eTable 1. Means (standard deviations) of actual and estimated EuroQol 5-dimension three-level utilities, R2 values, product-moment correlations, intra-class correlations, mean squared errors, and mean absolute errors for models in the PROMIS Wave 1 sample (N = 13,955) and Cleveland Clinic neurological disease patient sample (N = 6,906).

Mean (SD) / R2 / r / ICC / MSE / MAE
PROMIS
Wave 1
Sample / Actual / 0.817 (0.175) / -- / -- / -- / -- / --
REVReg / 0.724 (0.104) / 0.616 / 0.785 / 0.530 / 0.021 / 0.127
REVLE / 0.812 (0.169) / 0.615 / 0.785 / 0.784 / 0.013 / 0.083
REVequip / 0.809 (0.170) / 0.606 / 0.778 / 0.777 / 0.013 / 0.075
CATReg / 0.817 (0.142) / 0.664 / 0.815 / 0.798 / 0.010 / 0.076
CATLE / 0.814 (0.171) / 0.662 / 0.814 / 0.813 / 0.011 / 0.077
CATequip / 0.808 (0.167) / 0.648 / 0.805 / 0.803 / 0.012 / 0.069
Tobit / 0.852 (0.178) / 0.373 / 0.611 / 0.594 / 0.026 / 0.121
GAM / 0.817 (0.137) / 0.622 / 0.789 / 0.767 / 0.012 / 0.081
PLR / 0.832 (0.157) / 0.622 / 0.789 / 0.752 / 0.014 / 0.067
Cleveland
Clinic
Neurological
Disease
Sample / Actual / 0.634 (0.231) / -- / -- / -- / -- / --
REVReg / 0.619 (0.110) / 0.566 / 0.752 / 0.580 / 0.028 / 0.144
REVLE / 0.638 (0.184) / 0.566 / 0.752 / 0.733 / 0.023 / 0.123
REVequip / 0.634 (0.216) / 0.567 / 0.753 / 0.751 / 0.025 / 0.114
CATReg / 0.665 (0.168) / 0.613 / 0.783 / 0.734 / 0.022 / 0.115
CATLE / 0.630 (0.205) / 0.613 / 0.783 / 0.778 / 0.021 / 0.114
CATequip / 0.633 (0.213) / 0.608 / 0.780 / 0.777 / 0.022 / 0.107
Tobit / 0.673 (0.184) / 0.600 / 0.775 / 0.739 / 0.023 / 0.120
GAM / 0.687 (0.162) / 0.552 / 0.743 / 0.667 / 0.027 / 0.124
PLR / 0.683 (0.202) / 0.552 / 0.743 / 0.687 / 0.030 / 0.113

SD – standard deviation

R2 – square of the Pearson product-moment correlation between the observed and estimated EQ-5D utilities

r – Pearson product-moment correlation between the observed and estimated EQ-5D utilities

ICC – Intra-class correlation between the observed and estimated EQ-5D utilities

MSE – Mean Squared Error

MAE – Mean Absolute Error

REVReg – Revicki’s original linear regression model 3

REVLE – Revicki’s original linear regression model 3 with linear equating applied to the predicted values

REVequip – Revicki’s original linear regression model 3 with equipercentile equating applied to the predicted values

CATReg – Linear Regression model where PROMIS Global Health items are treated as categorical predictors

CATLE – Model that applies linear equating to the predicted values of a linear regression model where PROMIS Global Health items are treated as categorical predictors

CATequip – Model that applies equipercentile equating to the predicted values of a linear regression model where PROMIS Global Health items are treated as categorical predictors

Tobit – Tobit model where PROMIS Global Health items are treated as categorical predictors

GAM – Generalized additive model that applies smoothing functions to the PROMIS Global Health Physical and Mental Health T-scores. For each T-score, the smoothing function target equivalent degrees of freedom was set to 4

PLR – Each of the five EQ-5D-3L domain scores were estimated by polytomous logistic regression. For each subject and each EQ-5D-3L domain model, the most probable EQ-5D-3L domain response was inputted to the EQ-5D-3L disutility function based on the US version of the algorithm