New Tool Reveals Who Really Benefits in Randomised Trials, Beyond the Average

For decades, the randomised controlled trial has been the gold standard for deciding whether a medical intervention works. Its central logic is elegant: randomly assign people to a treatment or a control group, and any difference in outcomes between the groups can be attributed to the treatment itself. Yet a new study published in the International Journal for Equity in Health argues that this celebrated design has been answering only half the question. Led by Zachary D. V. Abel of the University of Oxford, together with Sophie Cole, Guido Erreygers, Laurence S. J. Roope and Philip M. Clarke, the research lays out a systematic method for measuring not just whether an intervention lifts the average, but who captures those gains — and whether the poorest in society are left behind.

The problem the researchers identify is striking in its scale. When a sample of 200 equity-relevant randomised trials was examined, only 4 percent reported any socioeconomic subgroup analysis. A separate systematic review of infectious disease trials found that just 9 percent reported participants’ socioeconomic status at all. This is a curious blind spot, the authors note, given that equity and equality are widely held principles of health systems and are enshrined in the third United Nations Sustainable Development Goal: to ensure healthy lives and promote well-being for all. If a health system genuinely aspires to equal outcomes, then knowing how a policy shifts the gap between rich and poor is not a luxury — it is a minimum requirement.

The technical heart of the proposal is the concentration index, a workhorse of health economics first developed by Kakwani and later introduced into health research by Wagstaff and colleagues. The index ranks every individual in a sample from poorest to richest and summarises how a health variable — vaccination status, disease burden, life expectancy — is distributed across that ranking. A value of zero means health is spread evenly across the socioeconomic spectrum; a positive value signals that health is concentrated among the better-off, while a negative value indicates the opposite. Extensions such as the generalised concentration index and the Erreygers adjusted index, which is suited to binary outcomes like vaccination and bounded between minus one and one, allow researchers to capture absolute rather than purely relative inequality.

What makes the new framework powerful is a simple statistical insight: because randomisation balances the socioeconomic composition of participants across trial arms at baseline, any difference in the concentration index between the treatment and control groups at follow-up can be interpreted as a causal effect of the intervention on income-related inequality. In other words, the same trial that estimates an average treatment effect can, at essentially no extra cost, estimate a distributional treatment effect. To the authors’ knowledge, only one previous trial — a study of an intervention for low birthweight babies in India — has ever reported concentration indices as an equity outcome, suggesting how rarely this opportunity has been seized.

To visualise both dimensions at once, the team revives the achievement index, originally developed by Wagstaff, which multiplies mean health by one minus the concentration index, and the achievement plane later introduced by Clarke and Hayes. In the trial setting, the average treatment effect is plotted on the horizontal axis and the change in absolute inequality on the vertical axis, with the placebo group anchored at the origin. Interventions landing in the upper-right quadrant deliver both higher average health and reduced pro-rich inequality — a win-win. Those in the lower-left are dominated on both counts. The off-diagonal quadrants expose the equality-efficiency trade-offs that policymakers routinely face but rarely see quantified: an intervention may raise the societal average precisely because it disproportionately benefits the wealthy, quietly widening the very gaps health systems claim to close.

As a proof of concept, the researchers turned to publicly available data from the Ghana Financial Incentives Trial, a cluster-randomised study conducted across six rural Ghanaian districts during the COVID-19 pandemic. Villages were assigned to one of four arms: a placebo video message, a CDC-styled informative health message, a low cash incentive of 20 Ghana cedis (about 3 US dollars), or a high cash incentive of 60 cedis (about 10 dollars). The analysis presented in the new paper focuses on the 2,271 participants whose vaccination status was verified at district hospitals between October and November 2022. Socioeconomic status was measured using equivalised weekly household food expenditure, adjusted with the OECD modified scale for household composition.

The results are instructive on two fronts. On the traditional measure of efficiency, the low cash incentive was the only intervention that significantly increased verified vaccination uptake relative to placebo. On the distributional front, health messaging produced a pro-rich inequality impact of minus 0.132 (95 percent confidence interval minus 0.28 to 0.02), while the low and high cash incentives yielded impacts of minus 0.108 (minus 0.29 to 0.07) and minus 0.081 (minus 0.26 to 0.11) respectively — none reaching statistical significance. In this framework, negative values indicate a shift toward pro-rich inequality, so the point estimates hint that every intervention may have nudged uptake slightly toward the wealthier end of the spectrum, but the evidence is far from conclusive.

That inconclusiveness is itself a lesson the authors want the field to absorb. Because the original trial was not designed or powered to detect distributional effects, the team computed the minimum detectable effect for each comparison, finding that only differences greater than roughly 0.17 in the concentration index could be ruled out. Smaller but potentially policy-relevant distributional shifts may exist undetected. The researchers stress that non-significant differences should never be equated with equivalence, and they recommend that future trials pre-specify equity outcomes and calculate sample sizes that account for the statistical power needed to detect distributional effects — treating inequality with the same rigour currently reserved for average treatment effects.

The framework also outperforms the subgroup analyses that occasionally appear in trial reports. Traditional heterogeneity testing compares average effects across broad, discretely defined groups, such as the poorest versus the wealthiest quintile, discarding most of the information in the continuous socioeconomic distribution. The concentration index, by contrast, uses the full ranked spectrum of socioeconomic status, weighting outcomes by relative position and capturing the entire gradient of inequality. Placed on the achievement plane, this makes equity a co-equal dimension of trial evaluation rather than a secondary afterthought, and allows competing interventions to be compared simultaneously on both efficiency and distribution.

The implications reach well beyond vaccination campaigns. The authors outline how the approach could integrate with distributional cost-effectiveness analysis, which weights health benefits by recipients’ socioeconomic position, and how the achievement index could serve as a way to adjust outcomes for distributional effects within that framework. They also sketch extensions to quasi-experimental designs such as difference-in-differences, regression discontinuity and instrumental variable studies, where assumptions analogous to randomisation could support causal claims about distributional impacts. For now, the Ghana example makes the core message vivid: during a pandemic in which equitable vaccine distribution was a global concern, the trial data could not rule out that incentives and messaging subtly favoured the better-off. Routine reporting of distributional treatment effects would ensure that such questions are answered by design rather than by accident — and that equality-blind decisions, which risk widening health inequalities while celebrating rising averages, become a thing of the past.

Subject of Research: A method for measuring socioeconomic health inequality effects within randomised controlled trials

Article Title: Moving beyond the average: a method to measure health related-inequalities within randomised trials

Article References: Moving beyond the average: a method to measure health related-inequalities within randomised trials. (n.d.). https://doi.org/10.1186/s12939-026-02977-x

Image Credits: AI Generated

DOI: 10.1186/s12939-026-02977-x

Keywords: randomised controlled trials, health inequality, concentration index, achievement plane, COVID-19 vaccination, Ghana, health economics, distributional treatment effects, equity, cash incentives, average treatment effect, distributional cost-effectiveness analysis