O'Brien Sean M, Gauvreau Kimberlee
Department of Biostatistics and Bioinformatics and Duke Clinical Research Institute, Duke University Medical Center, Durham, North Carolina 27715, USA.
Cardiol Young. 2008 Dec;18 Suppl 2:145-51. doi: 10.1017/S1047951108002898.
It is universally agreed that efforts to improve quality benefit from the analysis of outcomes. Yet, it is challenging to compare results across institutions because factors other than performance also impact outcomes. Two factors that complicate the analysis of outcomes after congenital cardiac surgery are case-mix and random statistical variation. Case-mix refers to differences in the mix of patients and their risk-factors at different institutions that may cause some centres to have more frequent complications and lower survival regardless of their true performance. Random statistical variation refers to fluctuations in outcomes that occur at random and follow the laws of probability. A variety of statistical methods exist to address these issues and make provider comparisons more fair. We explain a few common approaches including stratification, regression analysis, and confidence intervals. Concepts are illustrated using artificial data from two hypothetical hospitals, as well as real data from a multi-institution registry.
人们普遍认为，通过对结果进行分析，有助于提高质量。然而，跨机构比较结果具有挑战性，因为除了医疗表现之外，其他因素也会影响结果。先天性心脏手术后结果分析复杂化的两个因素是病例组合和随机统计变异。病例组合指不同机构患者及其风险因素组合的差异，这可能导致一些中心出现更频繁的并发症和更低的生存率，而与它们的实际医疗表现无关。随机统计变异指结果中随机出现且遵循概率法则的波动。存在多种统计方法来解决这些问题，并使医疗服务提供者之间的比较更加公平。我们解释一些常见方法，包括分层、回归分析和置信区间。使用来自两家假设医院的人工数据以及来自多机构登记处的真实数据来说明这些概念。
