Bayesian analysis of the geographical variation of type 1 diabetes mellitus in under 15 yr olds in northeast Spain, 1991–2009


Corresponding author:

Maria Pilar Rodrigo,

Department of Public Health,

Government of Aragon,

Via Universitas, 37, Zaragoza, Spain.

Tel: +34976715915;




High variability has been observed in type 1 diabetes mellitus (DM1) incidence rates (IRs) in childhood. The aim of this study was to characterize DM1 in Aragón and to identify variations in the geographical pattern by gender.


A descriptive and ecological study was conducted to determine geographical variations in the DM1 incidence for the period 1991–2009. The source of information was the registry of DM1. To determine data completeness, a capture–recapture analysis was performed. Cases were georeferenced according to the Basic Healthcare Area (BHA) of residence. IRs for both genders, age group, 5 yr of diagnosis, and 95% confidence intervals (95% CIs) were calculated. Geographical pattern was studied applying Bayesian statistical model. The standardized incidence ratios (SIRs), smoothed SIR, and the posteriori risk probability (PRP) were represented cartographically for BHA stratified by gender.


The completeness was 93.5%. The global IR was 19.2 cases/105 person-years (95% CI: 17.6–20.8), boys 21.4 (95% CI: 19.0–23.8) and girls 16.8 (95% CI: 14.7–19.1). The age-specific rates were significantly lower in the 0–4 year age group with respect to the groups of 5–9 and 10–14 years.

For boys, areas with a statistically significant excess of risk were found in the north [smoothed SIR: 118–167.9 and PRP of what the smoothed SIR would be greater than 100 (PRP) above 0.8] and below average risk in the south (smoothed SIR: 65.9–79.1 and PRP less than 0.2).


DM1 IR presented a north–south geographical pattern in boys. This pattern was not observed in girls or when both genders were considered together. Later studies should include gender as an essential variable.