Equity-Weighted Analysis

An economic evaluation that applies differential weights to health gains in different population groups, giving greater weight to gains for disadvantaged groups.

Source: Cookson et al. 2021

Health gains are multiplied by weights that differ according to who receives them, so a quality-adjusted life year accruing to a disadvantaged group counts for more than one accruing to an advantaged group. The weighted totals are then compared across options in the usual way. The weights make explicit a judgement that conventional analysis makes implicitly, since treating every unit of health as equal is itself a distributive position rather than a neutral one. The approach therefore does not introduce a value judgement into an otherwise neutral calculation; it replaces one distributive assumption with another and states it.

Source: Cookson et al. 2021

There is no agreed source, which is the central difficulty. Weights can be derived from surveys asking the public how much aggregate health they would sacrifice for a more equal distribution, from the implied weights in past decisions, or specified by the decision maker as a policy parameter. Because none of these produces a settled figure, analyses normally report results across a range of weights rather than adopting one, so the reader can see at what point the conclusion changes. Presenting the threshold weight at which the preferred option changes is frequently more useful than presenting a result at any single weight.

Source: Cookson et al. 2021

Distributional analysis reports how gains and losses fall across groups and leaves the weighing to the decision maker. Equity weighting applies the weights within the analysis and returns a single adjusted result. The first preserves transparency and requires judgement afterwards; the second embeds the judgement and produces a comparable figure. They are frequently used together, with the distribution reported and the weighted result shown as a scenario. In practice the two are complementary, and reporting the underlying distribution alongside any weighted figure preserves the information that weighting compresses. Both approaches depend on the same underlying subgroup evidence, so neither is available where effects have not been estimated by group.

Source: Cookson et al. 2021

It requires effects to be estimated separately by group, including differences in baseline risk, uptake, adherence and the treatment effect itself, which trials rarely report by socioeconomic position. It also requires an assumption about where the displaced spending would have fallen, since the health forgone is distributed as well as the health gained. Both are frequently assumed rather than observed, and the assumptions can drive the equity conclusion more than the evidence does. Where subgroup effects are assumed rather than measured, the analysis should say so plainly, since a weighted result built on assumed differences is a scenario rather than an estimate.

Source: Drummond et al. 2015

That the weights are arbitrary in the absence of an agreed basis, so the analysis can produce whichever answer the chosen weight implies. That weighting by group membership treats individuals as representatives of categories. That it addresses one dimension of disadvantage while others go unweighted. And that combining efficiency and equity into a single number removes the trade-off from view, which is the opposite of what an explicit treatment of equity was meant to achieve. Reporting results across a range of weights rather than adopting one answers most of these objections, at the cost of not producing the single figure that made the approach attractive.

Source: Neumann, Sanders et al. 2016