Background Area-based inequalities in mortality are well known in UK and elsewhere, with poorer areas having higher mortality than more affluent areas. Despite area level measures of deprivation being used extensively in research and also to inform policy, they are not without problems. The modifiable areal unit problem occurs when data are combined into units for analysis and this can result in biassed results. A further issue is area sizes can differ in population. Both phenomena have implications for the size of inequality between most deprived and least deprived areas. This analysis examines inequalities in mortality by area deprivation using three administrative geographic levels with different population sizes.

Methods Carstairs scores are a measure of material disadvantage derived from four variables from the 2011 Census in Scotland. Carstairs was calculated for postcode sectors (PS: average population 5,233); data zones (DZ: average population 815); and Census output areas (OA: average population 114). Mortality data for 2010–2012 for Scotland were combined with population data from the 2011 Census for men and women aged 0–64 years. Poisson regression was used to assess age adjusted inequalities in all cause mortality and mortality due to ischaemic heart disease (IHD) and cancer.

Results There were 44,048 deaths at ages 0–64 between 2010–2012 in a population of 4,666,267. Rate ratios (RR) tended to increase as the geographic areas became smaller. For all-cause mortality the RR for men for PS was 3.08 [95% confidence interval (2.95–3.22)]; DZ 4.34 (4.15–4.53); and OA 5.01 (4.19–5.24). For women the corresponding RR (95% CI) were PS 2.51 (2.38–2.64); DZ 3.39 (3.21–3.57); OA 3.77 (3.57–3.98). For IHD mortality the RR for men for PS was 3.59 (3.21–4.00); DZ 5.34 (4.78–5.97); and OA 6.08 (5.43–6.80). The corresponding RR for women: PS 3.99 (3.29–4.84); DZ 6.72 (5.50–8.20); and OA 7.46 (6.09–9.13). For cancer mortality the RR for men for PS was 2.08 (1.93–2.25); DZ 2.57 (2.38–2.77); and OA 2.83 (2.62–3.06). The corresponding RR for women: PS 1.74 (1.61–1.88); DZ 2.03 (1.88–2.20); and OA 2.20 (2.03–2.38).

Conclusion Inequalities increase for all-cause mortality and cause-specific mortality as geographies get smaller. One possible reason is pockets of deprivation are easier to identify when using smaller areas – small areas tend to be more internally homogenous than large areas. Alternatively, it could be individual deprivation driving the relationship with health and the smaller the area, the closer deprivation captures individual status. Focusing interventions on the smallest geographic areas may help to reduce widening health inequalities.