Objectives: Our objective was to evaluate the role of potential predictors in explaining spatial variation among diabetes hospitalization rates in Germany.
Study design: This was an ecological analysis using hospital routine data.
Methods: County-level hospitalization rates (n = 402) in 2015 were calculated based on the German Diagnosis Related Groups database. We used a funnel plot to identify counties with high hospitalization rates. To examine the impact of predictors such as socio-economic status or structure of primary care, we performed linear and logistic regression analyses.
Results: The crude hospitalization rate was 262 admissions per 100,000 population. In multivariable logistic models, we found the percentage of employees with academic degree (odds ratio [OR]: 0.72, 95% confidence interval [CI]: 0.56-0.91), high hospital bed rate (4th quartile vs 1st quartile; OR: 2.73, CI: 1.03-7.24), and diabetes prevalence (OR: 1.49, CI: 1.17-1.90) to be significant predictors for high hospitalization rates. In multivariable linear models, the percentage of unemployed (regression coefficient b: 4.79, CI: 0.81-8.78) and rurality (b: 0.52, CI: 0.19-0.85) explained the variation in addition to predictors from logistic regression. Primary care structure was not a significant predictor in multivariable models.
Conclusions: The non-significant impact of primary care in adjusted models casts the use of diabetes hospitalizations as indicators for access and quality of primary care into doubt. Diabetes hospitalizations may rather reflect demand for care.
Keywords: Ambulatory care–sensitive conditions; Diabetes complications; Quality indicators; Small-area analysis.
Copyright © 2019 The Royal Society for Public Health. Published by Elsevier Ltd. All rights reserved.