Ethnic inequalities in health have been well documented in the UK.1^–5 Since the Black Report,6 there has been continued focus upon the influence of socioeconomic factors in explaining how those at the lower end of the social hierarchy suffer a poorer health profile than those less disadvantaged. Emphasis has also been placed upon understanding the importance of biological, cultural or behaviour risk factors to health, which may also be differentially distributed between and within ethnic minority groups.7

However, such investigations in the UK have been largely based upon data from migrants, either because they have been specifically focused on migrants or because the older population, where morbidity and mortality become prevalent, is almost exclusively composed of migrants. Immigrant mortality studies in the UK3 and elsewhere8 9 suggest that the health profiles of these individuals have been shaped by a healthy migrant effect, whereby those who migrate are atypically healthier and less likely to be disadvantaged than those individuals who remained in the country of origin. Importantly, a link has been established between increasing duration of residence and declining health in a range of UK ethnic minorities.10^–12 This suggests that the positive health characteristics of the first generation are not retained indefinitely and that some loss of the healthy profile is inevitable over time. Despite international studies identifying migrant population health as approximating to that of the host population within one or two generations,13 14 there are few studies documenting intergenerational changes in health status in the UK, or the causal pathways by which change is mediated, largely because of the young age profile of the second and third generations.

There is substantial evidence of differential rates of intergenerational upward social mobility in the main UK ethnic minority groups.15 16 These differences in socioeconomic success reflect the different capabilities and circumstances of each group, leading to differences in the reduction in the exposure to risk factors to health over generations. There is also evidence for intergenerational modification of culturally specific behaviour and social norms.17 Acculturative changes in health behaviours over time might also influence, and possibly diminish, the health advantage of the migrant generation. Common examples of this acculturative phenomenon are the uptake of less healthy behaviours such as increased smoking rates, declining standards of diet18^–20 and a worsening in maternal behaviours.21 It would seem likely, then, that exposure to both socioeconomic and behavioural risk factors to health are fixed neither across generations nor over time, and that associated health outcomes might be expected to vary accordingly.

Previous cross-sectional analyses have shown consistently higher rates of reported fair/poor self-rated general health across a range of UK ethnic minorities when compared with the white majority population.1 4 5 This study takes advantage of the large sample sizes in the ethnically boosted years of the Health Survey for England (1999 and 2004) to investigate the extent of intergenerational change in ethnic inequalities in self-rated general health and health behaviours, how far this varies between groups, and to identify those factors underlying any observed changes.

METHODS

Health survey for England

The Health Survey for England (HSE) used a multistage stratified probability sample design to select a nationally representative sample of the general population. The 1999 and 2004 surveys oversampled ethnic minority households and, when these surveys are combined, the sample is large enough to allow for the study of generation effects. Comparative data for a white English population are drawn from the 1998 and 2003 HSE surveys.

Ethnicity and generation

Ethnicity was recorded according to informant self-reports of their family origins. The groups included in the analysis were black Caribbean, Indian, Pakistani, Bangladeshi, Chinese, Irish and white English. First-generation ethnic minorities were defined as foreign born and migrating to the UK aged 12 years or older. The second generation were classed as UK born, or foreign born and migrating to the UK when under 12 years of age. The cut-off age of 12 years is used because it correlates with subsequent exposure to a significant period of childhood and secondary schooling in the UK, which will be important influences for both social mobility and changes in behaviours. This cut-off provides as numerically balanced a sample as possible and has been used in previous large-scale cross-sectional studies differentiating between generations.22^–25 Similar results are obtained when country of birth is used as the cut-off.

As the different generations had only partially overlapping age structures with each other, age censoring was required. This was carried out to ensure that at least 20 individuals from each generation were contained in 5-year age bands. Consequently, we only included in the analysis: black Caribbean aged 21–55 years, Indian aged 21–50 years, Pakistani aged 16–50 years, Bangladeshi aged 16–45 years, Chinese aged 16–45 years and Irish aged 21–85 years (this wider age distribution reflecting their longer migratory history). The white English reference population covered ages 16–55 years. The Irish age distribution was further censored to ages 16–55 years when modelled for a direct comparison with the white reference group.

Health outcome

Informants were asked to rate their health according to a five-point scale: very good, good, fair, poor and very poor. This was coded to a binary variable: fair/poor/very poor and good/very good. This dichotomy has been shown to capture the ordered nature of self-rated health26 and has been used previously in HSE analyses1 5 and in the investigation of the Fourth National Survey of Ethnic Minorities in Britain.4 This measure is an important predictor of all-cause mortality,27 28 validated in different ethnic groups,29 30 so is a relationship that seems to be universal rather than ethnically specific.31 32

Explanatory variables

Socioeconomic factors

A single socioeconomic indicator inadequately reflects the social position occupied by individuals of each ethnic minority group.4 33 For example, there is much variation in the income levels among occupations that make up each occupational class. Consequently, three indicators were used to more accurately assess the social circumstances of each group. The Registrar General’s social class system classified informant occupations by social position. Household income, equivalised to account for the number of individuals in a household, was coded into quintiles to account for the highly skewed income distribution. Educational level was taken as the top qualification achieved by the informant. This comprised seven categories ranging from no qualification to NVQ5 or degree, and is generally regarded as an accurate representation of the level of skills available to the labour market. The small numbers of people with foreign qualifications were treated as having no qualification.

Health-related behaviours

Measures covered smoking, alcohol consumption and diet. For modelling purposes, behaviours were coded to binary variables. Smokers were classified as those who were current smokers as opposed to current non-smokers. Patterns of alcoholic intake were identified as any regular drinking vs abstention from alcohol. Diet was assessed in terms of fruit and vegetable consumption, contrasting those consuming less than one portion of fruit or vegetables in a week with others.

Analysis

Stepwise construction of binary logistic regression models began with the calculation of an unadjusted odds ratio (OR) for the poor/fair health in the second generation relative to the first generation. These odds were subsequently adjusted to account for differences in the age and sex distributions between generations. The effects of socioeconomic status and health behaviours were then examined separately, and simultaneously, to estimate the extent to which each factor explained intergenerational variation in fair/poor general health.

The final models examined whether the odds of fair/poor health for ethnic minority groups were converging to, or diverging from, the rates of fair/poor health in the white reference population. All analyses were performed using STATA™ version 9.2 (StataCorp, College Station, TX, USA), which allowed us to account for sample weights (used to account for different known probabilities of selection into the sample) and the stratification and clustering of the sample.

RESULTS

Table 1 shows the demographic and socioeconomic characteristics of each generation across the range of ethnic groups. All ethnic minority groups tended to show upward intergenerational social mobility as measured through educational attainment, social class and equivalised income, although the extent of this mobility varied between groups. The Indian group demonstrated the greatest mobility, with around 45% of the second generation in the upper professional classes compared with 35% in the first generation, and similar differences were observed in levels of educational attainment and income. These intergenerational changes contrasted sharply with the Bangladeshi group, where half of the first generation sample occupied the lowest income quintile, decreasing to 43% into the second generation. The Bangladeshi group showed marked improvements in educational attainment, however, with 64% of the first generation having no qualification declining sharply to 21% in the second generation. As a consequence of a low starting position and weak upward socioeconomic mobility, the second generation of both the Bangladeshi and the Pakistani groups remained markedly socioeconomically disadvantaged relative to the white population.

Table 2 shows that there was a general trend for the health behaviours of second- compared with first-generation ethnic minority people to approximate towards those observed in the majority population. A statistically significant increase in the rate of smoking across generations was observed for the black Caribbean group. In the Indian and Pakistani groups, smoking rates were also higher in the second generation, although changes were not statistically significant. These changes represented a convergence towards the smoking rates observed in the majority white population. Most groups showed a slight intergenerational worsening of the diet, reaching statistical significance for the Indian group for low vegetable and fruit consumption. Patterns of daily alcohol intake varied widely by ethnic group, with the vast majority of those in the Pakistani and Bangladeshi groups abstaining. Nevertheless, with the exception of the Irish, all ethnic minority groups showed a statistically significant increased likelihood of drinking in the second generation, with movement towards the prevalence of drinking in the white majority population.

The odds ratios for fair or poor general health adjusted for age and sex, and then a range of socioeconomic and health behaviour variables are shown in table 3. There appeared to be little difference between the generations after accounting for age and sex differences. Adjustment for socioeconomic factors resulted in a slight increase in the odds of reporting fair or poor health in the second generation for all but the Bangladeshi and Irish groups. This increase was statistically significant for both Pakistani and Chinese groups. In contrast, adjustment for health-related behaviours had little effect on the models. So, in the model that adjusted for both health-related behaviour and socioeconomic position, the odds of fair/poor health in the second compared with the first generation approximated to that where only socioeconomic position was adjusted for. Results suggest that, once improved socioeconomic circumstances in the second generation were accounted for, fair/poor health was more likely in the second generation, with the exception of the Bangladeshi and Irish groups. It would appear that the overall tendency to show comparable health in the second generation was a consequence of upward intergenerational social mobility.

Table 4 shows age- and sex-adjusted ORs for having fair/poor health compared with white English people, separately for first- and second-generation people in each ethnic minority group. With the exception of the Bangladeshi group, there were no differences between generations. The Bangladeshi group showed a statistically significant convergence from first to second generation towards the rates of fair/poor general health in the white English population. This finding is consistent with earlier models (table 3), which also observed an improvement in levels of general health between the Bangladeshi generations. It was notable that the Bangladeshi first generation reported considerably poorer general health than all other groups, providing scope for the marked improvement seen in the second generation. There was a consistently elevated rate of fair/poor health within the black Caribbean, Pakistani, Indian, as well as the Bangladeshi group, whereas the rates for both Chinese and Irish groups were more similar to the rates observed in the white population.

DISCUSSION

This study examined the extent to which intergenerational modifications in socioeconomic circumstances and health behaviours are associated with generational shifts in general health. The well-established association between higher social position and better health suggested that the intergenerational social mobility observed here, and elsewhere,16 34 35 would consequently lead to improved second-generation general health. This was not the case, however. In fact, once the upward mobility of the second generation was accounted for, their health appeared to be worse than that of the first generation.

A possible underlying explanation concerns the process of health selection. The healthy migrant effect posits that the first generation have a better health profile than their social position might otherwise suggest. Furthermore, migrants are liable to face a period of immediate downward mobility post migration, undertaking social roles that were not befitting of their social status before migration.15 Such downward mobility exacerbates the mismatch between relatively high health status and low social position post migration. This phenomenon may account for the finding that the second generation appears to require greater levels of socioeconomic advantage to achieve the same level of self-reported health as the first generation.

While socioeconomic disadvantage is likely to contribute to patterns of health inequality, other risk factors are clearly involved. These risks are situated within the differing life courses of each generation. The second generation is exposed to different sets of exposures in early life compared with the first generation, having spent critical periods of childhood and development within the UK environment. For example, postnatal growth has been shown to affect later life disease risk.36 As the second generation ages, it is likely that differential exposure to psychosocial risks in the workplace will also be encountered, as the second-generation populations establish themselves within areas of the labour market unoccupied by the first generation.34 These risks may be experienced alongside new levels and forms of discrimination, with the second generation exposed to less extreme forms of discrimination and racism, but liable to a greater sense of the injustice caused by the gap between expectations of economic or legal equality and the realities of exclusion.37 Barriers to healthcare, treatment and effective illness management may also differ by generation, which may subsequently affect how an individual reports their health. Causal analysis of such pathways is beyond the scope of this cross-sectional analysis, so continued development of panel survey data will be vital to the further exploration of differences in the experiences of each generation.

Whichever speculated pathway may be in operation, the data presented here does not support a role for acculturative changes in health behaviours in mediating intergenerational trends in fair/poor general health. Differences in health behaviours (eg, alcohol intake) provide evidence that health risks can be determined by general characteristics common to specific ethnic groups such as religion. Despite changes in these health behaviours across generations, suggesting a (variable) degree of acculturation taking place, these did not translate into a significant change in general health. It is possible that the self-reported general health outcome was insensitive to the effects of changing behaviours in the relatively young sample used. For instance, the second-generation black Caribbean group are likely to be too young to report an adverse effect on their general health as a consequence of a significant smoking uptake. Future analyses that make use of biomarkers that are specifically linked to key health behaviours, and are predictive of later life illness, may cast more light on the influence of the post-migratory cultural environment on health outcomes in the relatively young second generations.

There is, however, the possibility that the observed rates of social mobility are a consequence of the way data describe the first generation. Owing to possible downward mobility post migration for the first generation, it is unclear to what degree upward social mobility might be mediated by post-migration class reassertion and whether these trends in mobility will persist across future generations. It is plausible, then, that the next generations will find upward mobility more difficult than their predecessors, perhaps leading to a widening of the socioeconomic inequality gap, with an accompanying increase in the health gap between ethnic minorities and the general population.

The cross-sectional design of the Health Survey for England has important implications for intergenerational investigations. As with any cross-sectional study, any link between improving socioeconomic circumstances and general health is associational and not necessarily causal. Furthermore, the data are poorly equipped to manage differential cohort effects within each generation and ethnic group, resulting from exposure to different historical contexts as a consequence of differences in the period of migration. For example, despite attempts to standardise for age and social circumstances, recent migratory history meant that it is likely that some of the Bangladeshi second-generation sample in particular had not yet reached their final social position owing to their relatively young age. More specific cohort effects also require greater attention in future investigations as it is likely that changes in socioeconomic position take place alongside changes in identity which are driven by contemporary contexts. As a consequence of these changing attitudes, health perception may also differ across and within generations. Although previous studies have shown a strong association between self-rated health and mortality across a variety of ethnic groups, there are no published data available to validate the intergenerational persistence of this trend. It is also possible that generational variations in the length of residence mediated group differences in general health. Unfortunately, it was not possible to examine the impact of the length of residence reliably due to the co-linearity which existed between generational status, the length of residence and age. Lastly, this study was unable to examine sex differences in self-rated health, owing to the small sample size.

Summary

Improving socioeconomic circumstances mediated the extent to which fair/poor general health was reported. However, upward intergenerational mobility did not subsequently lead to improved levels of general health but, instead, only maintained the same level of health inequality across generations.