Möglichkeiten und Grenzen der Evaluation von Krebsfrüherkennungsuntersuchungen anhand von Routinedaten der Gesetzlichen Krankenversicherung in Deutschland

Abstract

In Germany, in 2014, about 223,000 people died of cancer. The prevention of cancers, therefore, offers the potential to considerably reduce the societal burden of these diseases. Cancer screening tests and examinations aim to detect cancer or precancerous conditions at an early stage and thus to increase chances for treatment, cure and survival. Since early detection strategies have both beneficial and adverse effects, they need to be monitored and evaluated continuously. This cumulative dissertation examined the opportunities and limitations of German health insurance claims data for monitoring and evaluating cancer screening tests and examinations implemented on a national scale. Based on claims data provided by the BARMER, specific research questions were addressed in four separate publications which came to the conclusions that i) the coding quality of outpatient breast, prostate and colorectal cancer diagnoses in health insurance claims data is limited, ii) the number of incident breast, prostate and colorectal cancer cases identified in health insurance claims data depends on the length of the lookback period, the length of the confirmation period for confirmatory events and the definition of confirmatory events, iii) health insurance claims data-based comorbidity measures are insufficient for controlling selection bias in observational studies of mammography screening, and iv) health insurance claims data are suitable for assessing and explaining geographic variations in mammography screening participation. Health insurance claims data offer opportunities for monitoring and evaluating cancer screening tests and examinations thanks to their ready availability, large study populations, long observation periods, high level of detail, diversity of applicable study designs and diversity of examinable outcomes. Limitations arise from incomplete information on confounders, missing information on the cause of death and tumor stages, missing internal and external validation studies, selection bias, misclassification, restrictions on internal and external validity as well as data protection regulations. To fully exploit the potential of health insurance claims data for monitoring and evaluating cancer screening tests and examinations implemented nationwide, the data should be linked to survey and cancer registry data. Moreover, further validation studies should be conducted and the methods for monitoring and evaluating national cancer screening tests and examinations should be standardized focusing particularly on the control of selection bias and quantification of effects.