Background: Evaluations of public health interventions typically report benefits and harms aggregated over the population. However, benefits and harms are not always evenly distributed. Examining disaggregated outcomes enables decision makers to consider health benefits and harms accruing to both intended intervention recipients and others in the population.
Methods: We provide a graphical framework for categorizing and comparing public health interventions that examines the distribution of benefit and harm between and within population subgroups for a single intervention and compares distributions of harm and benefit for multiple interventions. We demonstrate the framework through a case study of a hypothetical increase in the price of meat (5%, 10%, 25%, or 50%) that, via elasticity of demand, reduces consumption and consequently reduces body mass index. We examine how inequalities in benefits and harms (measured by quality-adjusted life-years) are distributed across a population of white and black males and females.
Results: A 50% meat price increase would yield the greatest net benefit to the population. However, because of reduced consumption among low-weight individuals, black males would bear disproportionate harm relative to the benefit they receive. With increasing meat price, the distribution of harm relative to benefit becomes less "internal" to those receiving benefit and more "distributed" to those not receiving commensurate benefit. When we segment the population by sex only, this result does not hold.
Conclusions: Disaggregating harms and benefits to understand their differential impact on subgroups can strongly affect which decision alternative is deemed optimal, as can the approach to segmenting the population. Our framework provides a useful tool for illuminating key tradeoffs relevant to harm-averse decision makers and those concerned with both equity and efficiency.
Keywords: distributional effects; harm aversion; health technology assessment; multicriteria decision analysis.