Introduction: There is no national register of childhood type 1 diabetes mellitus for England. Our aim was to assess the feasibility of using routine hospital admissions data as a surrogate for a childhood diabetes register across England, and to create a geographically referenced childhood diabetes dataset for use in epidemiologic studies and health service research.

Methods: Hospital Episodes Statistics data for England from April 1992 to March 2006 referring to a type 1 diabetes diagnosis in 0–14 yr olds were cleaned to approximate an incident dataset. The cleaned data were validated against regional population-based register data, available for Yorkshire and the area of the former Oxford Regional Health Authority.

Results: There were 32 665 unique cases of type 1 and type unknown diabetes over the study period. The hospital-derived data improved in quality over time (91% concordance with regional register data over the period 2000–2006 vs. 52% concordance over the period 1992–1999), and data quality was better for younger (0–9 yr) (86.5% concordance with regional register data) than older cases (10–14 yr). Overall incidence was 24.99 (95% confidence interval 24.71–25.26) per 100 000. Basic trends in age distribution, seasonality of onset, and incidence matched well with previously reported findings.

Conclusion: We were able to create a surrogate register of childhood diabetes based on national hospital admissions data, containing approximately 2300 cases/yr, and geo-coded to a high resolution. For younger cases (0–9 yr) and more recent years (from 2000) these data will be a useful resource for epidemiological studies exploring the determinants of childhood diabetes.