Weighted two-group social welfare and the switching weight

Scores each option as total QALYs of group A plus a weight w times total QALYs of group B, where w is the value placed on a QALY accruing to the worse-off group relative to group A. A weight of 1 gives the utilitarian sum. The switching weight is the value of w at which two options X and Y score the same, found by setting their weighted totals equal; above it the option that does more for group B is preferred.

Signature

W_X = QA_X + w * QB_X; W_Y = QA_Y + w * QB_Y; w_s = (QA_X - QA_Y) / (QB_Y - QB_X)
Inputs
InputsDefinitionUnit
QA_XTotal QALYs of group A under XQALYs
wRelative value of a QALY accruing to group B, 1 for the utilitarian sumdimensionless
QB_XTotal QALYs of group B under XQALYs
QA_YTotal QALYs of group A under YQALYs
QB_YTotal QALYs of group B under Y, different from QB_XQALYs
Output
W_XWeighted total QALYs under option Xweighted QALYs
W_YWeighted total QALYs under option Yweighted QALYs
w_sWeight on group B at which X and Y are ranked equallydimensionless

Function

Welfare change measurement and aggregation function

Converts each person's gain or loss from a change into money, with the compensating or equivalent variation or the consumer surplus approximation, and then ranks the change for society: by adding the money measures, as the Kaldor-Hicks potential compensation test does, or by combining individual outcomes with an explicit social welfare function that states how gains to one person are set against losses to another.

Try this function

Implementations

  • Excel

    Weighted welfare scores and switching weight

    With the QALY totals in QAlyAX, QAlyBX, QAlyAY and QAlyBY and the weight in WeightB, the formulas return the two weighted scores and the switching weight.

    =QAlyAX+WeightB*QAlyBX; =QAlyAY+WeightB*QAlyBY; =(QAlyAX-QAlyAY)/(QAlyBY-QAlyBX)

Assumptions

  • Linear weighting within each group

    Every QALY in a group receives the same weight, so the function is linear. An inequality-averse function that is concave in each person's health gives weights that change with the level of health.

  • Weights are value judgements

    The weight expresses a distributional judgement, even when elicited from public preference surveys. Reporting the switching weight lets decision-makers see how strong that judgement must be to change the ranking.

Worked examples

  • Reconfiguration against outreach with a weight of 2

    In the article's example, option X gives groups A and B 20,030 and 11,990 QALYs and option Y 20,000 and 12,015. With a weight of 2 on group B, X scores 44,010 and Y 44,030, so Y is preferred; the switching weight is 30 divided by 25, or 1.2.

    QA_X = 20030; QB_X = 11990; QA_Y = 20000; QB_Y = 12015; w = 2; W_X = 44010; W_Y = 44030; w_s = 1.2
  • Same options under the utilitarian weight

    With a weight of 1 the function is the utilitarian sum: X scores 32,020 and Y 32,015, so X is preferred, as in the cost-benefit result.

    QA_X = 20030; QB_X = 11990; QA_Y = 20000; QB_Y = 12015; w = 1; W_X = 32020; W_Y = 32015; w_s = 1.2

Common errors

  • Reading a switching weight below 1 as favouring the better-off group

    The switching weight only says where the ranking changes. Its direction depends on which option does more for group B; here Y does, so every weight above 1.2 favours Y and every weight below it favours X.

  • Applying group totals when group sizes differ

    Comparing group totals ranks options in the same way as QALYs per person only when the groups are the same size and everyone within a group is affected alike. Otherwise a Rawlsian or per-person comparison needs per-person values.

Sources

  • Social welfare functions over the distribution of QALYs

    Wagstaff A. QALYs and the equity-efficiency trade-off. Journal of Health Economics. 1991;10(1):21-41. A social welfare function defined over individuals' health that weights QALYs by whom they accrue to, making the trade-off between total QALYs and their distribution explicit.

    View source →

Canonical Identity

Stable URI · Machine-readable · Resolvable · CC BY 4.0