Institute for Choice, University of South Australia Business School, Level 3 Way Lee Building, North Terrace, Adelaide, SA, 5001, Australia.
Centre for Health Economics, Monash Business School, Monash University, Melbourne, Australia.
Pharmacoeconomics. 2019 Sep;37(9):1139-1153. doi: 10.1007/s40273-019-00808-2.
Mapping algorithms have been indicated as a second-best solution for estimating health state utilities for the calculation of quality-adjusted life-years within cost-utility analysis when no generic preference-based measure is incorporated into the study. However, the predictive performance of these algorithms may be variable and hence it is important to assess their external validity before application in different settings.
The aim of this study was to assess the external validity and generalisability of existing mapping algorithms for predicting preference-based Child Health Utility 9D (CHU9D) utilities from non-preference-based Pediatric Quality of Life Inventory (PedsQL) scores among children and adolescents living with or without disabilities or health conditions.
Five existing mapping algorithms, three developed using data from an Australian community population and two using data from a UK population with one or more self-reported health conditions, were externally validated on data from the Longitudinal Study of Australian Children (n = 6623). The predictive accuracy of each mapping algorithm was assessed using the mean absolute error (MAE) and the mean squared error (MSE).
Values for the MAE (0.0741-0.2302) for all validations were within the range of published estimates. In general, across all ages, the algorithms amongst children and adolescents with disabilities/health conditions (Australia MAE: 0.2085-0.2302; UK MAE: 0.0854-0.1162) performed worse relative to those amongst children and adolescents without disabilities/health conditions (Australia MAE: 0.1424-0.1645; UK MAE: 0.0741-0.0931).
The published mapping algorithms have acceptable predictive accuracy as measured by MAE and MSE. The findings of this study indicate that the choice of the most appropriate mapping algorithm to apply may vary according to the population under consideration.
当没有通用偏好量表纳入研究时，映射算法已被认为是在成本效用分析中估算健康状态效用以计算质量调整生命年的次优解决方案。然而，这些算法的预测性能可能存在差异，因此在不同的环境中应用之前，评估其外部有效性非常重要。
本研究旨在评估现有的映射算法对于预测有无残疾或健康状况的儿童和青少年从非偏好量表儿科生活质量量表（PedsQL）得分得出偏好量表儿童健康效用 9D（CHU9D）效用的外部有效性和通用性。
对来自澳大利亚社区人群的三项数据和来自英国一项或多项自我报告健康状况人群的两项数据开发的五项现有映射算法进行了外部验证，数据来自澳大利亚儿童纵向研究（n=6623）。使用平均绝对误差（MAE）和均方误差（MSE）评估每种映射算法的预测准确性。
所有验证的 MAE 值（0.0741-0.2302）均在已发表估计值的范围内。一般来说，在所有年龄段中，残疾/健康状况儿童和青少年的算法（澳大利亚 MAE：0.2085-0.2302；英国 MAE：0.0854-0.1162）的表现均差于无残疾/健康状况儿童和青少年的算法（澳大利亚 MAE：0.1424-0.1645；英国 MAE：0.0741-0.0931）。
从 MAE 和 MSE 衡量，发表的映射算法具有可接受的预测准确性。本研究的结果表明，应用最合适的映射算法的选择可能因所考虑的人群而异。
