Institut für Pharmakoökonomie und Arzneimittellogistik, Hochschule Wismar, Wismar, Germany.
Eur J Health Econ. 2013 Jun;14(3):551-68. doi: 10.1007/s10198-012-0410-y. Epub 2012 Jul 20.
The purpose of this study was to describe the methodological framework underlying nonadherence (NA) measurement based on pharmacy claims data, its quantitative impact on the results of NA studies, and to identify those methodological categories most likely to explain diabetes-related clinical outcomes. We use the example of oral antidiabetics in the treatment of diabetes mellitus type 2; 113,108 patients derived from a German statutory health insurance fund were analyzed.
We identified 12 methodological categories as pervasive features in pharmacy claims data based NA analyses. The influence of the different methodological categories and their parameters on analysis results was tested using sensitivity analysis. To validate alternative methodological framework options, we performed multivariate logistical regression estimates using diabetes-related hospitalization/clinical events as a combined dichotomized dependent variable.
The choice of parameters within the identified 12 methodological categories available has exceptional impact on the results of pharmacy data based claims NA analyses. When the full range of theoretically possible cases is considered in our sample, it can be seen that the resulting NA range is between 15.7% and 97.0%. The definition of the required daily dose, the decision to use either a prescription-/interval-based approach, and the classes of medication analyzed exert a notable influence on the study results. In our analysis, 69.4% of the 216 different study design options analyzed significantly explain the likelihood of diabetes-related clinical events.
We recommend strongly that methodological transparency is awarded a much more important role in the conduct of NA analyses made on the basis of pharmacy claims data.
本研究旨在描述基于药房理赔数据的不依从（NA）测量的方法框架，及其对 NA 研究结果的定量影响，并确定最有可能解释糖尿病相关临床结局的方法类别。我们以德国法定健康保险基金中治疗 2 型糖尿病的口服抗糖尿病药物为例，分析了 113108 名患者。
我们根据基于药房理赔数据的 NA 分析确定了 12 种方法类别作为普遍特征。使用敏感性分析测试了不同方法类别及其参数对分析结果的影响。为了验证替代方法框架选项，我们使用多元逻辑回归估计，将糖尿病相关住院/临床事件作为一个综合的二分类因变量。
在所确定的 12 种方法类别中，参数的选择对基于药房理赔数据的 NA 分析结果有特殊影响。当考虑我们样本中理论上所有可能的情况时，可以看出，由此产生的不依从率在 15.7%至 97.0%之间。所需日剂量的定义、决定使用处方/间隔为基础的方法以及分析的药物类别对研究结果有显著影响。在我们的分析中，69.4%的 216 种不同的研究设计方案显著解释了糖尿病相关临床事件的可能性。
我们强烈建议，在基于药房理赔数据进行的不依从分析中，方法学透明度应发挥更重要的作用。
