Regional differences in diabetes across Europe - regression and causal forest analyses

We examine regional differences in diabetes within Europe, and relate them to variations in socio-economic conditions, comorbidities, health behaviour and diabetes management. We use the SHARE (Survey of Health, Ageing and Retirement in Europe) data of 15 European countries and 28,454 individuals, who participated both in the 4th and 7th (year 2011 and 2017) waves of the survey. First, we estimate multivariate regressions, where the outcome variables are diabetes prevalence, diabetes incidence, and weight loss due to diet as an indicator of management. Second, we study the heterogeneous impact of demographic, socio-economic, health and lifestyle indicators on the regional differences in diabetes incidence with causal random forests. Compared to Western Europe, the odds of a new diabetes diagnosis over a six-year horizon is 2.2-fold higher in Southern and 2.6-fold higher in Eastern Europe. Adjusting for individual characteristics, the odds ratio decreases to 1.8 in the South-West and to 2.0 in the East-West dimension. These remaining differences are mostly explained by country-specific healthcare indicators. Based on the causal forest approach, the adjusted East-West difference is essentially zero for the lowest risk groups (tertiary education, employment, no hypertension, no overweight) and increases substantially with these risk factors, but the South-West difference is much less heterogeneous. The prevalence of diet-related weight loss around the time of diagnosis also exhibits regional variation. The results suggest that the regional differences in diabetes incidence could be reduced by putting more emphasis on diabetes prevention among high-risk individuals in Eastern and Southern Europe.