Direct Mapping of the QLQ-C30 to EQ-5D Preferences: A Comparison of Regression Methods

• DOI: 10.1007/s41669-017-0049-9
• Publication Date: 2017-08-07T02:28:26Z
• Authors: Ralph Crott
• Keywords: Ordinary least squares; Regression diagnostic; Censored regression model; Local regression

Several mapping or cross-walking algorithms for deriving utilities from the European Organisation Research and Treatment of Cancer Quality Life Questionnaire (EORTC QLQ-C30) scores have been published in recent years. However, large majority used ordinary least squares (OLS) regression, which proved to be not very accurate because specifics quality-of-life measures.Our objective was compare regression methods that map EuroQol 5 Dimensions 3 Levels (EQ-5D-3L) utility values general EORTC QLQ-C30 using OLS as a benchmark while fixing number explanatory variables explore an alternative three-part model.We conducted analysis predicted EQ-5D-3L generated data observational study ambulatory patients with non-small-cell lung cancer Toronto hospital. Six were compared simple benchmark. The six models Tobit, censored absolute deviation, normal mixture, beta, zero-one inflated beta mix piecewise logistic regression.The best predictive fit obtained by regression(s) lower than 1 cut-off point 0.50 separate binary equal one. Zero-one also promising. most mean. prediction one poor all approaches.Three-part separately target low, medium high (<0.50, 0.51-0.99 1) seem better power data, although seems quite robust. Exploration approaches single should further tested other similar datasets individual pooled various clinical studies. use goodness-of-fit measures studies their influence on choice performing investigated.