Background A distribution of public health intervention impact across socioeconomic groups can be estimated by examining the cumulative impact of socioeconomic differences across a staircase from need (e.g. prevalence) to intervention characteristics (e.g. effectiveness) using distributional cost effectiveness analysis (DCEA). The extent to which the amount of evidence for inequality at different steps of the staircase contributes to uncertainty in population level impact is not well understood. In this study, we used a DCEA of smoking cessation interventions to explore how socioeconomic inequality in model inputs impacts upon final conclusions about health inequality and value for money.

Methods A smoking cessation DCEA was developed to examine the impacts of interventions on different socioeconomic groups.

Impacts on total population health and health inequality were assessed using incremental population net health benefit (NHB) and incremental ‘equally distributed equivalent’ (EDE) health, both expressed in quality adjusted life years (QALYs). EDE reflects the extent by which the social value of NHB is reduced by inequality in its distribution that favours more advantaged groups.

Scenario analyses were used to explore: (i) the impact of ignoring socioeconomic differences in inputs, e.g., setting mortality in all groups to the average; (ii) the value of eliminating the differences by ‘levelling up’ uptake to the ‘best’; and (iii) how the results differ when applying local level patterns of prevalence (e.g. York). The DCEA was adapted to reflect uncertainty in the extent of the differences between socioeconomic groups. Probabilistic sensitivity analysis (PSA) was used to determine the importance of uncertainty in each input for determining uncertainty in outputs.

Results Using English data, interventions improved NHB (Varenicline: 522,143 QALYs; e-cigarette: 334,874 QALYs) and EDE (Varenicline: 421,457 QALYs; e-cigarette: 270,097 QALYs), but increased health inequality (incremental EDE<incremental NHB). Setting mortality to the average, interventions provided an additional 4% NHB and 2% EDE. Setting uptake to the ‘best’, interventions provided an additional 33% NHB and 56–57% EDE.

Using the data for York, there was uncertainty as to whether interventions reduced health inequality (probability: 16.9% for Varenicline and 22.6% for e-cigarette). The PSA indicated the key drivers for uncertainty were socioeconomic differences in effectiveness, smoking prevalence and uptake.

Conclusion Smoking cessation interventions provide value for money in all the scenarios and interventions to eliminate differences in uptake efficacy could provide additional EDE QALYs. Uncertainty in socioeconomic differences in smoking prevalence contributes the most to uncertainty about the health inequality impact of smoking cessation interventions.