Language-group mortality differentials in Finland in 1988–2004: assessment of the contribution of cause of death, sex and age

Abstract

Background: The higher mortality among Finnish-speaking than Swedish-speaking Finns is well established and has been shown to be very persistent. However, the most recent comprehensive documentation on the language-group differential extends only to the mid-1980s. We study recent trends in sex- and age-specific language-group mortality from 1988 to 2004, focusing on the relative contribution of the main causes of death and differences in sociodemographic structure. Methods: The register data are based on an 11% sample of Finns aged ≥30 years at the end of 1987—which further includes an oversample of deaths covering 80% of all deaths in the period 1988–2004. Results: Finnish-speaking men had 19% and women 11% higher age-adjusted mortality rates than the Swedish speakers. Further adjustments for sociodemographic composition explained 47% and 36%, respectively, of these differences. Age group-specific analyses showed the steepest differences in the young and middle-aged groups (30–49 years) in both men and women. The cause-specific differences were most pronounced in alcohol-related diseases, suicide and other external causes, for which a slight increase in differentials was also observed. Conclusion: The relative difference in mortality between the two main language groups in Finland persists, with a slight narrowing of the gap among women. Moreover, the largest differences and indications of diverging trends were found among the youngest age groups. The causes of death in which the difference is greatest highlight the importance of health-related behaviour and cultural differences in lifestyle.

Introduction

Most studies on ethnic minority-group inequalities in health and mortality indicate poorer health and higher mortality among minorities.^1–3 Swedish-speaking Finns are a notable exception to this almost universal phenomenon although describing them in terms of ethnicity is at best debatable because they are born in Finland, are citizens of Finland, and their origins go back at least a thousand years.⁴

Although the proportion of Swedish speakers in Finland has been steadily declining since the records began, they still comprise some 6% of the total population and form a distinct and well-established part of the Finnish national identity. Swedish has a status of official language in Finland alongside with Finnish.

The fact that Swedish-speaking Finns tend to live longer than their Finnish-speaking counterparts has been recognized for >50 years.⁵ Part of the difference is attributable to their better socioeconomic, demographic (e.g. marital status) and geographical circumstances.⁶ Differences in genetic composition or circumstances in early childhood⁷ and health-beneficial psycho-social processes such as a larger stock of social capital and trust have been suggested as contributory factors.^8,9 No consensus concerning the relevance of these alternative (albeit not exclusive) explanations exists, and it is likely that the different explanatory mechanisms do not apply to men and women, for various causes of death and in different phases of life. Since the most recent documentation on language-group differences extends only to the mid-1980s⁶ this study aims to provide a clear and up-to-date picture on how they have evolved by basic demographic (age and sex) groups and cause of death, and to assess the contribution of several socio-demographic explanatory factors.

Current evidence on the contribution of age, sex and cause of death

Despite the consistency of the overall phenomenon, the contribution of different age groups and causes of death to the gradient has varied considerably over time.^5,10–12 Fougstedt⁵ found it to be steeper among females than among males; in terms of life expectancy the difference was 3.5 and 2.5 years, respectively (Fougstedt's data covered the years 1936–45). By the 1970s the difference had narrowed to 1.2 years in women, but almost sustained the earlier level in men (1.9 years).¹² The most recent results show no changes in the magnitude of the sex-specific differences up to 1985.⁶

Following the first comprehensive study on language-group mortality differences, which identified infant mortality as the most important contributory group,⁵ the decline in overall infant mortality has been significant. Most subsequent studies have concentrated on the working-age population, more recent results^6,12 indicating that there are no longer any systematic differences in absolute mortality in childhood, and that the relative post-retirement difference is much smaller. However, as most deaths occur at older ages, their public health importance is significant, and they contribute strongly to the language-group difference in mortality.

So far, only a few studies have attempted to assess the relative contribution of different causes of death to the language-group mortality differential.^6,10,11 All of the previous research highlights the importance of circulatory diseases, suicide, accidents and violence, and alcohol-related causes. The only small reverse association has been found in neoplasms other than lung cancer among men and all neoplasms among women. In all, the group of main contributory causes suggests towards the importance of differences in health behaviour behind the language-group difference in mortality.

The study representing the most extensive effort thus far, to untangle the contribution of compositional differences in the language groups to the mortality difference was conducted by Koskinen and Martelin on data from 1971–85.⁶ The variety of possible compositional explanatory factors comprised educational attainment, occupational class, marital status, region of residence and level of urbanization in the municipality of residence. Overall they were able to explain totally the female mortality difference. The difference was larger to begin with among males and their adjustments explained only 37%. The modelling strategy used involved a set of expanding hierarchical models, which makes it difficult to compare the relative importance of different covariates. Moreover, the classification of the variables used was somewhat crude, which could have made the adjustment inadequate.

Aims

We used register-based data on the Finnish population in order to study trends in both overall and cause-specific mortality differences between Finnish- and Swedish-speaking men and women.

The specific aims of this study are:

1. To establish the trends in language-group mortality differential in Finnish men and women over the age of 30 years from 1988 to 2004.

2. To assess the relative importance of different age groups (30–49, 50–64, 65–79 and ≥80 years) to this difference and trends.

3. To study the relative importance of different causes of death to the mortality difference.

4. To study the relative contribution of different aspects of sociodemographic composition (education, social class, income, main economic activity and family status) and region of residence to the gradient.

Methods

Study population

The data are based on an 11% sample of Finns aged ≥30 years obtained from the annual Statistics Finland Labour Market data file covering all Finns for the period 1998–2004. The data set is complemented with an additional 80% random sample from the register of all deaths for the same period. Given the nature of the sampling procedure, appropriate weights are used in all analyses. To analyse mortality trends, we divided follow-up into four periods (1988–91, 1992–95, 1996–99 and for each period we had a separate baseline (1988, 1992, 1996 and 2000) and all the study variables were measured at these time-points.

Measurements

We defined two language groups: (i) Finnish speakers, (ii) Swedish speakers: all other language groups (between 0.36% in 1988 and 1.70% in 2000) were excluded from the analyses. The information on mother tongue comes from the population register, defined by the parents and registered simultaneously with the child's name during the 2 months following birth.

The explanatory variables used were largely based on previous research. It has been shown that Finnish and Swedish speakers differ with respect to socioeconomic status and education,^13,14 economic activity,^15,16 age composition, and income and family status (Table 1) and region of residence.

The occupation-based social class classification used in this study was based on the Statistics Finland standard (Classification of Socio-economic Groups 1989) comprised eight categories (i) upper non-manual, (ii) lower non-manual in autonomous or (iii) non-autonomous work, (iv) skilled and (v) non-skilled manual, (vi) farmers, (vii) self-employed and (viii) students or unknown. Data were available for 1985, 1990, 1995 and 2000. The 1985 information was used for 1988–89, the 1990 information for 1990–94, and so forth. Economically inactive persons were classified according to their occupation in earlier periods or to that of the head of the household, as it has been shown that exclusion or non-classification of the inactive population may underestimate the socioeconomic differences in mortality markedly.¹⁷

Educational level was categorised into six groups (Classification of educational groups 1997, Statistics Finland): (i) no schooling after the basic level (or unknown), (ii) lower and (iii) upper secondary, (iv) lower and (v) upper tertiary and (vi) licentiate or doctoral degree.

Income was measured at the household level in order to assess the disposable income, and consisted of the total income (salaries, entrepreneurial income, pensions, unemployment benefits, and most other social income transfers) of all household members. It was divided by the number of consumption units, defined as 1 for the first adult in the household, 0.5 for all other members aged ≥14 years and 0.3 for all younger members (OECD standard modified by Statistics Finland).

Family status comprised four categories: (i) married or co-habiting without children, (ii) married or co-habiting with child(ren), (iii) single parents and (iv) those living alone or unknown (2.6%).

The main economic activity of the previous year was categorized into four groups: (i) employed, (ii) unemployed, (iii) retired and (iv) others (homemakers and students, for example).

We based the region of residence on the so-called Major Regions (NUTS 2 EU standard) classification, which divides Finland into six rather large areas (i) Uusimaa, (ii) Southern Finland, (iii) Eastern Finland, (iv) Mid-Finland, (v) Northern Finland and (vi) Åland. We further separated the Helsinki Metropolitan area from the rest of Uusimaa, and thus had seven regional units.¹⁸

Cause of death

The cause of death was classified into broad groups separating (see Table 4 for the ICD-10 codes) cardiovascular diseases (CVD), lung cancer, other neoplasms, alcohol-related causes, other diseases, suicides and other violence and accidents.

Statistical methods

The distributions of the covariates and numbers of deaths were calculated over the four separate follow-up periods. Furthermore, we calculated age-adjusted death rates for each follow-up period (Table 1). Subsequent analyses were performed using Cox regression in STATA 9¹⁹ with sampling weights.

Results

The Swedish-speaking minority had, on average, higher socioeconomic position, education and income, was less often divorced or unemployed, and lived in much smaller proportions in eastern or northern Finland (Table 1). They were also considerably older than the Finnish-speaking majority. In terms of absolute mortality there was a considerable difference between the language groups; in males the difference ranged from 4.29 in the first period to 3.21 deaths per 1000 person-years in the last period for the advantage of the Swedish-speakers. In females the corresponding figures were 1.65 and 1.09. However, as the overall mortality rate is also declining, the relative difference between the language-groups remains nearly the same throughout the follow-up.

The age-adjusted total mortality rates of Finnish-speaking men and women were 19 and 11% higher, respectively, than those of the Swedish speakers (Table 2). There were no real changes among males within the 17-year follow-up, while there was some evidence of a levelling off among females towards the end of the follow-up (Table 2).

The difference was largest in the young age groups among both women and men. Furthermore, in the youngest group (30–49) the excess mortality of Finnish-speaking men increased from 54% in 1988–91 to 89% in 2000–04 (Table 2). In females the difference was also largest in the youngest group: the excess mortality increased from 14% and 45% over the follow-up period. However, neither of these increases reached the 5% level of statistical significance. Also, because death is relatively rare in this age group, the large relative mortality differences and the changes in them within the follow-up are not manifest when the total mortality of all age groups is considered (Table 2).

Adjusting for sociodemographic composition (SES, education, income, family status, main activity and region of residence) explained 47% of the overall age-adjusted mortality differences in males and 36% in females (Table 3). The variables that best explained the language-group differential were traditional measures of socioeconomic position such as occupational SES, education and income.

In males the relative mortality difference was easily the biggest for external causes and alcohol-related mortality (RR 1.58 and 1.97). Adjustment for all explanatory factors explained 38% and 33% of these differences, respectively (Table 3). The most marked differences in females were in alcohol related causes and suicide (Table 3): mortality among Swedish speakers was only about 50% and 60%, respectively, of that among Finnish speakers. This difference was only slightly (<10%) attenuated by the introduction of any of the explanatory factors alone (Table 3). The model with full adjustment had no further effect over the one produced by adjusting for occupational class and education alone. The only causes of death favouring Finnish-speaking women were lung cancer (RR 0.71) and, less markedly, other neoplasms.

The age-adjusted excess mortality due to neoplasms other than lung cancer among Swedish-speaking males was 6%. Male lung-cancer mortality was higher (14%) among Finnish speakers, although the difference was fully explained by the introduction of the full set of explanatory variables.

The analysis of cause-specific trends (Table 4) revealed a phenomenon that was somewhat common to both sexes. The mortality differences tended to widen specifically for those causes of death in which they were largest to begin with, and which were particularly common at younger ages. This widening was most marked in alcohol-related diseases (both sexes), and suicide and accidents and violence (women only). The difference due to non-alcohol-related diseases stayed at the same level throughout the follow-up (CVD in males), or more often showed a tendency to slightly narrow (CVD in females, lung cancer in males and other diseases in both sexes).

Discussion

Our analyses of recent developments in the language-group mortality differential in Finland show both continuity and discontinuity during the past two decades compared with earlier findings.⁶ The overall difference between the Swedish and Finnish speakers is still more or less intact: 19% among men and 11% among women. It may have slightly narrowed among females, but not in males. There were also divergent trends within the follow-up period by age group and cause of death. Overall, we found slightly narrowing trends among older females (specifically 65–79 years), and in mortality due to non-alcohol-related diseases. However, some indications of contrary processes were often found in younger age groups and for external and alcohol-related causes.

The degree to which the sociodemographic compositional differences explained the language-group difference in total mortality remained stable throughout the period at ∼30–50% of which socioeconomic factors comprised about half. For men the proportions explained for all-cause mortality were slightly larger than those presented in a previous study.⁶ This is likely to partly reflect the fact that we were able to more accurately measure different aspects of social stratification, but may also indicate that in more contemporaneous data sets the explanatory power of socioeconomic structure has increased for men. For women we observed the opposite: while we were able to explain about a third of the language-group mortality difference, Koskinen and Martelin⁶ explained all of it. Furthermore, in females the difference in relative mortality was smaller to begin with. It is not entirely clear why such a large difference is observed between the studies. Obviously part of the discrepancy may be due to differences in study settings—our age-limit was set 5 years lower and follow-up for mortality was 17 years rather than 15 years. In particular, our study covered mortality for the period 1988–2004 and the study by Koskinen and Martelin⁶ the period 1971–85. It is thus also likely that the differences between the studies reflect true changes in the level and causes of language group differences in mortality for women in Finland.

As mentioned, the largest language-group differences were observed for alcohol-related diseases, suicide and other accidental and violent causes of death for both men and women. Among men between 30% and 40% of these differences were explained in terms of differences in socioeconomic structure, particularly social class, income and main economic activity while among women the proportions were much smaller. This may indicate greater importance of work-related attributes to well-being among men. At the very proximate level of explanation alcohol-related diseases, of course, must be attributed to differences in alcohol consumption, thus health behaviours play an important role in the explanation of mortality differences between the language groups. Nevertheless, such explanations do not explain why Finnish speakers consume more alcohol, or consume it in more harmful ways.

The explanation for the diverging trend in younger people may stem at least to some extent from the severe recession of the early 90s, which was followed by a period of rapid economic growth in the latter part of the decade. At its worst the level of unemployment was close to 20% nationally, although it was lower in Western and Southern costal areas where most of the Swedish-speaking population live. It has been shown that the consumption of alcohol and the mortality related to it declined during the recession, but started to rise rapidly in 1996.^20–22 Our analyses (results not shown) of language-group-specific trends in absolute mortality due to alcohol-related diseases and to somewhat lesser extent also for external causes show that the rise in mortality after 1996 did not affect the Swedish-speaking minority, and mortality differentials between the language groups have widened accordingly. These findings may imply that the recession and economic recovery did not affect the Swedish speakers perhaps because they lived in areas that suffered least and recovered soonest. This interpretation is not fully supported by our data because adjusting for unemployment status did not fully explain the mortality difference. Additionally, there may be language-group-specific differences in the ability to cope with adversities such as unemployment or in particular how people react to declining real price of alcohol as economy recovers.²²

Because only somewhat less than half of the language-group difference can be explained in terms of sociodemographic structure alternative explanations must also play a role. These may relate to, e.g. social control of unhealthy or unsocial behaviour²³ that is inherent in the Swedish-speaking communities or broad cultural differences in alcohol consumption patterns.²⁴ It has further been suggested^8,9 that there are marked culture-based differences in resources such as social capital and support between the two population groups. If these resources are not to same extent available in the cultural context of the Finnish-speaking majority, or in the areas dominated by it, it may lead to coping through excessive alcohol consumption and other harmful health behaviour. These explanations for the level of language-group differences and the changes in them may work on the micro-level in families or on the macro-level in communities, or both. Furthermore, macro-level processes may work either through social activity in various kinds of associations (as suggested by Hyyppä and Mäki) or purely in local communities (e.g. neighbourhoods or municipalities), and may affect either the Swedish-speaking population only or both Swedish and Finnish speakers in these areas. Either way, given the distinct location of a large proportion of the Swedish-speaking minority, geography and locality are inevitably present.

In conclusion, we have shown that language-group differences in mortality persist in Finland, although they may be narrowing slowly among older women. More than half of these differences remain unexplained by differences in various socio-demographic dimensions or geography, when it is considered in very broad terms. On the international level, an important reason for studying Finnish language-group differentials in health and mortality is to pinpoint the positive processes or mechanisms that grant the minority group extended longevity it experiences compared with the majority population. The next step is to examine the geographical dimension more closely, as it is evidently a major factor here, and to focus on younger ages and alcohol-related and accidental and violent causes of death, for which the differentials may be increasing.

Funding

Gyllenberg Foundation; Finnish Cultural Foundation; Academy of Finland.

Acknowledgements

We thank Statistics Finland for permission (permission TK-53-1783-96) to use the data.

References