This study with adults and older adults evidence of Brazil’s persistent racial inequalities in health and contributes to the understanding of the epidemiology of multimorbidity, especially in the context of low- and middle-income countries. Although multimorbidity is a common situation among ELSA participants, blacks displayed greater prevalence than whites and more morbidities in almost all quantiles of the distribution of the number of morbidities.
The prevalences found alert to the extent of multimorbidity in Brazil, which may be close to high-income countries with longer life expectancies, such as Portugal [42], Finland, Spain, and Poland [8]. Meanwhile, our prevalence was higher than those reported in analyses of the overall population of Latin America and the Caribbean, regarding both combined prevalence (43%; CI: 35%-51%) and combined prevalence excluding Brazilian studies (35%; CI: 26%-43%) [43].
The differences between the racial groups diverged from other Brazilian studies of the general population, which have reported greater prevalence in whites [22, 44], but these studies include adults 18 years of age and older, and diagnoses are self-reported. Rather, they approximate to those of the Longitudinal Study of the Health of Brazilian Older Adults, which specified multimorbidity as ≥ 2 and ≥ 3 morbidities, finding, respectively, 71.4% and 50.1% among blacks, 68.5% and 48.4% among whites, and 66.2% and 45.3% among mixed-race individuals [23].
Regarding the mean counts of morbidities, they were close to those reported in the Health and Retirement Study (USA), in which whites were found, on average, to live with 2.04 morbidities and blacks with 2.34 when follow up began [11], and, after adjustment, the count for blacks continued higher [10]. In ELSA, black and mixed-race participants returned higher counts in around 80% of the quantiles (particularly in the higher quantiles) of the distribution of number of morbidities after adjustment for age and sex. That evidence is an alert to the need to understand the mechanisms leading certain population groups to fall ill more than others.
Concerning the assessment of multimorbidity, the quantile analysis favours an approach reaching beyond the “average patient”. While the mean morbidity counts between racial groups were close (between 2.45 and 2.78 morbidities), and a linear or Poisson statistical model would work with these mean values, the quantile model illustrated that differences between racial groups widen as the number of coexisting morbidities is higher. For a population-based approach, it reveals the groups towards which health-promoting and disease-preventing interventions need to be directed so that the distribution curve of coexisting morbidities is shifted to the left.
Consistent with previous reports, an inverse relationship was found between prevalence and the cutoff used to specify multimorbidity [17, 45]. The traditional definition of multimorbidity (≥ 2 morbidities) revealed a high overall prevalence (70%) that was reduced to 46% when the cutoff changed to ≥ 3 morbidities. Concomitantly, the increase in the cutoff resulted in a greater estimate of association between race/skin colour and multimorbidity (especially for black participants). This evidence corroborates other discussions that suggest that increasing the cutoff in the operationalisation of multimorbidity may improve the specificity and provide greater differentiation between groups [17, 46]. For approaches at the individual level, it may identify those in a situation of complex multimorbidity, with multiple impaired body systems, which from a clinical point of view may be more meaningful to define patients who need differentiated integrated care.
Both the analysis by quantiles and the analysis with cutoff to define the multimorbidity led to the same conclusion: black individuals, especially, are in a situation of higher illness, with more coexisting morbidities. The quantile analysis is a resource when data about individuals are complete, and it is possible to define the exact number of morbidities. Alternatively, a cutoff for multimorbidity can be used in the presence of missing data when it is not possible to determine with accuracy the number of coexisting morbidities (e.g., when from a list of twenty morbidities there are records for only ten of them, but among these ten the individual has six).
Consistent with previous reports, our results suggest that the prevalence of multimorbidity is higher with advancing age and among women [3, 5, 6, 8]. The age acted as a negative confounder, underestimating the association between race/skin colour and multimorbidity when not controlled. Although the proportional distribution of racial groups by age group is close, the white group had 23.68% of individuals aged 60 years or more, against 17.34% of the mixed-race group and 19.6% of the black group. Since illness and diagnosis are related to advancing age, not controlling for age underestimated the true association. In addition, sex acted as a positive confounder in the association between black race/skin colour and multimorbidity (overestimating the association when not controlled). However, sex was a negative confounder in the association between mixed-race and multimorbidity. These effects are possibly explained by the proportion of women among the racial groups, since the black group presented a higher proportion of women (60.81%) in relation to the white group (53.76%), while the mixed-race group presented a lower proportion (51.62%).
This study found differences, by racial group, in the prevalences of multimorbidity and some morbidities (Additional File 3), complementing prior observations of racial inequalities in isolated diseases [47,48,49], but mixed-race and black women seem to be in worse health situations. That intersectionality (race and gender) can improve the documentation of health inequalities by addressing contextual factors that act jointly in producing illness [50]. The sensitivity analyses conducted here, considering an intersectional category for the independent variable, revealed no difference in multimorbidity in mixed-race and black men compared with white men, but rather that black women are in the worst situation (Additional File 5).
The results of this study should be read with certain points in mind. Adjustment for socioeconomic factors (Additional File 4) attenuated the magnitude of the association without fully explaining the associations found (especially for blacks). There are at least four reasons why race/skin colour still matters for health even when socioeconomic status is considered in analyses [34]. Firstly, health is affected by exposure to adversity throughout the life course [34]. Brazilian studies, population-based and nationwide, revealed that racial inequalities in health in the country are present from the beginning of life, e.g., with mixed-race and black women presenting a greater chance of inadequate prenatal care (which involves time to the start of prenatal care, number of appointments, exams, and receiving orientation), and with white children presenting better nutritional status in childhood [51, 52].
Secondly, there is a non-equivalence in the indicators of socioeconomic position across racial groups [34]. In our data, even though the mixed-race and black group has more than 50% of the individuals in the highest levels of education, they also have more than 50% of the individuals in the lowest levels of per capita family income. This is not the case for the white group. Thirdly, the most critical distinctive social exposure experienced by some racial groups is an additional effect of racism [34]. In ELSA-Brasil, black and mixed-race participants were more likely to live in economically segregated neighbourhoods [53], which may affect access to health-promoting conditions. Fourthly, some racial groups have an elevated risk of exposure to various psychosocial stressors [34], including experiences of discrimination throughout life – which were more frequently reported by mixed-race and black participants in ELSA-Brasil [54].
In addition, although Brazil does have a free, universal, public health system, a historical process of structural racism [21] has left substantial differences between racial groups in their access to health care and in care received [55, 56] which can have important implications for our results. The lack of diagnosis may mask even greater differences among the groups and, accordingly, our results may be underestimated, since ten of the 16 conditions on our list were self-reported. Corroborating this hypothesis, in the analysis considering six morbidities (Additional File 7), excluding the ten morbidities that are only self-reported in ELSA-Brasil, the differences between prevalences were amplified, and the association of mixed-race and black individuals with multimorbidity was stronger.
Another point relates to survival: life expectancy is shorter for blacks than for whites and levels of premature death (< 65 years) are higher [57, 58], so that blacks who reach advanced age may be less exposed to some risk factors throughout life and, as a result, display better state of health than the mean for their group or there may be differential survival among those with multimorbidity, which would be consistent with other situations of illness [59].
The data of this study offer no support for causal directionality, although it is reasonable to think that race/skin colour and accompanying experiences of racial discrimination precede the development of morbidities. Future studies are needed to investigate these factors’ roles in the prognosis and survival of different racial groups. The analysis of longitudinal data from ELSA-Brasil may answer whether the observed inequalities are related to disparities in the incidence and accumulation of morbidities, which have been documented regarding blacks in the USA [10, 11].
The patterns or severity of diseases or their impacts on daily and occupational activities or quality of life were not examined. Although the identification of which groups have the greatest number of morbidities is important to direct actions toward a target population, future studies are necessary to explore the differences in the situation of complex multimorbidity, in the identification of compromised body systems and in the pattern of diseases, which may differ between racial groups and differently impact the quality of life.
Lastly, generalisation of these findings to the overall population may be limited, because ELSA-Brasil included only civil servants in major cities, and black and mixed-race Brazilians historically have greater participation in informal occupations or are not occupied [60, 61]. For that reason, it is believed that racial inequalities in multimorbidity are even greater than this study has managed to measure.
The strengths of this study are that it used data on adults aged 35 and over who were neither hospitalised nor recruited by health services. Research into multimorbidity in younger adults deserves attention given the burden that multiple diseases place on the quality of life and health systems [3, 6, 13, 14]. In addition, it is essential to identify subgroups with more illness and vulnerable to multimorbidity, develop more equitable policies, and establish inter-sector strategies to act on the social determinants of health and mitigate their effects.
Another strength of this study is that the ELSA participants share certain occupational and socioeconomic similarities, and even after adjustment for variables of socioeconomic position and health risk behaviour (Additional File 4), differences in multimorbidity were attenuated, but not eliminated, especially for blacks, which is consistent with other reports [9, 33, 34]. Leading to the belief that the racial inequalities reported here reveal another facet of illness that is not captured by such indicators, but which may be life course-related in societies where unequal racial relations are determinants of illness.
