Our analysis, based on a multi-dimensional measure of health-related quality-of-life (HRQoL) applied across all adult age groups, illustrates the dynamic nature of socioeconomic inequalities in health in England. We found that HRQoL declined with increasing age across all five dimensions measured by the EQ-5D-5L. However, there were clear interactions between age and deprivation, with quality of life for people living in deprived areas in England declining substantially faster with age than those living in more advantaged areas. Our analysis shows that inequalities in overall HRQoL follow an approximately U-shape pattern: deprivation-related gaps emerge around the age of 35, reach a peak at around 60 to 64, and then decline again after retirement age. These patterns were consistent when using either the simple absolute or relative difference measures or the more complex concentration index that includes information on all deprivation quintiles. Inequalities in individual dimensions of HRQoL follow a similar pattern, except for mobility, where the gap continues to increase with age. Overall, our results provide further evidence that socioeconomic status is a key predictor of lifetime health in England.

The key strengths of our study are the representativeness of the HSE of the general population in England, the use of an established and widely accepted HRQoL measure such as the EQ-5D-5L, the lower incidence of ceiling effects compared to the three level version of the instrument (EQ-5D-3L) used in previous studies of the English population [15, 16], and the large sample size which permits stratifying HRQoL analyses by age, sex and area-based measures of deprivation.

However, there are also limitations to the study. First, we use the two most recent waves of the HSE, which offer an accurate picture of the current levels of inequality in HRQoL by age and sex but may be subject to cohort effect that limit the ability to extrapolate to past or future cohorts e.g. when calculating quality-adjusted life expectancy [15, 25, 26]. However, a previous study examining changes in EQ-5D-5L responses in England found that scores were stable for most domains and age groups between 2012 and 2017, with the exception of scores for anxiety/depression, which deteriorated for the under 35 s and for women, with the greatest change (leading to a 1.3% decrease in overall HRQoL) in the fifth of women in the most deprived areas [17].

Second, approximately 11% of participants did not report their HRQoL and were excluded from the study. Men and those living in deprived neighbourhoods were more likely to have missing HRQoL information, which may have affected our analysis. We did not impute missing values given that Love-Koh et al. [15] found imputation to have at best a marginal impact on HRQoL scores by deprivation quintile group: mean estimates of HRQoL between naive and imputed datasets differed by less than 0.01.

Third, our finding of inequality in mean EQ-5D-5L index scores (but not in dimension responses) is contingent on the value set chosen to derive these scores. A UK valuation study for the EQ-5D-5L is currently underway and this may affect the results presented here [27].

Fourth, our results need to be interpreted in the context of differential life expectancy across socio-economic deprivation groups, with people in more deprived neighbourhoods in England expected to live significantly shorter lives [28]. Dead people have a defined EQ-5D utility score of zero but are not included in the HSE. As a result, our study measures inequality in HRQoL conditional on being alive.

Finally, it is possible that individuals of different socioeconomic backgrounds report the same level of HRQoL differently on the EQ-5D instrument. For example, individuals may differ in how they interpret limitations of their ‘usual activities’ according to occupation, education or lifestyle [29, 30]. More research is needed to quantify such reporting heterogeneity and adjust for it in the calculation of inequalities in HRQoL.

As expected, we found that HRQoL varies with age, sex and deprivation. With respect to age, mean EQ-5D-5L utility scores start to decline from age 45–49 onwards, and this pattern is also found in three of the HRQoL dimensions—mobility, self-care and usual activities—for which there is generally low prevalence of problems in younger people. There are similar increases for pain/discomfort from middle age, but in this case a significant minority of younger people also report at least slight problems with pain. There is a different pattern for anxiety/depression, with around a quarter of respondents reporting problems in every age group.

Previous survey studies using the EQ-5D have found inequalities in all domains, with the greatest gaps for the pain and anxiety/depression domains [16]. Linked studies have also shown that obesity and chronic conditions, particularly stroke and mental illness, are strong predictors of lower HRQoL scores, but these impacts are mitigated to some extent by higher social status [18]. Shah et al. [17] examining responses to the national GP Patient Survey in England, found lower average overall HRQoL scores for women and a steep deterioration with age. Previous studies have not, however, measured age-related inequalities by domain. In this study, we also found a clear socioeconomic gradient in mean overall EQ-5D-5L scores, with increasing deprivation associated with lower HRQoL, but this gradient only started to emerge in middle age, after which age-related declines in HRQoL were much steeper in more deprived areas.

This interaction between age and deprivation leads to some striking inequalities in HRQoL; for example, in 2017–2018, average HRQoL was lower for 45–49 year old males living in the most deprived fifth of neighbourhoods than for 75–79 year old males living in the least deprived fifth. For four of the five dimensions of the EQ-5D-5L socioeconomic inequalities followed a U-shaped distribution, increasing between ages 45–49 and 60–64 and then decreasing in the older age groups. This was due to declines in HRQoL emerging in younger age groups in more deprived areas, with rates of reported problems eventually converging on similar levels across all quintiles in the oldest age groups. This convergence may in part reflect a healthy survivor effect, which would be consistent with prior evidence on the interaction of age, deprivation and multi-morbidity [7]. Despite inequalities in HRQoL not becoming apparent before middle age, it is likely that they reflect socioeconomic conditions present from childhood and the prenatal period that impact health in ways that may not be detectable by EQ-5D-5L (for example, low birthweight and obesity), but which compromise adult health in the long term [31]. These socially determined shortfalls in health may then be compounded by deprived groups receiving less health care than more advantaged groups, relative to their additional needs [32].

For mobility, inequalities continued to increase with age, with respondents in less deprived areas never reaching the same average levels of problems with walking as those in more deprived areas, even in the oldest age groups. This was particularly apparent in the proportion reporting severe or extreme problems, which ranged from 17% for women and 14% for men in the least deprived fifth of neighbourhoods to 27% for women and 20% for men in the most deprived fifth for the 85+ age group. In addition to differences in prevalence and severity of conditions associated with restricted mobility, this may also reflect differences in daily activities; the social and physical environment; and access to appropriate assistance, mobility aids and means of transport to mitigate the impacts of reduced mobility [33]. Age and deprivation-related patterns in HRQoL, and the interactions between them, were very similar for both sexes, with women tending to report more problems for all ages and all levels of deprivation across all domains.

Our findings highlight the need to use a range of measures in additions to life expectancy and summary measures of morbidity when monitoring health inequalities. Quality-adjusted life expectancy has been proposed [25, 26] as a metric that combines life expectancy with information on HRQoL to capture the cumulative impact of morbidity over individuals’ life spans. However, measuring mean differences in HRQoL across socioeconomic groups conditional on age as done in previous studies [15, 16] may fail to account for differences in inequalities in HRQoL at different points of the life course. We found that inequalities in HRQoL between the most and least deprived socioeconomic groups change with age, following an approximately U-shaped pattern for most dimensions. These results could be used to refine QALE estimates by accounting explicitly for changes in HRQoL inequalities over the life cycle. Future research could also explore inequalities by age for other equity-sensitive characteristics such as ethnicity [34].