The prevalence of fair/poor self-rated health found in this study was rather high with no significant between-gender differences: 59.7% among men and 61.8% among women. These proportions exceed those reported in 1990s and 2000s in central and western European countries[46, 47], but are comparable to those in Baltic countries (56–66% among men and 64–68% among women), and are lower than in Russia (61.6% and 78.9%, respectively). Women tend to report fair/poor health more often than men[20, 28, 47–49]. However, between gender differences in self-rated health are frequently insignificant[2, 21, 50, 51]. A recent study from Sweden suggests that gender differences in self-rated health would disappear if women were as secure financially as men and were not treated in a condescending manner to a larger extent than men.
The proportion of women with fair/poor self-rated health in this study is considerably lower than the previously reported numbers from Armavir marz, Armenia (80.0% in 2001). For better comparability with these numbers, we examined the prevalence of fair/poor self-rated health in the subsample of women from Armavir marz in this study and found even lower proportion (53.4%), which indicated a positive dynamic in this indicator in Armavir marz and possibly in Armenia from 2001 to 2006. According to the country’s official statistics, this period was characterized by economic growth and corresponding decline in the proportions of both extremely poor (from 16.0% in 2001 to 4.1% in 2006) and poor (from 34.9% to 26.5%). The Armavir study found that material deprivation was the strongest predictor of poor self-rated health, with a clear dose–response relationship. In this study, the effect of poverty on self-rated health was mediated largely by psychosocial variables. Thus, in the final model, poverty remained an independent predictor of fair/poor self-rated health only among women. A growing body of evidence showed that the association between material conditions and health outcomes intensifies as income inequality increases and that a threshold of income inequality exists beyond which its negative impacts on health begin to emerge. The importance of psychosocial pathways through which material circumstances affect health indirectly also is well recognized and can partly explain our findings.
In the final fitted models for both genders, the strongest association with fair/poor self-rated health was found for psychosocial variables – being at risk for depression and reporting weak social support. Previous studies well documented the detrimental effect of depression on health[7, 30, 52, 53]. Similar to this study, a study from Hungary found depression to be a stronger predictor of self-rated health than socio-economic deprivation and mediate between the latter and self-rated morbidity rates, especially among men. The study authors hypothesized that in suddenly changing societies material deprivation and depressive symptomatology could worsen each other creating a vicious cycle that leads to higher self-rated morbidity rates.
Social support was found to be protective of health in many studies[7, 22, 25, 53–58]. Different emotional and instrumental aid pathways through which social support might influence health were suggested, including better coping abilities with stress, higher sense of self-esteem, self-efficacy and coherence, positive feelings of belonging and attachment, more engagement in health-promoting behaviors and refraining from health-damaging ones. We found greater impact of weak social support on perceived health for men than for women, a finding consistent with that from the French Gazel cohort study.
For men, education also showed strong independent dose–response association with self-rated health. This is a common finding in many transition countries[19, 29]. However, this association in transition countries is weaker than that seen in the west. The most common explanation for this difference is that the link between education, occupation and material well-being in these countries is not as consistent as in the west[2, 20, 61]. Thus, in countries like Armenia, better knowledge of health and better copping abilities among educated might play a more important role in the observed positive association between education and health than a secure position in the labor market that higher education usually guarantees in the west.
We found that educational level was independently associated with self-rated health among men, but not among women. This finding is consistent with other studies, suggesting that educational level has stronger health effect in men than in women[22, 61]. Instead, material conditions, affordability of healthcare and employment were found to be independent predictors of self-rated health among women, but not men. These factors are known determinants of self-rated health[21, 23, 24, 32, 38]. However, their independent effect on women’s perceived health can be explained by the fact, that women are more engaged in household duties than men[61, 62]. Together with unemployment that affected women disproportionately more in Armenia, this reality places women in a situation where the influence of household-related factors on their health is perhaps stronger than that of outside factors. A number of studies have shown the independent effect of the amount of household labor and housing attributes on women’s perceived health[43, 62, 63]. The influence of housing factors supersede the effect of educational attainment on perceived health. In this study, we did not measure the effect of housing and household labor on women’s self-rated health directly. However, poverty and low affordability of healthcare can serve as indirect measures of poor housing conditions making women’s household labor more strenuous for many reasons including lack of resources and amenities to facilitate the labor.
Previous research has repeatedly shown the relation of unemployment with poor health[22, 24, 26, 32]. Unemployment can influence health through different pathways including reduced income, psychosocial stress and loss of social networks. In transition periods, however, when unemployment is a widespread phenomenon, the psychosocial stress caused by it becomes less pronounced. This latter explanation is consistent with the situation in Armenian and our finding for men that unemployment was not an independent predictor of fair/poor perceived health. Employment might contribute differently to perceived health status for women in Armenia. Employment allows women partially to trade household duties for greater control over their life, which was shown to be associated with better perceived health[2, 19, 20, 28, 56]. Unemployed women, however, are fully dependent on their housing environment.
This study revealed a protective effect of moderate drinking on self-rated health among men. Previous studies have found a U-shaped relation between alcohol consumption and health: heavy drinkers and those with past history of alcohol abuse usually reported poor health, while those consuming small to moderate amounts of alcohol were more likely to report better health than abstainers[3, 14, 48, 65]. The positive effect of moderate alcohol consumption on health was attributed to its ability to reduce the risk of cardiovascular and cerebrovascular adverse health events. However, in cross-sectional studies, a reverse causation is also a possibility, when poor health prevents drinking. In a study in Russia, frequent drinking was linked to better self-rated health but was also predictive for higher mortality.
Our study did not find any other independent relations between self-rated health and behavioral/attitudinal variables. These results are consistent with previous studies, which also have shown that behavioral factors usually explain only a small portion of the socioeconomic differences in self-rated health[66, 67].
The refusal rate in this household health survey (21.1%) was in the acceptable range for population-based studies. The characteristics of those who refused to participate are unknown, but it is unlikely that response bias could affect the results. Although the sampled households were representative for each marz, the respondents were selected to preferably include married women or those having children less than 18 years and their husbands. This method could have resulted in an under-representation of older age groups. We compared the age structure of our sample with that of Armenia’s general population and found slight over-representation of middle aged (25–40 years old) and slight under-representation of older aged (65 and over) in our female sample. In the male sample, younger men (under 29) were slightly underrepresented, while older men (65 and over) were slightly overrepresented. This modest disparity might have hindered our ability to find between-gender differences in the prevalence of fair/poor self-rated health. However, age-adjusted logistic regression analysis of the whole sample (not shown) did not indicate significant between-gender differences in self-rated health.
The cross-sectional design of this study makes drawing conclusions on which factors are determinants and which are consequences of fair/poor self-rated health impossible. However, the set of factors independently associated with self-rated health in this study largely mirrors those identified by a number of longitudinal studies as determinants of ill health[7, 14, 24, 38, 53, 54, 65].
As the study was based on the secondary analysis of previously collected data, our choice of independent variables was limited to what these data could provide. Hence, several important potential predictors of self-rated health like perceived life control, details of employment and household labor were left out from this study.