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

Abstract

BackgroundSeveral 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.ObjectiveOur 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.MethodsWe 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.ResultsThe 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.ConclusionsThree-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.Electronic supplementary materialThe online version of this article (doi:10.1007/s41669-017-0049-9) contains supplementary material, which is available to authorized users.

Mapping the Paediatric Quality of Life Inventory (PedsQL™) Generic Core Scales onto the Child Health Utility Index–9 Dimension (CHU-9D) Score for Economic Evaluation in Children

“…As well as the three functional forms chosen, other models have been used for mapping such as the beta-binomial estimator and finite mixture models used to accommodate skewed distributions [15]. However, neither of these models have been shown to be better than GLM or OLS when predicting utility value at near perfect health state [31,32]. Furthermore, the MM-estimator [33] has the potential to cope with heteroscedasticity and the undesired effect of outliers within the estimation sample, and has been shown to have the lowest predictive error in a previous paper that mapped PedsQL TM onto CHU-9D in an older population [15].…”

Exploring the Benefits of Transformations in Health Utility Mapping

“…It is important to mention here that all the investigation in the present study does not consider any issues related to measurement error, which could be present if the instruments involved as target or source do not measure the respective construct accurately. We do that for 2 reasons: first, all instruments involved in the real data used here have been previously validated and found to be reliable; second, we follow the practice of similar studies [15][16][17][18] of comparing and investigating the benefits of some statistical models (here OLS with transformations) under the assumption of negligible measurement error, which is a typical and important first step for this type of investigation. However, even though we consider it to be outside of the scope of our study, we recognize the importance of studying the effects of measurement error to the performance and validity of health utility mapping applications and related statistical methods used.…”

A value‐based healthcare approach: Health‐related quality of life and psychosocial functioning in women with Turner syndrome

“…In case of collinearity, we entered the variable with the strongest correlation with the outcome into the multivariate linear regression analysis. As previously described the tri‐modal distribution of the EQ‐5D outcome measure hampers ordinary least‐squares (OLS) regression, three‐part regression methods seem to have better prediction power than OLS with EQ‐5D data, although OLS seems quite robust . Therefore, we conducted separate linear regression for the .5‐.99 range as an additional ‘sensitivity analysis’.…”