Mapping the Minnesota living with heart failure questionnaire (MLHFQ) to EQ-5D-5L in patients with heart failure

Table 2 Goodness of fit results from full estimation sample (N = 141)

Model Specification	Mean utility (SD)	Minimum	Maximum	P.25	Median	P.75	RMSE	MAE	Abs diff < 0.03	Abs diff < 0.05	ICC
Observed	0.6619	−0.2630	1.0000	0.5530	0.7080	0.8770
Model 1
OLS	0.6619	0.2421	0.8575	0.5396	0.6952	0.7966	0.2219	0.1782	9.9	16.2	0.670
GLM	0.6619	0.2421	0.8575	0.5396	0.6952	0.7966	0.2219	0.1782	9.9	16.2	0.670
CLAD	0.6849	0.2920	0.8679	0.5705	0.7160	0.8109	0.2234	0.1740	11.3	18.3	0.654
MFP	0.7164	0.2679	0.9254	0.5858	0.7520	0.8604	0.2288	0.1713	12.7	18.3	0.684
MM	0.7365	0.3104	0.9350	0.6124	0.7703	0.8732	0.2342	0.1724	12.7	19.7	0.673
BETA	0.7267	0.1593	0.9253	0.5885	0.8044	0.8912	0.2386	0.1778	10.6	16.2	0.705
Model 2
OLS	0.6619	0.2308	0.8631	0.5519	0.6818	0.8028	0.2185	0.1750	9.2	15.5	0.690
GLM	0.6619	0.2308	0.8631	0.5519	0.6818	0.8028	0.2185	0.1750	9.2	15.5	0.690
CLAD	0.7066	0.2762	0.9072	0.5964	0.7272	0.8473	0.2230	0.1689	10.6	19.0	0.689
MFP	0.7224	0.2920	0.9263	0.6132	0.7414	0.8653	0.2268	0.1684	11.3	16.9	0.688
MM	0.7373	0.2897	0.9429	0.6125	0.7586	0.8871	0.2313	0.1695	9.2	16.2	0.697
BETA	0.7256	0.1505	0.9335	0.6154	0.7887	0.9019	0.2349	0.1761	4.2	12.0	0.717
Model 3
OLS	0.6619	0.2360	0.8222	0.5697	0.6984	0.8222	0.2151	0.1721	11.3	16.9	0.707
GLM	0.6619	0.2360	0.8222	0.5697	0.6984	0.8222	0.2151	0.1721	11.3	16.9	0.707
CLAD	0.6967	0.2920	0.8370	0.6126	0.7322	0.8370	0.2217	0.1703	13.4	22.5	0.667
MFP	0.7086	0.2883	0.8655	0.6177	0.7473	0.8655	0.2204	0.1676	13.4	20.4	0.700
MM	0.7260	0.2719	0.8953	0.6278	0.7795	0.8953	0.2248	0.1695	9.9	16.9	0.718
BETA	0.7345	0.1775	0.9032	0.6098	0.8027	0.9032	0.2321	0.1745	10.6	16.9	0.731
Indirect mapping
OLOGIT	0.6443	0.4790	0.7680	0.5410	0.5920	0.7680	0.2363	0.1927	10.6	15.5	0.512

Dependant variable: EQ-5D-5 L utility score; Independent variables: Model 1 - MLHF total score; Model 2 – MLHF domain scores; Model 3 – MLHF item scores (Item 04, 17 and 21).
Abs diff. < 0.03 (0.05)% - proportion of predicted utilities whose absolute values deviate from the mean of the observed utility values by less than 0.03 (0.05); P.25 – 25th percentile; P.75 75th percentile, RMSE Root Mean Square Error; MAE Mean Absolute Error; ICC – Intra Class Correlation.
OLS Ordinary least square, GLM Generalized linear modelling, CLAD Censored least absolute deviations, MFP Multivariable fractional polynomials, MM Robust MM estimator, BETA Mixture beta regression model, OLOGIT ordered logit (indirect response mapping).

ISSN: 1477-7525