Developing mapping algorithms to predict EQ-5D health utility values from Bath Ankylosing Spondylitis Disease Activity Index and Bath Ankylosing Spondylitis Functional Index among patients with Ankylosing Spondylitis

Table 3 Predictive performance of direct and indirect mapping models for the EQ‑5D‑3L (N = 238)

Model type	Mean	SD	Min	Max	MAE	RMSE	Spearman	CCC	Average rank
EQ-5D-3L value	0.8811	0.1207	0.1702	1.0000
Full sample
OLS	0.8811	0.0466	0.7590	0.9444
CLAD	0.8788	0.0737	0.6788	1.0002
Tobit	0.9100	0.0680	0.7317	1.0107
ALDVMM	0.8799	0.0468	0.7463	0.9427
BM	0.8944	0.0469	0.7254	0.9491
Ologit	0.8791	0.0896	0.5966	0.9868
Mlogit	0.8795	0.0881	0.6573	0.9882
Cross validation
OLS	0.8810	0.0473	0.7383	0.9499	0.0869	0.1130	0.4059	0.2410	5
CLAD	0.9100	0.0687	0.7077	1.0266	0.0850	0.1197	0.4057	0.2890	5
Tobit	0.8958	0.0745	0.6847	1.0207	0.0856	0.1174	0.4158	0.3230	3
ALDVMM	0.8788	0.0501	0.7224	0.9483	0.0896	0.1168	0.4114	0.2360	7
BM	0.8677	0.0553	0.7175	0.9569	0.0858	0.1132	0.4195	0.2810	3
Ologit	0.8786	0.0901	0.5758	0.9885	0.0848	0.1202	0.4365	0.3640	2
Mlogit	0.8786	0.0893	0.6124	0.9905	0.0836	0.1199	0.4395	0.3630	1

CLAD censored least absolute deviations, ALDVMM adjusted limited dependent variable mixture model, BM beta-mixture model, Ologit ordered logit, Mlogit multinomial logit. Bold indicates the best model in this performance indicator

ISSN: 1477-7525