnostic gaps in England
A national electronic health record analysis of 406 rare diseases has mapped a level of ethnic variation that broad census categories had been masking for years. Among 62.5 million people alive in England on 31 July 2023, 1.05 million had at least one of the rare diseases captured in the study, representing 1.7% of the population. The scale alone matters: rare diseases are individually uncommon, but in aggregate they occupy a large and clinically consequential slice of the health system, particularly when early recognition depends on pattern matching across specialties, age groups, and ancestry-linked phenotypes.
The most striking message is not just prevalence, but heterogeneity. Age- and sex-standardised rates differed widely across ethnic groups, and the differences became much more visible when broad Office for National Statistics categories were split into the more granular National Health Service classifications. In other words, the usual “broad Black,” “broad Asian,” or “broad White” bins were too coarse to capture the real distribution of several conditions. For clinicians, that means apparent group averages can conceal disease burdens that are concentrated in specific communities, delaying suspicion, referral, and diagnosis.
Granular ethnicity categories uncovered disease patterns the broad labels obscured
One example stood out clearly: discoid lupus erythematosus was markedly more prevalent in Black Caribbean individuals, but not in Black African individuals, a distinction that disappeared inside the aggregate “Black” category. This is more than a statistical curiosity. Skin disease phenotypes often vary with ancestry, pigmentation, environmental exposures, and access to specialist assessment, so collapsing different populations into a single label can flatten clinically important biology into an unusable average.
The same logic applies across the broader rare disease portfolio. When prevalence is measured only at a national level, the result can suggest a stable background burden. When the same data are disaggregated by ethnicity, the burden becomes uneven and geographically invisible at the bedside. A clinician seeing one patient with a subtle multisystem presentation may not have the population-level context to recognize that the patient’s ethnic background is associated with a higher or lower likelihood of a particular rare diagnosis. Population mapping cannot replace clinical judgment, but it can make judgment more informed.
Recorded COVID-19 infection was higher in people with rare disease
Recorded COVID-19 infection was higher in the rare disease group than in the non-rare-disease population: 32.8% versus 29.0%. The difference is modest in absolute terms, yet it is meaningful in a cohort of this size and in the context of chronic multisystem illness. People with rare diseases often have organ-specific vulnerability, immune dysregulation, disability, frequent health-care contact, or treatment-related susceptibility, any of which can alter exposure risk or clinical detection of infection. The study did not attempt to parse those mechanisms, but the observation fits the broader expectation that complex chronic illness can increase the chance of a recorded viral diagnosis.
Within the rare disease group, infection rates also varied by ethnicity. Recorded infection was 27.7% in Pakistani individuals and 23.8% in Black African individuals, compared with 33.7% in White British individuals. Across the 25 most common rare diseases, infection rates ranged from 24.6% to 41.2%. These spreads suggest that both ancestry-linked social determinants and disease-specific vulnerabilities likely shaped who was recorded as infected during the pandemic. The study design cannot separate biologic susceptibility from testing access, healthcare-seeking behavior, household exposure, occupation, shielding practices, or differential recording in the health record. But it does show that the burden was not uniform.
One practical consequence follows immediately: a single rare disease registry or clinic pathway cannot assume a uniform COVID-19 exposure profile across all ethnic groups. The people living with rare conditions were not affected as a single block, and the infection data make that plain.
What the England-wide data can and cannot support
This was a cross-sectional study conducted within the NHS England Secure Data Environment, using linked primary care, hospital, and mortality records. That architecture is a strength because it allows population-scale estimates rather than clinic-based sampling, which often overrepresents more severe or better-ascertained disease. It also means the prevalence figures are grounded in routinely captured care data from a very large denominator, which is essential when evaluating uncommon conditions.
At the same time, routine data bring familiar constraints. The study measured recorded COVID-19 infection, not necessarily total infection burden, so asymptomatic disease or infections never tested or documented would be missed. Rare disease identification also depends on coding quality, diagnostic completion, and how completely a condition is captured in linked records. Any one of those factors can shift prevalence estimates, particularly for rare disorders that may take years to diagnose or may be recognized only in specialist care.
Even with those limitations, the main pattern is consistent and clinically important: prevalence varied substantially across ethnic groups, and those differences were often hidden until the data were split into finer categories. The highest recorded COVID-19 infection rates appeared in the ONS White and NHS White British groups, while the rare-disease burden itself showed substantial heterogeneity across the smaller ethnic subgroups. This dual signal matters because it challenges a common assumption that rare disease epidemiology is too sparse to reveal meaningful population structure. In England, with enough linked data, it clearly can.
Why these prevalence maps matter at the bedside
Rare disease care depends on early recognition, but recognition usually begins with a clinician asking whether a symptom cluster fits a known pattern. Ethnicity-resolved prevalence maps can sharpen that question. If a condition is concentrated in one subgroup and nearly absent in another, the threshold for investigation changes. If a disease appears elevated only in a granular subgroup and not in a broad umbrella category, the umbrella category is clinically insufficient.
The study provides a baseline for earlier and fairer diagnosis, but it also points to a more immediate operational need: diagnostic services, outreach programs, and specialist referral pathways should not rely on coarse ethnic labels when planning case finding. The data suggest that better stratification could help identify who is being missed and where. In a field where many patients still endure prolonged diagnostic odysseys, the difference between a broad category and a more precise one can determine whether a clinician recognizes the disease at all.