Purpose: Several preference-weighted measurement systems are available to estimate societal health state values for reference case economic evaluation. However, further information about their measurement properties is necessary to guide system choice. Our objective was to characterize the longitudinal validity of several systems among Spine Patient Outcomes Research Trial (SPORT) participants over a one-year period.

 Methods: Health state values obtained from EuroQol with US and UK scoring (EQ5D-US, EQ5D-UK), the Health Utilities Index with Mark 2 and 3 scoring (HUI2, HUI3), and SF-36-derived Quality of Well Being Scale (eQWB) and SF-6D at baseline (V₀) and follow-up (V_f ) were used to compute change scores (V_f -V₀ ) for each system.  Paired t-tests were used to assess differences in mean change scores between system pairs.  Spearman correlations were used to describe associations between change scores. We estimated effect size (mean change /standard deviation (SD) at baseline), standardized response mean (SRM = mean change /SD_{change _score}), and minimally important difference (MID) for each system. MID was defined using the mean change score for those who reported minor progress or worsening of their spine problem at follow-up.

Results: Among 1,696 individuals ages 18-93 (mean age 55, 50% female), 80% reported improvement by global transition rating with MID computed among the 446 who reported minimal change. Marked differences between change scores were noted for all systems except EQ5D-US and HUI2. All systems were strongly correlated with each other (0.6 to1.0). Results were:

 

	Mean Score Change (95%CI)	Effect Size	SRM	MID
EQ-5D-UK	0.308(0.290, 0.326)	0.952	0.818	0.096
EQ-5D-US	0.205(0.193, 0.216)	0.967	0.820	0.063
HUI3	0.258(0.244, 0.272)	0.972	0.872	0.100
HUI2	0.206(0.195, 0.218)	0.977	0.865	0.072
SF-6D	0.158(0.150, 0.165)	1.338	0.972	0.041
eQWB	0.094(0.090, 0.099)	1.730	1.013	0.024

Conclusion: Although change scores across systems were highly correlated marked differences were noted. Observed differences in longitudinal response characteristics may have important implications for study planning and cost-effectiveness analyses of interventions targeted at spine problems.