Trends in the measurement of health utilities in published cost-utility analyses

The Panel on Cost-Effectiveness in Health and Medicine recommended the compilation of a catalog of health state utility weights for use in cost-utility analyses (CUAs), and has given methodological recommendations. This study presents an update, through 2001, to our current registry of utility weights (available at http://www.tufts-nemc.org/cearegistry; previously at http://www.hsph.harvard.edu/cearegistry), and documents recent changes in methods used for utility weight elicitation.

We searched the English-language medical literature for original CUAs reporting outcomes as cost per quality-adjusted life-year (QALY). Two trained readers independently audited each article, abstracting data on the health state descriptions, corresponding utility weights, methods of elicitation, and sources of the estimates. The utility elicitation methods from 1998 to 2001 were compared with the methods used to obtain utilities before 1998.

We identified 306 CUAs published after 1998, reporting 1210 separate health-related utility estimates, bringing the total in our catalog to 2159 weights. Most frequently, health states pertained to the circulatory system and oncology. Methods varied substantially: 36% of authors used direct elicitation (standard gamble, time trade-off or rating scale), 23% used generic health status instruments (EQ-5D, Health Utilities Index, etc.), and 25% estimated weights based on clinical judgment. Community preferences were used in 27% of the values. Compared with pre-1998, utilities published from 1998 to 2001 were more likely to be elicited using a generic instrument, more likely elicited from community samples, and less likely derived from expert opinion, with no formally employed methodology.

Increasingly, analysts conducting CUAs are using generic, preference-weighted instruments, and relying on community-based preferences. Our catalog of utility weights provides a useful reference tool for producers and consumers of CUAs, but also highlights the continued need for improvement in methods and transparency.