National Institute of Public Health of Québec, 945, Wolfe Avenue, Quebec, QC G1V 5B3, Canada.
Osteoporos Int. 2012 Feb;23(2):483-501. doi: 10.1007/s00198-011-1559-4. Epub 2011 Feb 19.
Physician-billing claims databases can be used to determine the incidence of fractures in the community. This study tested three algorithms designed to accurately and reliably identify fractures from a physician-billing claims database and concluded that they were useful for identifying all types of fractures, except vertebral, sacral, and coccyx fractures.
To develop and validate algorithms that identify fracture events from a physician-billing claims database (PCDs).
Three algorithms were developed using physician's specialty, diagnostic, and medical service codes used in a PCD from the province of Quebec. Algorithm validity was assessed via calculation of positive predictive values (PPV; via verification of a sample of algorithm-identified cases with hospitalization files) and sensitivities (via cross-referencing respective algorithm-identified fracture cases with a well-characterized fracture cohort).
PPV and sensitivity varied across fracture sites. For most fracture sites, the PPV with algorithm 3 was higher than with algorithms 1 or 2. Except for knee fracture, the PPVs ranged from 0.81 to 0.96. Sensitivities were low at the vertebral, sacral, and coccyx sites (0.40-0.50), but high at all other fracture sites. For 95% of fractures, the fracture site identified by algorithm agreed with the fracture site from patients' medical records. Fracture dates identified by algorithm were within 2 days of the actual fracture date in 88% of fracture cases. Among cases identified by algorithm 3 to have had an open reduction (N = 461), 95% underwent surgery according to their respective medical charts.
Algorithms using PCDs are accurate and reliable for identifying incident fractures associated with osteoporosis-related fracture sites. The identification of these fractures in the community is important for helping to estimate the burden associated with osteoporosis and the utility of programs designed to reduce the rates of fragility fracture.
医师计费索赔数据库可用于确定社区中骨折的发生率。本研究测试了三种旨在从医师计费索赔数据库（PCD）中准确可靠地识别骨折的算法，并得出结论，这些算法可用于识别所有类型的骨折，但不包括椎体、骶骨和尾骨骨折。
开发和验证从医师计费索赔数据库（PCD）中识别骨折事件的算法。
使用魁北克省 PCD 中的医师专业、诊断和医疗服务代码开发了三种算法。通过计算阳性预测值（PPV；通过验证算法识别的病例样本与住院档案）和敏感度（通过交叉引用各自算法识别的骨折病例与特征明确的骨折队列）来评估算法的有效性。
骨折部位的 PPV 和敏感度各不相同。对于大多数骨折部位，算法 3 的 PPV 高于算法 1 或 2。除了膝关节骨折外，PPV 范围为 0.81 至 0.96。椎体、骶骨和尾骨部位的敏感度较低（0.40-0.50），但其他所有骨折部位的敏感度均较高。对于 95%的骨折，算法识别的骨折部位与患者病历中的骨折部位一致。算法识别的骨折日期与实际骨折日期相差 2 天以内的占 88%的骨折病例。在算法 3 识别出的需要进行切开复位的病例中（N=461），根据各自的病历，95%的病例接受了手术。
使用 PCD 的算法可准确可靠地识别与骨质疏松性骨折部位相关的偶发性骨折。在社区中识别这些骨折对于帮助估计与骨质疏松症相关的负担以及减少脆性骨折发生率的计划的效用非常重要。
Zotero 插件