OBJECTIVE: The aim of this study was to develop a response mapping algorithm to predict EQ-5D-5L utilities from European Organisation for Research and Treatment Quality of Life Questionnaire Core 30 (EORTC QLQ-C30) scores and compare performance with direct mapping approaches to identify the best performing algorithm. METHODS: The Multi-Instrument Comparison dataset contains responses to both the EQ-5D-5L and QLQ-C30 questionnaires from 692 individuals with a broad range of cancers. Response mapping was conducted, fitting ordered logistic regressions to predict response levels for each of the five EQ-5D dimensions and utilities were predicted using the US and Japanese EQ-5D-5L value sets to test the algorithm performance. Various direct mapping models were fitted: ordinary least squares, tobit, two-part (TPM), adjusted limited dependent variable mixture and beta mixture models. Model assessment and recommendations regarding the best mapping algorithm was based on goodness-of-fit statistics, predictive ability (measures of error, distribution of predicted utilities) and in sample cross-validation. RESULTS: The response mapping model performed well in terms of predictive ability and measurement error using the US or Japanese value set, with mean absolute error ranging from 0.0708 to 0.0988, and comparably to the TPM, which was the best performing direct algorithm. CONCLUSION: The developed mapping algorithms enable the prediction of EQ-5D-5L utilities from QLQ-C30 scores when EQ-5D-5L data have not been directly collected in clinical trials. The response mapping model offers the possibility of predicting EQ-5D-5L utility values using any national value set and can be generalised to multiple countries and oncology settings.
OBJECTIVE: The aim of this study was to develop a response mapping algorithm to predict EQ-5D-5L utilities from European Organisation for Research and Treatment Quality of Life Questionnaire Core 30 (EORTC QLQ-C30) scores and compare performance with direct mapping approaches to identify the best performing algorithm. METHODS: The Multi-Instrument Comparison dataset contains responses to both the EQ-5D-5L and QLQ-C30 questionnaires from 692 individuals with a broad range of cancers. Response mapping was conducted, fitting ordered logistic regressions to predict response levels for each of the five EQ-5D dimensions and utilities were predicted using the US and Japanese EQ-5D-5L value sets to test the algorithm performance. Various direct mapping models were fitted: ordinary least squares, tobit, two-part (TPM), adjusted limited dependent variable mixture and beta mixture models. Model assessment and recommendations regarding the best mapping algorithm was based on goodness-of-fit statistics, predictive ability (measures of error, distribution of predicted utilities) and in sample cross-validation. RESULTS: The response mapping model performed well in terms of predictive ability and measurement error using the US or Japanese value set, with mean absolute error ranging from 0.0708 to 0.0988, and comparably to the TPM, which was the best performing direct algorithm. CONCLUSION: The developed mapping algorithms enable the prediction of EQ-5D-5L utilities from QLQ-C30 scores when EQ-5D-5L data have not been directly collected in clinical trials. The response mapping model offers the possibility of predicting EQ-5D-5L utility values using any national value set and can be generalised to multiple countries and oncology settings.
Authors: A Simon Pickard; Ernest H Law; Ruixuan Jiang; Eleanor Pullenayegum; James W Shaw; Feng Xie; Mark Oppe; Kristina S Boye; Richard H Chapman; Cynthia L Gong; Alan Balch; Jan J V Busschbach Journal: Value Health Date: 2019-05-25 Impact factor: 5.725
Authors: Clara Mukuria; Donna Rowen; Sue Harnan; Andrew Rawdin; Ruth Wong; Roberta Ara; John Brazier Journal: Appl Health Econ Health Policy Date: 2019-06 Impact factor: 2.561
Authors: Hosein Ameri; Mahmood Yousefi; Mehdi Yaseri; Azin Nahvijou; Mohammad Arab; Ali Akbari Sari Journal: Expert Rev Pharmacoecon Outcomes Res Date: 2018-09-03 Impact factor: 2.217
Authors: N K Aaronson; S Ahmedzai; B Bergman; M Bullinger; A Cull; N J Duez; A Filiberti; H Flechtner; S B Fleishman; J C de Haes Journal: J Natl Cancer Inst Date: 1993-03-03 Impact factor: 13.506