Health State Utility Value

A number, typically anchored between zero for dead and one for full health, representing the preference-based value of a specific health state.

Source: Torrance GW. Measurement of health state utilities for economic appraisal. Journal of Health Economics. 1986;5(1):1-30. doi:10.1016/0167-6296(86)90020-2.

Values are obtained by methods that force a trade-off. The time trade-off asks how much time in full health is regarded as equivalent to a longer period in the impaired state, and the ratio of durations gives the value. The standard gamble asks what probability of death a respondent would accept in a treatment offering full health rather than remaining impaired. Both anchor full health at one and death at zero, allowing states judged worse than death to take negative values, and both are usually administered to the general population. Visual analogue scales are not generally accepted as a valuation method, since rating a state on a line requires no sacrifice.

Source: Torrance 1986

Most appraisal frameworks specify values elicited from the general public rather than from patients, on the reasoning that the public funds the service and bears the opportunity cost. Patients consistently rate states they live with less severely than the public imagining them, a pattern attributed partly to adaptation and partly to the difficulty of imagining an unfamiliar condition. Using public values therefore produces larger estimated gains from treatment than patient values would, and the choice is normative rather than technical. Analyses should state whose preferences were used, since the choice affects the estimated benefit of every intervention in the model.

Source: Brazier, Ratcliffe, Salomon & Tsuchiya 2017

Directly eliciting values for every state in a model is rarely feasible, so studies use a standardised descriptive system covering dimensions such as mobility, self-care, usual activities, pain and mood. Patients report their state on that system, and a scoring algorithm derived from a prior population valuation converts the description into a value. This separates description, collected from patients, from valuation, collected once from the public, and it makes routine collection of utility data practicable. It also means the values in a model can be updated when a new population valuation is published, without recollecting from patients, provided the same descriptive system was used.

Source: Brazier, Ratcliffe, Salomon & Tsuchiya 2017

Instruments differ in which dimensions they describe, how many levels each has, and which population valued them by which method. A state described on one system may not be distinguishable on another, so an intervention improving something a system does not capture will show no gain. Values for apparently similar states can differ substantially between instruments, so an intervention's estimated benefit depends partly on the instrument chosen, and analyses using different measures are not directly comparable. The choice of instrument should therefore be made before the study begins and reported prominently, since retrofitting a different measure requires mapping that introduces uncertainty.

Source: Drummond et al. 2015

Descriptive systems are insensitive to changes that matter in some conditions, particularly mental health, sensory impairment and cognitive decline, so interventions there can appear less effective than they are. Values elicited by different methods diverge, and the standard gamble and time trade-off do not agree. Aggregating a multidimensional state into one number discards information about which aspect changed. Values are elicited for states described in isolation rather than in sequence, whereas experience of a condition depends on what preceded it, and for states described as permanent, whereas most conditions fluctuate. Applying a single value to a period during which the patient moved between states misdescribes their experience.

Source: Drummond et al. 2015