Statement of principal findings

We have shown that despite well-known recent declines in national asthma mortality and hospital admission rates,18 significant differences by SES and region remain. We have also highlighted an unusual and unexpected pattern of a modest but significant decrease in deaths from asthma with increasing levels of relative deprivation in children and adults<45 years.

This was not consistent with the trend in national emergency admissions, where there was a trebling with increasing deprivation, nor in self-reported levels of clinically significant asthma, where there was a doubling in prevalence, in the same age group. For older age groups, in each of these outcomes there remained a clear trend of increasing asthma deaths, admissions and prevalence with greater relative deprivation. The IMD trends in measures of asthma prevalence were attenuated, but remained, after adjustment for smoking.

Relative differences between the lowest and highest regional rates ranged between 1.5 to 1.9 times for the outcomes investigated here. Although the West Midlands still has among the highest rates for both asthma mortality and emergency admissions in England, regional patterns of admissions and prevalence across England do not generally reflect those for mortality.

Strengths and weaknesses

We have used national datasets over a number of overlapping years to assess asthma patterns by SES and region though we were unable to link these to compile a longitudinal dataset. UK deaths statistics have excellent completion rates14 though NRAD’s detailed investigation into cause of death showed a drop in the accuracy of deaths coding with increasing age.2 In those under 45 years, the NRAD Panel agreed that clinical findings matched asthma as the underlying cause in over 90% of cases. The youngest death in that study was aged 4 suggesting that our 5–44 years group has the most accurate asthma mortality coding of all age-groups. This proportion fell to 60–85% in those aged 45–74 years and under 50% in those sampled over 75 years when mis-classification with COPD and other comorbidities increase. However, the percentage of deaths with a mention of asthma which have asthma as the underlying cause has not changed appreciably over time19 and there has to our knowledge been no suggestion of any association of deaths coding with SES or region. Low numbers of asthma deaths in the youngest age group prevented more detailed breakdowns by age or other risk factors to investigate further. HES data have high completeness nationally and a recent review found a median 80% accuracy in diagnostic coding.20 Neither deaths nor admissions data include information on smoking. Prevalence from HSE is self-reported, but this nationally representative survey has good response rates from a large sample and uses validated prevalence questions.17 In contrast to the surveys used in earlier reviews,6 for the HSE 2010 non-response weights were calculated adjusting for age, sex, Strategic Health Authority, household type and the social class of the household reference person.

IMD is a widely used, regularly updated, area-based, multifaceted measure of SES with seven components including health, employment, income and education. It applies beyond working-aged groups, reflects households rather than individuals and may be less prone to reverse-causation than purely income or employment based measures. Although the IMD measures presented are from different years for some outcomes, those for 2004, 2007, 2010 and 2015 employed here use the same methodology and were designed to be broadly comparable in terms of relative deprivation.21 We found 3.6% missing IMD quintile in the mortality data, but there was little difference by age-group (3.8% 5–44y, 4.0% 45–74y and 3.4% 75+). We have no reason to believe IMD coding is less accurate in younger age groups though deprivation measures tend to improve with age. This does not explain however the inverse relationship seen for mortality compared with that for admissions or prevalence.

Interpretation in the light of previous research

Previous UK studies used social class (SC I to V) as a measure of SES.4 5 8 9 This was derived from the occupation of the individual or head of household but could not be applied to those over working age. Higgins et al 4 found a similar pattern to this investigation: little difference in mortality between SC in males aged 5–34 years (England and Wales 1979–87), but in men aged 35–64 years there were 36% more deaths in manual occupations compared with non-manual. Two-thirds of records for women were missing individual SC and in those remaining there was no clear trend. Jones et al 8 considered a broader age range and found that in those aged 5–55 years (from deaths in 401 local authority districts in England in 1988–92) asthma mortality was strongly associated with the proportion of households where the head was SC most deprived (RR=1.61, 95% CI 1.12 to 2.33). National mortality records (all-ages SERs, England and Wales 1991–93) showed a four-fold increase in asthma deaths between those in the most and least deprived SC.9

The more recent mortality rates among older adults presented here show higher rates in more deprived areas, but the observation of a significant trend in the opposite direction among children and younger adults may reveal an unexpected phenomenon. Among 5–44 year-olds, the prevalence of diagnosed-and-treated asthma varied little with area deprivation in contrast to a doubling in prevalence of severe symptoms (table 1). While this might suggest small differences in asthma labelling by SES (even after adjustment for smoking), most likely to be made by primary care physicians, the mortality results for this age-group are unlikely to be entirely attributable to diagnostic or labelling artefact. Since severe asthma symptoms and hospital admissions for asthma in the under-45s were less common in the more affluent areas it is unlikely that the higher mortality rate in this group is due to poorer long-term control of asthma symptoms. Indeed, one would expect that compliance with prescribed treatment would be better in these areas than in more deprived areas. We were unable to assess trends by SES in general practice data, but two studies which have done this showed higher prevalence and incidence of asthma consultations in more deprived areas,5 a socioeconomic trend which has been found even among non-smokers.9 SES measures from routine data are difficult to harmonise internationally so comparable SES measures between the UK and elsewhere are rare and we are not aware of any for asthma risk factors. Nor are we aware of any analyses of SES time-trends for asthma, again probably due to the difficulties in comparisons over time, though our own analysis shows consistently higher mortality rates in the most deprived quintile compared with the least deprived quintile (online supplementary figure 1).

Significant differences in asthma mortality, admissions and prevalence by English region have been noted in the past.4 6 The weak correlations found between regional all-ages asthma prevalence, admissions and mortality in the early 1990s6 match our results. This is perhaps unsurprising as the highest admission rates are in the young whereas the highest mortality rates occur in the elderly with prevalence more stable over the life course. We found that large regional differences persist despite the introduction, in the late 1990s, and the development, over several decades, of national treatment guidelines by the National Institute for Health and Care Excellence and BTS/SIGN as well as several years’ of incentive payments for asthma treatment by the Quality and Outcomes Framework programme. This concurs with GP reports of people ‘ever diagnosed’ with asthma showing wide variations across England in 2012.5 An analysis of 24 European countries found a very similar correlation (r=0.44, P=0.03) between them in age-standardised mortality and admissions for asthma.1 A comparison of mortality, admissions and prevalence of wheeze in children22 similarly found a high correlation between severe wheeze and admissions in 13–14 year olds (r=0.73, P=0.003) in 14 countries though little correlation between other outcomes.

Regional variations in asthma mortality in the UK have been shown to reflect geographical differences in access to and quality of healthcare services6 as well as socioeconomic factors.8 Information on more specific exposures such as smoking and ethnicity was not available in the national mortality and admissions data, but using the prevalence data we were able to confirm the observation of Hansell et al 6 8 that regional differences in asthma prevalence and severity were not altered greatly by adjustment for smoking. An analysis of deaths from all causes in England23 found that IMD accounted for a large proportion of the geographical variation in age-adjusted mortality and the independent contribution of regional differences was small in both sexes.

Interpretation and future research

Stringini et al 7 have estimated the rate-ratio of low compared with high SES on non-cardiovascular disease, non-cancer mortality to be 1.45 (95% CI 1.35 to 1.56) after adjustment for age, sex and ethnicity and 1.25 (95% CI 1.17 to 1.33) when further adjusted for high alcohol intake, physical inactivity, current smoking, hypertension, diabetes and obesity. They acknowledge that each of these may be interrelated, together with specific components of SES including low income, education levels and occupational position. Rona3 suggested the following possible reasons for SES as a risk factor in asthma: an aetiological factor; an exacerbating factor; a determinant of the quality of care; a contributor of psychosocial behaviour which impacts on the management and prognosis of the condition or as a component in asthma labelling. More specifically, the Global Atlas of Asthma24 suggests that low SES families are more likely to be exposed to environmental pollutants (eg, particles from diesel), indoor allergens (eg, mould and dust) and other respiratory irritants (eg, tobacco smoke) that adversely affect asthma as well as being more likely to be exposed to psychosocial stressors (eg, food, housing and financial insecurity, social marginalisation and community or domestic violence). NRAD2 found psychosocial factors contributed to 26% (51/195) of UK asthma deaths of all ages, with similar proportions among deaths under 10 (30%) and aged 10–19 (22%). ‘Psychological factors’ contributed to 16% of deaths and ‘social factors’ to 15% of deaths. There are well-recognised occupational causes of asthma associated with lower SES jobs (eg, bakers, farmers, painters, welders, woodworkers). These, as with most factors in the literature, help explain the increased asthma in lower SES groups seen in most outcomes and age-groups here and elsewhere.

There are few examples however of risk factors of any type for higher SES groups that might account for our mortality findings in the 5–44 year olds. The Global Atlas of Asthma suggests ‘community vitality/collective efficacy, increased maternal-child interaction, and effective utilisation of psychological coping strategies’ have positive effects on asthma outcomes and it could be that some of these form mitigating factors in lower SES groups. There are also a few occupational causes of asthma associated with higher SES (eg, latex in healthcare workers, aldehydes in dentists, animal excreta in laboratory researchers). It is not clear how much each of these might contribute to the inverse pattern seen here in younger deaths or if other factors might account for it. Higgins and Britton were not able to provide an explanation found for the similar pattern found in males in earlier decades.4 A study of mortality in England in the late 1990s25 found slightly higher mortality rates from ‘major causes’ in teenage boys in the least deprived quintile of IMD compared with the most deprived and almost equal rates in teenage girls. This category included lower respiratory diseases, cancers, circulatory disease and others (ICD9 1–239 390–519) but no reasons for this pattern were offered nor were more detailed causes investigated. A study comparing mortality with a census question on long-term illness or disability26 found SES all-cause mortality differentials were lowest in the teenage years and early 20s, but also found a similar pattern for self-reported ‘limiting long-term illness’, in contrast to our findings here for asthma. This suggests a disease-specific component to the difference between mortality and morbidity SES patterns for asthma.

One speculative explanation for the anomalous socioeconomic trend in younger asthma deaths is that in more affluent areas there is a higher prevalence of a less predictable (more ‘brittle’) form of asthma which occasionally progresses rapidly to a fatal outcome despite apparently adequate long-term symptomatic control. The higher prevalence of atopy and allergies in higher SES groups may also be a factor acting together with brittle asthma or in addition to it and it is possible that some of these cases are included within deaths coded as asthma.

Asthma poses a large healthcare burden in England costing the public sector an estimated £900 million12 and many deaths are avoidable with better care.2 Despite falling asthma mortality over time and decades of improvements in asthma management, the UK still suffers among the highest global mortality rates in the young1 and significant differences by SES and region remain across England, despite national guidelines and a health service free at the point of contact. These differences are even wider in emergency hospital admissions where there were more than three times as many emergency admissions in the poorest among the youngest and twice as many admissions and cases of severe symptoms among the poorest in older age groups.

These results should be monitored in England and, where possible, tested in other countries using existing international health studies. Greater numbers of deaths from longer English or UK time trends could be sought to investigate age-specific mortality-SES differentials within the 5–44 year old age-band. The relationships between prevalence of symptoms, consultations in general practice, emergency admission to hospital and mortality should be examined further and the reason for the inverse relation of mortality with IMD in those aged 5–44 years in relation to other age groups and other outcomes needs explanation as this is unlikely to be due to data artefact. Such analyses could use general practice databases linked to admissions and mortality which would enable more complete allowance than possible here, for confounding by smoking and other risk factors. More detailed clinical information could identify a distinct and troublesome asthma phenotype which has become more common in England in recent years, particularly in more affluent areas, as well as helping identify the role of concurrent allergic disease or any overlap with COPD in older age groups.