Estimation of the prevalence of chronic kidney disease in people with diabetes by combining information from multiple routine data collections

Health care claims databases maintained by social insurance institutions provide rich and sometimes easily accessible data sources for epidemiological research. Interpreting the registered claims, for example, drug prescriptions, as proxies for the condition of interest, for example, diabetes, they allow for nationwide prevalence estimation. We illustrate a more subtle use of health care claims data in estimating the stage-specific prevalence of chronic kidney disease in the Austrian population with diabetes. The main difficulty was that information on the type of disease (chronic or acute) and information on the stage of disease were only available for small, almost disjoint subsets of the health care claims data. Using high-dimensional regression models, we could combine the information and provide nationwide estimates of the stage-specific prevalence of diabetic chronic kidney disease. Validating our estimates by comparing to other studies, we found the level of agreement satisfying.

Automated summary

Health care claims databases maintained by social insurance institutions are valuable resources for epidemiological research, providing nationwide prevalence estimates for conditions like diabetes. Using high-dimensional regression models, information from small subsets of the data can be combined to estimate stage-specific prevalence rates for diseases such as chronic kidney disease in diabetic populations. Validation through comparison with other studies revealed satisfactory agreement levels.