INTRODUCTION
Activity limitations are difficulties an individual may experience performing physical tasks. They are considered negative aspects of functionality and constitute an important part of an individual’s disability status1,2, which is defined within the context of health as the person’s feature directly caused by disease, trauma or other health condition that requires medical care provided in the form of individual treatment by professionals (medical model)2. However, a person’s disability is also affected by the environmental characteristics of his/her surroundings (social model)3.
It is now argued that, as life expectancy continues to increase around the world, life will mean, on average, a longer period wasted by suffering due to frailty, disability, and age-related illness (e.g., cardiovascular diseases, cancer, diabetes mellitus, osteoporosis, among others)4,5,6. In Brazil, the population pyramid is in the process of contraction, resulting from a constant decrease of fertility rates, from 6.3 children per woman, in 1960, to 1.9 children per woman, in 2010. This is already below the replacement level, set as 2.1 children per woman7. At the same time, the epidemiological transition occurs differently from the model experienced by industrialized countries (e.g., Canada, the USA, Spain, Germany, England) and even by other nearby locations, such as Chile, Cuba, and Costa Rica. In Brazil, there is coexistence of old and new health problems, in which despite the predominance of chronic and degenerative diseases, the communicable ones still play an important role8.
In Brazil, social inequalities are prevalent in older adults due to the continuing inequitable distribution of resources among groups living under different socioeconomic conditions9. Social inequality plays an important role in the prevalence of disability among older adults10, especially among women11. It is likely that individuals living in areas with better socioeconomic conditions have greater access to medical goods and services and, consequently, are capable of maintaining good health conditions and functional capacity over time12. Those cumulative disadvantages mean that women are more likely than men to become poor and suffer from disabilities at an old age13.
In a previous study, 33% of older Brazilian adults had limitations to perform Activities of Daily Living (ADL) (37% of women and 27% of men). In the Northeast, this prevalence affected 39%, followed by the North and Midwest with 35% each, the South with 32% and the Southeast with 29%14. The prevalence of disability is significantly higher among users of the public system (32%) than those of the private health system (25%)15. The older adults who lived in Federative Units with higher Gini index were more likely to present higher level of disabilities compared to the older adults who lived in states with lower economic disadvantages12,16.
Many important studies provide an overview of the disabilities in Brazilian older adults, including gender differences and their determinants17,18,19,20,21. Other previous studies have measured the functionality of older adults in Brazil, where disabilities were associated with demographic (sex, age), socioeconomic (occupation, level of schooling, income), and health factors (chronic diseases)12,22,23. Differently of those previous investigations, this study uses the item response theory to aggregate two dimensions of functionality, i.e. ADL and Instrumental ADL (IADL), in a latent trait and estimate cross-level interaction analysis using multilevel modelling. We can, therefore, estimate the contextual effect (aggregate level) of an ecological exposure on individual risk (individual level). The aims of this study were to estimate the magnitude of gender differences in disability among older adults aged 60 and older in Brazil in 2013 and to examine whether those differences could be associated with social gender inequalities and socioeconomic contextual factors at the level of federative units.
METHODS
SAMPLE AND DATA COLLECTION
This was a cross-sectional study based on data from the Brazilian Health Survey 2013 (BNHS)23. The BNHS selected 81,187 households randomly, and individuals aged 18 years or older were interviewed. One individual from each household was selected in a complex sampling plan. Thus, 64,348 interviews were collected in the households, resulting in a non-response rate of 8.1%24. In total, 60,202 people participated in individual interviews25, in which 23,815 were older adults15. A selected population of the study was composed by 23,575 older adults aged 60 or older living in 27 Brazilian federative units. Calibrated individual weights were established in order to address participants’ non-response and sample attrition26. Response rates ranged from 84% in Pará to 96% in Bahia. A small number (1,8%) of individuals was excluded from the study due to lack of data for individual variables or because they were born with physical, hearing, cognitive, or visual impairments. Considering that the data used in this study are anonymous and available in the public domain, submission to the Ethics Committee was not required in accordance with the recommendations of the National Research Ethics Committee of Brazil27.
The National Health Survey is a household-based nationwide survey carried out by the Brazilian Department of Health in partnership with the Brazilian Institute of Geography and Statistics (IBGE). The scope of the survey is to establish the health status and lifestyles of the population - as well as how they look after their health - with regard to access and use of services, preventive actions, continuity of care, and health care financing.
ETHICAL APPROVAL
This study was exempted from submission to an ethics committee in accordance with the recommendations of the Brazilian National Committee for Ethics in Research, as the data used in this study are anonymous and available in the public domain.
DEPENDENT VARIABLE
The construct of the activity limitation index (ALI), which is used as a proxy to measure disability, was based on 12 questions in the BNHS. They were divided into two functionality dimensions: ADL (k1, k4, k7, k10, k13, k16 and k19)14 and IADL (k22, k25, k28, k31 and k34)14, but they were united in a single latent trait. The self-reported response categories in ADL were:
0: “no difficulty”;
1: “has little difficulty”;
2: “has great difficulty”;
3: “cannot” perform the activities.
The items related to the IADL were transformed into a dichotomous scale (0: “no difficulty”; 1: “has little difficulty, great difficulty, or cannot”) due to the scarce number of responses in some categories.
INDIVIDUAL INDEPENDENT VARIABLES
The main individual independent variable was sex. The following individual variables were also included in the analysis: age in years; restriction of usual activities during the last two weeks as a result of health problems (yes, no); self-perceived health (very good or good; fair, poor or very poor); medical health insurance plan (yes, no); participation in organized social activities (yes, no); educational level (illiterate or primary study; secondary or tertiary studies); Brazilian economic classification criteria, adapted and based in the sum of scores (by number of color television, bathroom, car, washing machine, DVD player, refrigerator, microwave), educational level and housekeeper (no = 0; yes = 4), as proposed by CCEB28. Thus, the social classes were categorized as follows: 0-13 (E and D); 14-50 (A, B and C) (Fiocruz, Rio de Janeiro, 2016, unpublished data).
CONTEXTUAL INDEPENDENT VARIABLES
The following contextual factors at federative unit level were used as additional independent variables:
The multidimensional gender inequality index (MGII) for 2010, which is based on six dimensions: education, income, work, politics, reproductive health, and protective factors. The methodological strategy used in the construction of the MGII reproduces the same mathematical procedure adopted by the United Nations Development Programme (UNDP)29. The MGII ranges from 0 to 1; the closer a score is to 1, the greater the level of social gender inequality30;
The per capita Gini coefficient for 2013 measures the per capita household income distribution. The measure varies between 0 and 1, where 0 corresponds to perfect equality and 1 to perfect inequality31;
Gross domestic product (GDP) per capita for 2013 is a proxy for living standards, indicating the levels of economic production of the territory33;
Life expectancy for 2013 reflects the average number of years a newborn can be expected to live. It is a summary measure of mortality as a proxy for improvement in living conditions and population health32.
STATISTICAL ANALYSIS
This study used a hybrid model (generalized partial credit model and two logistic parameters) of item response theory ((2LP)/GPCM IRT) to construct the ALI for each individual respondent. Despite the fact that the θ scale ranged from -∞ to +∞, it was transformed and restricted in continuous values between 0 (no activity limitation) and 1 (maximum activity limitation) for practical purposes, and thus allowing regression for nonnegative positively skewed dependent variable.
The second level of multilevel analysis considered those variables related to federative units, and the evaluation of ALI behavior at first level was seen as a function of the predictor variables for both levels. The coefficients of ALI and gender differences therein, as well as the 95% confidence intervals (95%CI) were estimated for each federative unit. To estimate the association between contextual variables and gender differences in disability, cross-level interaction effects were used in multilevel generalized linear models for non-normal (log-linked GLM Gamma) and zero-inflated (logit-linked GLM Binomial), with robust variance responses. Model I included activity limitation by sex, and there was no control for other individual variables. Model II, level 1, included all the individual and contextual variables; Model III (final) comprised all variables that were statistically significant in the multivariate analysis in model II. Intercept and sex factor were considered to be random effects on all models, fitted using gllamm command, using a statistical significance level of 0.05. The statistical analysis was performed using STATA 14.1 IC statistical software (Stata Corporation, College Station, TX).
RESULTS
GENDER DIFFERENCES IN DISABILITY
Among the 23,575 participants in the sample, 32.7% (95%CI 32.1 - 33.3) presented some difficulty in performing the evaluated activities, with a prevalence of 15.9% for ADL and 29.7% for IADL. The prevalence of disability (ADL + IADL) was significantly higher among women (37.6%) than in men (26.5%), similarly to the case with the average ALI (µwomen = 0.167, 95%CI 0.161 - 0.173; µmen = 0.119, 95%CI 0.112 - 0.125). In addition, older women presented disadvantages in terms of disability in the evaluated variables. More details on the characteristics of the survey participants can be found in Table 1.
Table 1. Description of study variables for men and women aged 60 years or older, 2013 National Brazilian Health Survey (n = 23,575).
Individual variables | Men | Women | ||
---|---|---|---|---|
n | % | n | % | |
Age group (years) | ||||
60 to 69 | 6,074 | 58.2 | 7,304 | 55.6 |
70 to 79 | 3,132 | 30.0 | 3,868 | 29.5 |
80+ | 1,237 | 11.9 | 1,960 | 14.9 |
Total | 10,443 | 44.3 | 13,132 | 55.7 |
Restriction of activities in the last two weeks | ||||
No | 9,449 | 90.5 | 11,481 | 87.4 |
Yes | 994 | 9.5 | 1,237 | 12.6 |
Total | 10,443 | 44.3 | 13,132 | 55.7 |
Self-perceived health | ||||
Very good or good | 4,847 | 46.4 | 5,530 | 42.1 |
Fair, poor or very poor | 5,596 | 53.6 | 7,602 | 57.9 |
Total | 10,443 | 44.3 | 13,132 | 55.7 |
Social activity participation | ||||
Yes | 994 | 90.5 | 1,651 | 87.4 |
No | 9,449 | 9.5 | 11,481 | 12.6 |
Total | 10,443 | 44.3 | 13,132 | 55.7 |
Health medical insurance | ||||
Yes | 1,646 | 15.8 | 2,529 | 19.3 |
No | 8,797 | 84.2 | 10,603 | 80.7 |
Total | 10,443 | 44.3 | 13,132 | 55.7 |
Educational level | ||||
Secondary or tertiary studies | 2,468 | 23.6 | 3,070 | 23.4 |
None or primary studies | 7,975 | 76.4 | 10,062 | 76.6 |
Total | 10,443 | 44.3 | 13,132 | 55.7 |
BR criteria (economic class) | ||||
A, B, C | 8,572 | 82.08 | 10,417 | 79.33 |
E and D | 1,871 | 17.92 | 2,715 | 20.67 |
Total | 10,443 | 44.3 | 13,132 | 55.7 |
Functional tasks*: | ||||
Activities of daily living | 1,425 | 13.7 | 2,315 | 17.6 |
Instrumental activities of daily living | 2,439 | 23.4 | 4,567 | 34.8 |
Total | 10,443 | 44.3 | 13,132 | 55.7 |
*n and percentage of limitation in any activity of daily living (ADL) or instrumental activities of daily living (IADL).
Gender differences in disability were observed in most of the federative units. Figure 1 shows the extent of gender differences in disability for each one. The differences and severity of the disability among women are greater in the North and Northeast regions. The closer the results are to the diagonal line, which represents gender equality, the fewer the differences in disability between men and women. Despite the fact that Maranhão (MA), Amapá (AP), Pará (PA) and São Paulo (SP) had the smallest differences between men and women, AP and PA had worse results than SP, but better than MA. However, Alagoas (AL) presented the greatest difference in disability between men and women, while maintaining a high level of it among the older adults. In this regard, we can extract two extreme models, AL and SP, each of them representing the negative and positive aspects of disability in Brazil.
GENDER DIFFERENCES IN DISABILITY AND CONTEXTUAL VARIABLES
In the first model, disability was significantly higher among women aged 60 or older (γwomen = 0.050, p < 0.001) when compared to men, as shown in Table 2. In an adjusted multilevel analysis (Model III), more advanced age, activity restrictions in the last two weeks prior to the survey, a health-self-perceived as “fair, poor or very poor” and being illiterate or having primary studies are factors associated with greater disability among older adults. On the other hand, social activity participation was associated with lower disability. Those functional disabilities were shown to be significantly influenced by income inequality in all federative units (γgini = 0.022, p < 0.001). Thus, income inequality has a negative impact on disability in older adults. The best model, in the end, was able to explain 61% of the differences between the federative units with the social gender inequalities. The differences in disability between men and women were increased when adjusted for the other variables (γmgiiXsex = 0.020, p = 0.004).
Table 2. Activity limitation index coefficients estimated using multilevel generalized linear models for non-normal response: a sample of Brazilian residents aged 60 or older from 27 federative units participating in the Brazilian Health Survey, 2013.
Model I | Model II | Model III | ||||
---|---|---|---|---|---|---|
Coefficient (SE) | p | Coefficient (SE) | p | Coefficient (SE) | p | |
Individual variables | ||||||
Sex (Men) | ||||||
Women | 0.050 (0.012) | < 0.001 | 0.043 (0.008) | < 0.001 | 0.045 (0.009) | < 0.001 |
Age group (60 to 69 years) | ||||||
70 to 79 | 0.133 (0.008) | < 0.001 | 0.132 (0.008) | < 0.001 | ||
80+ | 0.339 (0.017) | < 0.001 | 0.337 (0.015) | < 0.001 | ||
Restriction of activities in the last two weeks (Yes) | ||||||
No | 0.216 (0.015) | < 0.001 | 0.216 (0.015) | < 0.001 | ||
Self-perceived health (Very good or good) | ||||||
Fair, poor or very poor | 0.159 (0.018) | < 0.001 | 0.161 (0.019) | < 0.001 | ||
Social activity participation (No) | ||||||
Yes | -0.110 (0.012) | < 0.001 | -0.110 (0.011) | < 0.001 | ||
Health medical insurance (No) | ||||||
Yes | -0.018 (0.018) | 0.306 | ||||
Educational level (Secondary or tertiary studies) | ||||||
None or primary study | 0.076 (0.015) | < 0.001 | 0.083 (0.016) | < 0.001 | ||
BR criteria (A, B, C economic class) | ||||||
E and D economic class | 0.013 (0.013) | 0.337 | ||||
Contextual variables (intercept) | ||||||
MGII | 0.004 (0.013) | 0.764 | ||||
Gini | 0.021 (0.010) | 0.028 | 0.022 (0.004) | < 0.001 | ||
GDP | -0.006 (0.012) | 0.640 | ||||
Life expectancy | 0.001 (0.014) | 0.983 | ||||
Contextual variables (interaction with sex) | ||||||
MGII | 0.015 (0.007) | 0.024 | 0.020 (0.007) | 0.004 | ||
Gini | 0.004 (0.010) | 0.692 | ||||
GDP | -0.021 (0.013) | 0.092 | ||||
Life expectancy | 0.021 (0.014) | 0.132 | ||||
Intercept | -0.878 (0.013) | < 0.001 | -1.232 (0.026) | < 0.001 | -1.238 (0.030) | < 0.001 |
Variability | ||||||
Intercept variability (% of change*) | 0.253 | 0.063 (75%) | ||||
Interaction with sex (% of change*) | 0.137 | 0.053 (61%) |
*% of change in variability was calculated using this formula: [(initial variability-final variability)/initial variability]x100. Statistically significant variables (p < 0.05). In the response categories in brackets, the parameter is set to zero because it is redundant.
DISCUSSION
Results showed that the limitations in IADL were more prevalent than the ADL among older adults. This can be explained because the ability to perform the IADL is primarily affected, especially because it requires greater motor and cognitive skills, which tend to be lost early in the aging process33,34,35.
The participation of older adults in community activities and the restriction in the performance of activities in the last two weeks have a direct relationship with one another, pointing out an evaluation of the quality of social life among older adults towards the disability consequences. The participation in community activities is related to the control of the physical and social environment, reflecting better functioning and cognitive skills of older adults36.
The disabilities associated with self-perceived health status among older adults were consistent with other findings in previous research12,16. Hence, it has been suggested that ‘poor’ self-perceived health among women is a reflection of the higher burden of diseases they suffer. However, when this burden of diseases was compensated, the gender gap disappeared37.
Another finding was that the subscription to a health insurance was not associated with disability. It is likely that, in line with the theory of lack of information asymmetry, there is little significance in health condition variables that explain the private ownership status among older adults38,39,40. However, a Brazilian study showed that older adults who were enrolled in a health insurance had lower probability of high level of difficulty in accomplishing the ADL compared to those who did not have a health insurance12.
Regarding education and income, as the levels increase, the probability of the older adults to be able to perform ADL increase12. Low educational level has a direct association with greater disabilities among the older adults, compromising access to health education, which is a strategy that allows the adoption of healthy behaviors and social mobilization to improve living conditions41. Among older women, education seems to be associated with better functional capacity42.
GENDER DIFFERENCES IN DISABILITY
Similar to other studies10,12,16, disability was found to be worse among women in all the federative units. The high prevalence of disability in women is due to a combination of high incidence and long duration of disability, resulting from lower rates of rehabilitation and mortality among women with functional disability43. Gender may be a risk factor for explaining a worse physical function among women, but moving to equity could attenuate gender gaps in the physical function with advanced age13,44. According to BNHS data, all Brazilian regions showed similar levels of disability to the national average, with no differences between men and women34.
Two models of gender differences in disability can be highlighted: AL, with a high level of disability among older adults and a large disadvantage of women versus men, which is above the national average in both cases; and SP, with a low level of disability among older adults and a few differences between men and women, which is below the national average for both. Alagoas and São Paulo represent the socioeconomic extremes among the Brazilian federative units. In addition to the low level of human development, AL cannot meet the needs of the current generations without compromising the capacity of reaching the needs of future generations, due to the low sustainability. Therefore, urgent state interventions are demanded45,46. It is noteworthy that SP, besides the socioeconomic advantages that it presents ahead of many other federative units, is considered an “older adults friendly state”. In practical terms, SP adapts its structures and services to be accessible and inclusive to the older adults with varying needs and capacities47.
Therefore, SP is an example in which intersectoriality happens in policies directed towards older adults. The other federative units should see SP as a positive model, although not ideal, and replicate intersectoral policies with the aim of improving the older adult population’s life48. Thus, it is necessary to provide active and healthy aging, applying and implementing health information and education strategies, focusing on the prevention of disabilities and their determinants, going beyond prescriptive methodologies and placing older adults as transforming subjects of their reality.
GENDER DIFFERENCES IN DISABILITY AND CONTEXTUAL FACTORS
Social gender inequalities
In Brazil, the highest rates of social gender inequality were associated with higher gender differences in disability, with women at a greater disadvantage16. A previous study found that the greatest social gender inequalities in federative units seem to be associated with high incidence of long-term disability among older women43. In fact, gender differences in disability may be related to broader social gender inequalities, raising questions on how health, socioeconomic, and cultural factors influence gender patterns of seeking and using health care in later life in the country49,50.
Socioeconomic contextual factors
Income inequality is a main contextual factor in the social environment, and it affects directly health, increasing the probability of older adults presenting functional disability12. Therefore, income inequality deepens gender inequality and disadvantages for older women that have a history of low participation in the labor market, have minimal retirement benefits and depend on the public health system41. With lower incomes than men, women have less access to health and education that can ensure an improvement in their health quality, since highly educated women were more than 10 times as likely to age successfully as were women with low levels of education10,42.
In addition, it is noteworthy that disability is strongly associated with chronic disease, and functional disability relates to the performance of physical activities and social participation51. Therefore, prevention policies (for example, those that combat chronic disease) can help to delay the processes that lead to disability in older adults52. As in this study, other manuscripts did not find the GDP effect of disability on older adults in Brazil12,16. As for life expectancy, there was no effect either, even though it is an important indicator of the population’s quality of life. It has also demonstrated the negative impact of disability on the health of older adults and its differential effects on women, considering their higher disabled life expectancy18. Other studies could be conducted by adding new individual and contextual variables, including the disability-free life expectancy, and making a comparison with other countries.
STRENGTHS AND LIMITATIONS
This study provides a representative approach to self-reported functional performance of older Brazilians and contributes to the analysis of the health situation in Brazil, allowing researchers to trace the disability profile in Brazil. It also reinforces the importance of policies that reduce social inequalities in disability. Despite this, this study has some limitations. The response rates were lower in some federative units compared to others, although sample weights have been used to correct this. Another limitation is derived from the nature of the sample: information was only collected from older adults living in the community and excluded other living situations, such as nursing homes, convents, and hospitals34. Moreover, the database that represents the general population, has limitations, particularly because it addressed self-reported diagnosis data53. Older adults, at a disadvantaged socioeconomic level, might not have understood the questions asked during the BNHS or might be unaware of certain problems assessed in their locality, underestimating the prevalence. Furthermore, the study consisted of 27 samples in the second level of analysis, which could limit the ability to detect statistically significant differences between groups.
CONCLUSIONS
Women had higher disability disadvantages compared to men, and those differences were associated with social gender inequalities among the Brazilian federative units influenced by income inequality. However, it is important to consider that more studies are needed to elucidate other dimensions of functionality within and between the federative units, considering other social determinants of health.