Evaluation of an indirect record linkage method in an exemplary study on the risk of cancer and cancer mortality in type 2 diabetes patients treated with different antidiabetics

法定健康保险提供者的索赔数据是卫生服务流行病学研究和药物安全研究的重要资源。这些数据的好处在于患者数量多，随访时间长，有机会调查非常罕见的药物不良反应和潜伏期长的药物不良反应。由于收集这些数据是为了计费，而不是主要用于研究目的，因此缺乏有关肿瘤分期或死亡原因等疾病的重要信息。此外，数据中肿瘤诊断的有效性尚不清楚。索赔数据与癌症登记数据的联系提供了缺失的信息，从而对致癌物质影响的调查作出了重要贡献。在本研究提案中，对接受不同抗糖尿病治疗的2型糖尿病患者的癌症风险和癌症死亡率进行了示范性研究，旨在评估使用不同的方法将假名索赔数据与德国流行病学癌症登记处的数据联系起来，可以在多大程度上促进更容易的数据访问。首先，将使用出生年份、性别、社区代码和肿瘤诊断日期/季度进行间接记录链接。其次，结果将通过使用所谓的控制数字的一个完善的、精确的、复杂的联系程序来评估。由于这种关联需要非常严格的数据保护概念，并且非常耗时，因此并不是对每个研究问题都可行。使用示例性研究，将描述和评估两个关联数据集之间的质量差异。本项目的总体目标是优化这两种方法的关联程序，以便为开展具有特定数据保护要求的进一步研究提供成本和时间效率方面的建议。此外，还将提出建议，说明在何种情况下，这两种联系程序中的任何一种将导致有效的调查结果。该项目的结果将分发给使用德国二手数据的研究人员。

Claims data of statutory health insurance providers are an important resource for epidemiological studies in health services and drug safety research. The benefit of these data stems from the large number of patients and the potential for long follow-up periods offering the opportunity to investigate both very rare adverse drug reactions and adverse drug reactions with long latency times. Important information about diseases such as tumor stage or cause of death are lacking since the data was collected for billing and not primarily for research purposes. Furthermore, the validity of tumor diagnoses in the data is unknown. A linkage of claims data with data from a cancer registry provides the missing information, thereby making an essential contribution to the investigation of the effects of carcinogenic substances. In this research proposal, an exemplary study of cancer risk and cancer mortality in type 2 diabetes patients treated with different antidiabetic therapies serves to evaluate to which extent easier data access can be facilitated using different methods of linking pseudonymized claims data to data from epidemiological cancer registries in Germany. First, an indirect record linkage will be performed using year of birth, sex, community code, and date/quarter of the tumor diagnosis. Second, the results will be evaluated by a well-established and exact, yet elaborate linkage procedure using so-called control numbers. As such a linkage requires a very stringent data protection concept and is very time-consuming, it is not feasible for every research question. Using the exemplary study, the differences in quality between the two linked datasets will be described and assessed. The overall objective of this project is to optimize the linkage procedures of both methods in order to give recommendations for the cost and time-efficient conduct of further studies with specific data protection requirements. Furthermore, recommendations will be given under which circumstances either one of the two linkage procedures will lead to valid findings. The results of the project will be disseminated to researchers using secondary data in Germany.