Abstract
An early requirement of the UK's Diabetes National Service Framework is enumeration of the total affected population. Existing estimates tend to be based on incomplete lists. In a study conducted over one year in North Liverpool, we compared crude prevalence rates for type 1 and type 2 diabetes with estimates obtained by capture-recapture (CR) analysis of multiple incomplete patient lists, to assess the extent of unascertained but diagnosed cases. Patient databases were constructed from six sources-a hospital diabetes centre; general practitioner registers; hospital admissions with a diagnosis of diabetes; a hospital diabetic retinal clinic; a research list of patients with diabetes admitted with stroke; and a local children's hospital. Log linear modelling was used to estimate missing cases, hence total prevalence. The crude prevalence of diabetes was 1.5% (95% confidence interval [CI] 1.41, 1.52), compared with a CR-adjusted rate of 3.1% (CI 3.03, 3.19). Age-banded CR-adjusted prevalence was always higher in males than in females and the difference became more pronounced with increasing age. Among males, CR-adjusted prevalence rose from 0.4% at age 10-19 years to 18.3% at 80+ years; in females the corresponding figures were 0.4% and 9.3%. The gap between crude and CR-estimated prevalence points to a rate of 'hidden diabetes' that has substantial implications for future diabetes care.