SINTEF, Department of Health Research, Abels gt 5, N-7465, Trondheim, Norway.
Frisch Centre, Oslo, Norway.
Soc Sci Med. 2010 Feb;70(3):439-446. doi: 10.1016/j.socscimed.2009.11.002. Epub 2009 Nov 18.
The performance of health service providers may be monitored by measuring productivity. However, the policy value of such measures may depend crucially on the accuracy of input and output measures. In particular, an important question is how to adjust adequately for case-mix in the production of health care. In this study, we assess productivity growth in Norwegian outpatient child and adolescent mental health service units (CAMHS) over a period characterized by governmental utilization of simple productivity indices, a substantial increase in capacity and a concurrent change in case-mix. We analyze the sensitivity of the productivity growth estimates using different specifications of output to adjust for case-mix differences. Case-mix adjustment is achieved by distributing patients into eight groups depending on reason for referral, age and gender, as well as correcting for the number of consultations. We utilize the nonparametric Data Envelopment Analysis (DEA) method to implicitly calculate weights that maximize each unit's efficiency. Malmquist indices of technical productivity growth are estimated and bootstrap procedures are performed to calculate confidence intervals and to test alternative specifications of outputs. The dataset consist of an unbalanced panel of 48-60 CAMHS in the period 1998-2006. The mean productivity growth estimate from a simple unadjusted patient model (one single output) is 35%; adjusting for case-mix (eight outputs) reduces the growth estimate to 15%. Adding consultations increases the estimate to 28%. The latter reflects an increase in number of consultations per patient. We find that the governmental productivity indices strongly tend to overestimate productivity growth. Case-mix adjustment is of major importance and governmental utilization of performance indicators necessitates careful considerations of output specifications.
服务提供者的绩效可以通过衡量生产力来进行监测。然而，这些措施的政策价值可能取决于投入和产出措施的准确性。特别是，一个重要的问题是如何在医疗保健生产中充分调整病例组合。在这项研究中，我们评估了挪威门诊儿童和青少年心理健康服务单位（CAMHS）在一段时期内的生产力增长，该时期的特点是政府利用简单的生产力指数、容量大幅增加以及病例组合同时发生变化。我们使用不同的输出规格来调整病例组合差异，从而评估生产力增长估计的敏感性。病例组合调整是通过将患者分为 8 组来实现的，分组依据是转诊原因、年龄和性别，以及对咨询次数进行校正。我们利用非参数数据包络分析（DEA）方法，隐含地计算出最大化每个单位效率的权重。技术生产力增长的 Malmquist 指数被估计，并且进行了 bootstrap 程序来计算置信区间和测试替代输出规格。该数据集包括 1998-2006 年期间 48-60 个 CAMHS 的非平衡面板。从简单的未调整患者模型（一个单一输出）得出的平均生产力增长估计值为 35%；通过病例组合调整（八个输出）将增长估计值降低至 15%。增加咨询次数将估计值提高到 28%。后者反映了每位患者咨询次数的增加。我们发现，政府的生产力指数往往会高估生产力增长。病例组合调整非常重要，政府对绩效指标的利用需要仔细考虑输出规格。
