An adult rare diseases surveillance registry was established in Wales in June 2020 as part of the COVID-19 pandemic response. The initial focus was to collect data on the conditions included on the clinically extremely vulnerable (shielding) list, including Behçet’s disease.
Identification of Behçet’s disease cases for registration was a dual process. Firstly, anonymised data were transferred by registry staff from Public Health Wales’s Congenital Anomalies Register and Information Service (CARIS) for individuals now over 18 years of age. Secondly, data were accessed from the Patient Case Episodes Database Wales (PEDW). This secondary care database was implemented in 1991 and holds data on all emergency and planned hospital in-patient admissions in Wales. Demographic data were cross-referenced with the Welsh Demographic Service (WDS, a used by NHS services to identify cases and their addresses) to reconcile case data and address missing variables. As WDS records deaths, checks could be made on whether cases were still alive. As many rare disease patients have multiple contacts with the NHS, it was possible to check consistency of information across multiple admissions. Clinical portals were accessed, containing information including clinical letters, discharge summaries, referral letters and microbiological and radiological test results. This was to ensure that coding in PEDW was correct and to validate diagnoses, with escalation to a clinician if required, for example by manually checking for the most recent diagnosis where individuals had received previous misdiagnoses. Initial misdiagnosis or consideration of differential diagnoses is a common occurrence for patients with rare diseases as diagnosis can be challenging. To calculate population-level prevalence, paediatric cases of Behçet’s disease were also identified from the CARIS database.
The first step in handling data was to manually de-duplicate the records, acknowledging the potential for cases to be recorded more than once due to data being gathered from multiple datasets. Cases were then removed where diagnosis of Behçet’s disease was suspected, but not verified in the clinical records.
In order to adjust for deprivation, a new variable, the Welsh Index of Multiple Deprivation (WIMD) quintile of residence for each individual, was created. WIMD is the Welsh Government’s official measure of relative deprivation for small areas in Wales . Using a postcode to WIMD rank look-up tool , a deprivation quintile between 1 and 5, with 1 being the most deprived, was inputted for each record.
To facilitate survival analysis, another new variable, survival in months from diagnosis, was created. Date of death was used as the censor point for those cases known to have died. Living cases had survival measured in months from date of diagnosis (the earliest mention of a coded episode) to the end of the study period.
Population prevalence was calculated by combining paediatric cases of Behçet’s disease recorded by CARIS, and living adult cases from the adult rare disease registry as the numerator. The denominator was the population of Wales as per the most recent Office for National Statistics mid-year population estimate at the study end date . For the purposes of the epidemiological description, only adult cases were included.
The data were exported to SPSS (version 24.0, IBM, Chicago, IL)  for statistical analyses. Continuous data were summarised as median (range) and non-parametric tests used. Categorical and continuous data were compared with the Chi-square and Mann–Whitney U-tests respectively. Spearman’s Rho was used to investigate the relationship between date of diagnosis and age at diagnosis, to ascertain whether age at diagnosis had varied over time. The significance value was set at 5% (p < 0.05).
Univariable and multivariable survival analyses were conducted using Kaplan–Meier and Cox Regression (backward likelihood ratio) respectively. As the proportion of deaths across the cohort was relatively low (6.6%), formal median survival according to survival analyses were not possible, however survival was calculated manually using survival from date in diagnosis in months. The variables entered into the Cox regression were gender, WIMD quintile of residence and age at diagnosis.