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