How Robust Is the Health Utilities Index Mark 2 Utility Function?

Abstract

Purpose . The utility function for the Health Utilities Index Mark 2 (HUI2) system is based on preference measurements from a random sample of parents with exclusion of inconsistent respondents. Would results without exclusions or from a different group of parents have differed? Methods . Scores were obtained from parents of patients (n = 59) undergoing treatment for cancer. Mean scores from the 2 sets of parents were compared: parents of patients and parents from the general population. Three multiattribute utility functions were estimated. Mean scores for HUI2 states using the functions were compared. Results . Most differences in mean scores between different groups were not statistically significant (P < 0.05). Differences in parameter estimates among the 3 utility functions were 0.05 or less. The exponent on the power function for the parent-of-patient group was 2.16, within 6% of that for random sample parents. The intraclass correlation between scores for 144 health states derived from the random-sample-parents and parents-of-patients functions was 0.99; the mean difference per state in scores was 0.018. Conclusion . The HUI2 scoring function generalizes well in that different groups of parents give similar results. The HUI2 scoring function is robust in that the functions without and with exclusions generate scores that are very close in value.