Introduction
Women live longer than men. This is not a new phenomenon as in the eighteenth century when time series on life expectancy began in Sweden, a gender difference in life expectancy at birth of almost 3 years in favour of women was found (Human Mortality Database
2010,
http://www.mortality.org/). Similar gender gaps were present in European countries where time series started in the nineteenth century (France, Denmark, Netherlands, Belgium and Norway), except for Belgium where life expectancy at birth of men and women was equal (Human Mortality Database
2010,
http://www.mortality.org/). In the course of the twentieth century, the overall mortality reduction was more beneficial for women and resulted in a substantial widening of the male–female longevity gap. In the low mortality countries, this tendency has reversed in the last decades (European Observatory on the Social Situation—Demography Network
2009b). Until today, life expectancy of women exceeds that of men, although the size of the gender gap varies between populations (Jagger et al.
2008; Commission of the European Communities
2009).
The longer life expectancy of women may suggest that women are healthier than men. However, substantial evidence exists that women report worse self-rated health, more health problems, more frequent disability and use more health services, both in terms of hospitalisation and medications than same-aged men throughout life (Verbrugge
1989; Crimmins et al.
1996; Leveille et al.
2000; Case and Paxson
2005; Oksuzyan et al.
2008,
2009). This mortality advantage and health disadvantage is known as the ‘health-survival’ paradox (Oksuzyan et al.
2008). For a comprehensive assessment of health, in particular gender differences in health, it is important to consider both survival and health, rather than looking at either life expectancy or the health status of the population.
Health expectancy indicators, that combine mortality and morbidity data into a single composite indicator, are increasingly used to assess the health of populations or population groups (Robine et al.
1992,
2003; Robine and Jagger
2003). Health expectancy indicators share important attractive properties with life expectancy, such as their measurement in expected years of life and their independence from the age structure of the population, if small age intervals are used. An additional attractive feature is that health expectancy takes into account both mortality and the health status of the surviving population and thus provides information on the length of life (adding years to life), and the healthfulness of life (adding life to years). A well-known example is the disability-free life expectancy (DFLE). DFLE indicates how many years of the total life expectancy a person of a given age can expect to live without disability, and the difference between the total life expectancy and the DFLE is the life expectancy with disability. Different variants of DFLE exist, depending on how disability is measured.
Healthy Life Years (HLY) or the expected life years without activity limitations, which was selected in 2004 to be one of the structural indicators of the EU (European Commission
2010), combines information on mortality with disability based on the global activity limitation index (GALI). The GALI aims to capture long-term limitation (at least 6 months) in usual activities, caused by ill-health and includes three severity levels: not limited, limited but not severely, and severely limited (Van Oyen et al.
2006). Detailed information on Health Expectancies in Europe, including HLY for all EU member states is available from the European Health Expectancy Monitoring Unit (
http://www.ehemu.eu/) and the general public website devoted to HLY (
http://www.healthy-life-years.eu/). HLY differ substantially between EU member states and within member states (e.g., between men and women). Most striking is the gap between the original (EU15) member states (Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, the Netherlands, Portugal, Spain, Sweden and United Kingdom) and the new EU (EU10) member states that joined after 2004 (Cyprus, Czech Republic, Estonia, Hungary, Latvia, Lithuania, Malta, Poland, Slovakia and Slovenia). HLY in women at age 50 were 1.5 years lower in the pooled EU10 countries than in the pooled EU15 countries, and this disadvantage was 3 years in men (Jagger et al.
2008).
Reducing inequalities in HLY within member states receives high priority on the EU (European Commission
2009) and national policy agendas (VWS
2008), both from an equity point of view and because improving health of the disadvantaged group may improve health of the population at large. To reduce inequalities, insight is needed into the underlying causes of the inequalities, as these may point at possible interventions, either at the societal or the individual level, to reduce them. A first step is assessing to what extent differences in HLY are due to differences in mortality and disability, and at which ages these differences occur. In mortality research, decomposition (i.e., partitioning) techniques have been widely used to assess the contribution that age groups make to differences in life expectancy (Arriaga
1984; Pollard
1988) including gender disparities in life expectancy (Bah
1998; Trovato and Lalu
1998; Nusselder and Looman
2004). Similar tools are now available to partition differences in health expectancy into additive contributions of mortality versus disability differences and of age groups (Andreev et al.
2002; Nusselder and Looman
2004). For the Netherlands, using these tools has revealed that DFLE did not differ between men and women because the effect of higher prevalence of disability in women, reducing DFLE, and lower mortality, increasing DFLE, largely compensated each other (Nusselder and Looman
2004). But the same study also showed that due to the combination of the higher prevalence of disability and lower mortality women spent substantially more years with disability than men.
The aim of the current study is to describe and compare gender differences in HLY and unhealthy life years (ULY), defined as the expected life years with activity limitations, in the pooled EU10 and EU15 populations, and to explain differences in the overall patterns by looking at the contribution of mortality and disability differences and of different age groups.
Discussion
Both in the pooled EU10 and EU15 populations, women live longer than men, but women also spend more years and a larger proportion of their remaining life expectancy with disability. The gender gap in HLY, however, differs between the two populations. While in the EU15 population the gender gap in HLY is virtually absent, it is more than 3 years in the EU10 population. Our study thus shows that the health-survival paradox is present in both the EU10 and EU15 populations, but it also points at clear differences in the overall patterns of the gender gaps. In the EU10 the male mortality disadvantage is larger. Additionally, the commonly found disability advantage in men appears to be much smaller in EU10 males, and is virtually absent below age 65. Hence EU10 males face a double disadvantage: higher mortality than EU10 females and EU15 males and a smaller disability advantage than EU15 males. As a result, EU10 men not only have a shorter total life expectancy, but also the number of HLY is substantially lower than in women of the same age. In the EU15 population, this male disadvantage in HLY is not found.
A potential limitation of our study relates to using self-reported disability data from one single question in surveys of the non-institutionalised population. As medical or other registrations generally do not include information on disability, surveys are the standard source of disability information. Disability in large-scale surveys is mostly self-reported; only a few surveys and none that are across all EU countries include performance-based measures of disability. Relying on self-reports of disability may have biased our results, in particular because reporting behaviour may differ between men and women. However, comparing self-reports with performance measures confirms that women are truly more disabled than men (Merrill et al.
1997).
Using one single question to assess disability may have introduced bias, although we expect this to be less than that from using all non-harmonised disability measures, particularly since a previous validation study showed that the GALI question has an acceptable reliability (Cox et al.
2009). In addition, another recent study comparing the GALI with other disability measures, showed that the GALI appears to satisfactorily reflect levels of function and disability as assessed by longer-standing objective and subjective measures, both across Europe and in a similar way between countries (Jagger et al.
2010). The international comparability of the disability measures in our study is better than earlier efforts to obtain EU wide disability measures, including the ECHP. However, the SILC 2006 is still not fully comparable between the countries (Ekholm and Bronnum-Hansen
2009). In particular for Germany and Denmark the GALI question is not comparable to those in other countries. We reran our analyses excluding these two countries, but found no differences in our results. Excluding the institutionalised population may also have affected our results, as the rate of institutionalisation is known to differ by gender. However, given the higher institutionalisation rate in women, we expect that if anything, our conclusion on the double disadvantage in EU10 men as compared to EU10 women would become stronger if we had included data among the institutionalised population.
Some limitations of the methods need to be considered. The calculation of HLY and the decomposition of HLY are based on the Sullivan method. The Sullivan method is the standard way to calculate health expectancy on a routine basis, however, it does not produce a pure period indicator such as (period) life expectancy. While the deviation from a pure period indicator can introduce bias (Barendregt et al.
1994; Van de Water et al.
1995; Mathers and Robine
1997), this should not be important for comparisons between population groups. Regarding the decomposition method, it is noteworthy that the tool identifies the extent to which differences in disability prevalence and
total mortality (in each age group) contribute to health expectancy differences. It cannot assess the contribution of underlying flows, such as incidence of disability and recovery from disability (Nusselder and Looman
2004).
Another limitation is that our study treated the pooled EU10 and EU15 population as homogeneous groups, ignoring differences between the member states within each groups (Jagger et al.
2008; Commission of the European Communities
2009). The results of our study can therefore not be generalised to specific member states, but serves to understand better the overall picture. The next step will be to assess the variation in gender gaps within individual member states, and to assess which countries do and do not fit to the general picture.
Our study confirms the presence of the health-survival paradox (Case and Paxson
2005; Oksuzyan et al.
2009); both in the EU10 and EU15 populations women do live longer but spend more years with disability. There are different explanations for this paradox. One explanation is that women are healthier than men, but simply
report more health problems in surveys than men. Possible reasons for differences in reporting are that women are more likely to factor less serious ailments into their reports of poor health (Spiers et al.
2003), or that men are more reluctant to participate and/or accurately report in surveys if they have disabilities or diseases (Oksuzyan et al.
2009). There is substantial evidence that the health-survival paradox is not just an artefact. For instance, observed performance of physical tasks confirms that women have more disability than men of the same age (Merrill et al.
1997). A recent study of Oksuyan (Oksuzyan et al.
2009) found that the contribution of selection bias due to lower participation of men when they have ill-health or disability, is likely to be small. The same study also found no evidence for sex-specific reporting of medication use. The second explanation is that women do
have higher morbidity than men, and explanations include: a negative impact of estrogen and the second X chromosome, differences in the immune system to avoid the harmful effects of infections, a relatively higher compatibility of sick roles with other female responsibilities, engagement in more risk taking behaviour among men, as well as better awareness of disease symptoms, and timely seeking for medical advice (for an overview, see: Oksuzyan et al.
2009). Also the lower peak bone mass in women, accelerated bone loss beginning at menopause and less muscle strength in women may contribute to gender differences (Leveille et al.
2000). These factors may have resulted in gender differences in the distributions of chronic conditions (Verbrugge
1989; Case and Paxson
2005). Women are more likely than men to have non-fatal diseases, such as arthritis, whereas men are more likely to have conditions such as cardiovascular disease (CVD) or respiratory conditions with high case-fatality (Leveille et al.
2000). Additionally, these factors may have resulted in men having higher chances of experiencing hospital episodes and dying from the same chronic conditions, implying that men may experience more severe forms of these conditions.
A novel finding is that the pattern of gender differences varies between the EU10 and EU15 populations. In particular, in the EU10 the much stronger and younger mortality disadvantage (of the 8.4 years gap in life expectancy 5.1 years originates below age 65) in men, in combination with the virtually absent disability advantage below age 65, is striking. A possible explanation for this unfavourable situation is the rapid economic transition in the former communist states. According to McKee (McKee and Shkolnikov
2001) young men were especially vulnerable to the consequences of the policies pursued by the communist regimes in eastern Europe before 1990 and the ensuing transition. The leading causes of high mortality in these countries, in particular among men were injuries and violence, cardiovascular diseases and cancers (McKee and Shkolnikov
2001; European Observatory on Social Situation and Demography
2009a). High levels of alcohol consumption, especially binge drinking, were the main contributors to the high mortality from these causes in men, as did smoking and poor nutrition, but to a lesser extent (McKee and Shkolnikov
2001). Also the lower quality of medical care in the former Central and Eastern European Countries may have contributed to the more unfavourable position of EU10 men. A study of Newey et al. (
2003) not only showed a clear gender gap in preventable mortality in the new member states, but also indicated that treatable mortality, i.e., from causes that are responsive to medical intervention, through secondary prevention and treatment, was higher in men than in women in this region.
Our study has important implications. First, given the health-survival paradox, gender differences in health should be assessed by health indicators that take into account both mortality and morbidity. Additionally both healthy and unhealthy years need consideration. The gender gap in HLY masks important gender differences in mortality and disability within the EU15, while the similar size of the gender gap in ULY in the EU10 and EU15 masks the more unfavourable position of EU10 men, both due to higher mortality and a smaller disability advantage as compared to EU10 women and EU15 men. Second, the dire situation of men in the EU10, in particular below age 65, may have important consequences for their ability to contribute fully in society, and may preclude men from participating fully in working life, family life and society.