Mapping the Seattle Angina Questionnaire to EQ-5D-5L in patients with coronary heart disease

Table 3 Goodness of fit of direct and indirect mapping approach from full sample

Model	Mean	Minimum	Maximum	Mean changes between baseline and follow-up survey	MAE	RMSE	ρ	CCC
Observed	0.8725	0.1690	1.0000	0.1256	-	-	-	-
OLS	0.8725	0.5302	1.0877	0.1304	0.0656	0.0889	0.7846	0.7621
Tobit	0.9090	0.4843	1.2055	0.1990	0.0843	0.1058	0.7788	0.7539
GLM	0.8729	0.6004	1.0980	0.1313	0.0678	0.0923	0.7653	0.7331
CLAD	0.8781	0.6384	1.0730	0.1444	0.0646	0.0960	0.7469	0.6910
RMM	0.9008	0.7057	1.0555	0.0993	0.0653	0.1066	0.7284	0.5797
BM	0.8734	0.3573	0.9990	0.1160	0.0566	0.0813	0.8267	0.8063
ALDVMM	0.8776	0.5616	0.9991	0.1130	0.0592	0.0877	0.8010	0.7529
Indirect approach	0.8663	0.2985	0.9961	0.1176	0.0572	0.0796	0.8333	0.8243

Notes: Bold number indicate a best result on that indicator. OLS, order least square; GLM, generalized linear model; CLAD, censored least absolute deviations; RMM, MM-robust regression; BM, mixture beta regression model; ALDVMM, adjusted limited dependent variable mixture model. ρ, Pearson’s correlation coefficients between observed EQ-5D-5L scores and prediction; MAE, mean absolute error; RMSE, root mean square error, CCC, Lin’s Concordance correlation coefficient

ISSN: 1477-7525