COMPAQH, CERMES, INSERM U988, Institut Gustave Roussy, Villejuif, France.
Med Care. 2011 Jun;49(6):569-76. doi: 10.1097/MLR.0b013e31820fc386.
To determine the impact on hospital ranking of different aggregation methods when creating a composite score from a set of quality indicators relating to a single clinical condition.
The analysis was based on 14966 medical records taken from all French hospitals that treated over 30 patients with acute myocardial infarction in 2008 (n=275). Five quality indicators measuring the quality of care delivered to patients with acute myocardial infarction at hospital discharge were aggregated by 5 methods issued from a variety of activity sectors (indicator average, all-or-none, budget allocation process, benefit of the doubt, and unobserved component model).
Each aggregation method was used to rank hospitals into 3 categories depending on the position of the 95% confidence interval of the composite score relative to the overall mean. Variations in rank according to method were estimated using weighted κ coefficients.
Agreement between methods ranged from poor (κ=0.20) to almost perfect (κ=0.84). A change of method led to a change in rank for 71% (196 of 275) of hospitals. Only 14 of 121 hospitals which were ranked top and 20 of 118 which were ranked bottom, by at least 1 of the 5 methods, held their rank on a switch to the 4 other methods.
Hospital ranking varied widely according to 5 aggregation methods. If one method has to be chosen, for instance for reporting to governments, regulatory agencies, payers, health care professionals, and the public, it is necessary to provide its rationale and characteristics, and information on score uncertainty.
确定在创建与单一临床情况相关的一组质量指标的综合评分时，不同聚合方法对医院排名的影响。
该分析基于 2008 年法国所有治疗超过 30 例急性心肌梗死患者的医院（n=275）的 14966 份病历。使用来自不同活动领域的 5 种聚合方法聚合了 5 个衡量患者出院时急性心肌梗死护理质量的质量指标（指标平均值、全有或全无、预算分配过程、怀疑利益和未观察到的组成部分模型）。
使用每种聚合方法根据综合评分的 95%置信区间相对于总体平均值的位置将医院分为 3 类。使用加权 κ 系数估计方法之间的排名变化。
方法之间的一致性范围从差（κ=0.20）到几乎完美（κ=0.84）。方法的改变导致 71%（275 个中的 196 个）的医院排名发生变化。只有在至少 5 种方法中的 1 种方法中排名前 121 位的医院中的 14 家，以及在至少 1 种方法中排名后 118 位的医院中的 20 家，在切换到其他 4 种方法时保持其排名。
根据 5 种聚合方法，医院排名差异很大。如果必须选择一种方法，例如向政府、监管机构、付款人、医疗保健专业人员和公众报告，那么就有必要提供其基本原理和特点，以及有关评分不确定性的信息。
