INSERM, U657, F-33000, Bordeaux, France.
Pharmacoepidemiol Drug Saf. 2012 Jun;21(6):573-83. doi: 10.1002/pds.3221. Epub 2012 Feb 15.
To determine healthcare claim patterns associated using nonsteroidal anti-inflammatory drugs (NSAIDs) for rheumatoid arthritis (RA).
The CADEUS study randomly identified NSAID users within the French health insurance database. One-year claims data were extracted, and NSAID indication was obtained from prescribers. Logistic regression was used in a development sample to identify claim patterns predictive of RA and models applied to a validation sample. Analyses were stratified on the dispensation of immunosuppressive agents or specific antirheumatism treatment, and the area under the receiver operating characteristic curve was used to estimate discriminant power.
NSAID indication was provided for 26,259 of the 45,217 patients included in the CADEUS cohort; it was RA for 956 patients. Two models were constructed using the development sample (n = 13,143), stratifying on the dispensation of an immunosuppressive agent or specific antirheumatism treatment. Discriminant power was high for both models (AUC > 0.80) and was not statistically different from that found when applied to the validation sample (n = 13,116).
The models derived from this study may help to identify patients prescribed NSAIDs who are likely to have RA in claims databases without medical data such as treatment indication.
确定与类风湿关节炎（RA）相关的非甾体抗炎药（NSAIDs）使用的医疗保健索赔模式。
CADEUS 研究在法国健康保险数据库中随机确定 NSAID 用户。提取了一年的索赔数据，并从处方医生处获得了 NSAID 适应证。在开发样本中使用逻辑回归来确定预测 RA 的索赔模式，并将模型应用于验证样本。在免疫抑制剂或特定抗风湿治疗的分配基础上进行分析，并使用受试者工作特征曲线下的面积来估计判别能力。
CADEUS 队列中包括的 45217 名患者中有 26259 名患者提供了 NSAID 适应证；其中有 956 名患者为 RA。使用开发样本（n=13143）构建了两个模型，对免疫抑制剂或特定抗风湿治疗的分配进行分层。两个模型的判别能力均较高（AUC>0.80），且与应用于验证样本（n=13116）时的判别能力无统计学差异。
该研究得出的模型可能有助于在没有治疗适应证等医疗数据的情况下，在索赔数据库中识别出开具 NSAIDs 的可能患有 RA 的患者。
