• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of amjphAmerican Journal of Public Health Web SiteAmerican Public Health Association Web SiteSubmissionsSubscriptionsAbout Us
Am J Public Health. 2009 January; 99(1): 81–86.
PMCID: PMC2636599

Reductions in Disability Prevalence Among the Highest Income Groups of Older Brazilians

Abstract

Objectives. We sought to identify the income–disability prevalence relationship among older Brazilians.

Methods. Data were from 63 985 individuals 60 years and older from the 1998 and 2003 Brazilian National Household Surveys. Generalized additive logistic models with cubic regression splines were used to estimate the disability–income relationships.

Results. There was a strong linear relationship between increased income and reduced disability prevalence for most of the income distribution. Benefits were still present above the 90th percentile of income but were more modest. Because incomes among the wealthiest few are disproportionately large, odds ratios of disability nevertheless showed marked improvements, even across the very highest income groups.

Conclusions. Among older Brazilians, reduced disability is associated with higher income, and these associations are present even above the 90th percentile of income. In addition to understanding mechanisms of disability reduction among impoverished individuals, work is needed to understand these mechanisms in middle- and high-income groups.

Discussions of health inequalities are commonly concerned with poverty. Most empirical research in the field focuses on disadvantage, with relatively little analysis of the role of wealth1 or higher incomes. Populations worldwide are aging, and there is increasing interest in health inequalities in older populations; however, as with other groups, it is only recently that the effects of middle to higher incomes on aging outcomes have begun to be explored.2

Many studies have identified associations between lower income and poorer health.3 In addition, it is clear that the income–health relationship is causal, although the pathways involved are complicated.4,5 It is generally believed that health improvements with increasing income are relatively large but become progressively smaller across the middle and higher end of the income range, yielding a curvilinear relationship.6 At the cross-national level, for example, a pattern of markedly diminishing improvements in life expectancy is observed when examining mid- to high-range national income countries.7

Disability (i.e., having difficulty carrying out everyday tasks) is a critical measure of health for societies facing the challenges of caring for increasing numbers of dependent older people. Disability is a strong predictor of needing personal care and of having higher medical costs.8,9 Strong associations between socioeconomic status and disability among older people have been reported in developed countries,1017 although less is known about patterns in middle-income countries. For Brazil, Melzer and Parahyba18,19 showed that disparities in income and educational attainment are the most important sociodemographic markers associated with differences in disability prevalence in old age. However, the shape of the income disability relationship among older people in Brazil and other developing countries has received little attention.

We aimed to estimate the effect of increasing income on disability prevalence in a sample of over 63 000 older people across Brazil. Brazil has extreme income inequalities, and the available nationally representative data provide an opportunity to establish whether reductions in disability prevalence diminish as income increases from middle income through the highest income levels.

METHODS

The 1998 and 2003 National Household Surveys (Pesquisa Nacional por Amostra de Domicílios [PNAD]) were conducted by the Brazilian Institute of Geography and Statistics (Instituto Brasileiro de Geografia e Estatística). The data used in this survey are representative of the whole population residing in Brazil between September 1998 and September 2003, excluding only sparsely populated rural areas in the northern region, which was not included in the study. People in institutions were not included, but census (2000) data indicate that less than 1% of elderly people in Brazil were living in institutions.

The survey employed a multistage probabilistic sample of households. At the first stage, the units (municipal districts) were sampled. In a second stage, census sectors were selected within each municipal district of the sample. A simple systematic sample of households was then drawn in the third stage. The final sample was designed to be representative of the Brazilian population.20 Unofficial settlements and shantytowns were included in the sample, although response rates in these areas were difficult to assess.

The survey interview covered demographics, employment and occupation, health and physical mobility, education, income, migration, and household condition. Of 198 755 households sampled, interviews were held in approximately 90%, and 63 985 people 60 years and older were included in the sample: 28 943 in 1998 and 35 042 in 2003 (out of a total of 344 975 and 384 834 respondents of all ages in 1998 and 2003, respectively). Comparisons of the unweighted survey data with census information confirmed the survey's representativeness for gender and age.

Family-per-person monthly income was computed from total family income in the month preceding the interview divided by the number of people in the family. The income measure was calculated excluding persons who pay to live in the household, domestic servants, and relatives of domestic servants. To compare 1998 and 2003 income values, 1998 and 2003 figures were inflated to September 2004 values with an official Brazilian price index (Índice Nacional de Preços ao Consumidor; available at http://www.ibge.gov.br). Education was measured in terms of years of schooling.

Disability Measures

Difficulty in carrying out activities of daily living is a good measure of the impact of disease and aging among older people. However, reported difficulty with some activities (e.g., taking a shower or going to the toilet) could be directly affected by the better bathroom equipment that higher incomes can secure. A measure less affected by such confounding is necessary to reflect real differences in health. Walking is an everyday activity that all older people would normally do, and difficulty walking moderate distances (e.g., 100 m) is relatively unaffected by facilities. Mobility disability is often an early manifestation of the disablement process and is highly predictive of disability progression.2124 In this study, we used reported difficulty walking 100 m as a principal marker of the disablement process among older people, although we also explored activity of daily living disabilities.

We combined the 2 independent PNAD samples from 1998 and 2003 in this analysis. The same methods were used and the same disability questions were asked in both surveys. Disability questions were in the following format (in Portuguese): Normally, because of a health problem, do you have difficulty: (1) feeding, taking a shower or going to the bathroom?; (2) running, lifting weight, doing sports or doing heavy work?; (3) pushing a table or doing housework?; (4) climbing steps?; (5) kneeling down or bending down?; (6) walking more than 1 km?; (7) walking approximately 100 m? Response codes included “unable,” “great difficulty,” “little difficulty,” “able,” or “unknown,” and respondents were classified as having a disability if they gave any of the first 3 responses.

We used difficulty walking 100 m as the principal marker of disability. An additional, broader marker was examined in a sensitivity analysis, with disability defined as inability to do 1 or more of the following: running, lifting weight, doing sports, doing heavy work, pushing a table or doing housework, kneeling down, or bending down.

Statistical Analysis

Of the 63 985 elderly people for whom interview data were available, there were missing values for 31 on education and 2081 on family income, and these individuals were excluded from analysis. Information was obtained from the participants themselves in 63.4% of cases, from an informant in the household in 32.0% of cases, and an informant living outside the household in 3.6% of cases.

We used generalized additive logistic models with cubic regression splines25 to explore the functional form of the relationship between difficulty walking 100 m and family income per person. These models were fitted in R, a freely available software environment for statistical computing and graphics (R Foundation for Statistical Computing, Vienna, Austria; http://www.r-project.org), with the mgcv package (Simon Wood, author; available at http://cran.r-project.org) for generalized additive modeling. Linearity of the relationship between risk of disability and income, in different regions of the income distribution, was assessed by visual inspection of the estimated spline functions and by examining changes in the estimated degrees of freedom (edf) for the smoothed income term when models were fitted for individuals with incomes below selected percentiles (90th, 95th, and 98th). Values of the edf close to 1 were taken as evidence of linearity. All models were adjusted for age group, years of education, region of residence, place of residence, and study year.

Logistic regression models were used to analyze the association between difficulty walking 100 m and grouped percentiles of family income per person. Models were adjusted for the same factors as for the cubic spline regressions. The statistical analysis presented was not weighted because the sample is self-weighting and the addition of weights made little difference to the results. All data, except those from the cubic regression models, were analyzed with Stata software, SE version 9.2 (StataCorp LP, College Station, TX).

RESULTS

The sample included 28 140 men and 35 845 women aged 60 years and older. Most of the sample lived in the 2 major regions of the country (Table 1), southeast and northeast, and were concentrated in urban areas. The number of older women was greater than the number of older men, and the majority of sample members were aged 60 to 69 years. Attainment of education was limited, and some participants were living on very low incomes.

TABLE 1
Sample Characteristics: Brazilian National Household Surveys, 1998 and 2003

The median family monthly income (excluding people who declared zero income) was around R $273 (US $88) per person in September of 2004, rising to R $5139 (US $1661) per person in the top percentile. Income was grossly skewed, and jumps in median income became progressively larger toward the top end of the income distribution.

Table 2 shows that the unadjusted prevalence for difficulties walking 100 m (or being unable to do so) for those below the median per capita family income was 21.8% (95% confidence interval [CI] = 20.8, 22.8) for men, and this figure decreased to 7.7% (95% CI = 5.9, 10.0) in the highest income group (above the 95th percentile). As expected, the prevalence rates of difficulty were higher in women but had similar large differences by income group: 33.6% (95% CI = 32.4, 34. 8) in the poorest 50% and 18.3% (95% CI = 15.1, 21. 9) in the richest 5%.

TABLE 2
Unadjusted Prevalence of Difficulty or Inability to Walk 100 m, by Gender: Brazilian National Household Surveys, 1998 and 2003

Figure 1 shows spline plots of the adjusted odds ratios (OR) for difficulty walking 100 m as a function of family income per person, presented by gender. These models were adjusted for age group, living in an urban area, education, region, and survey year. Only individuals with an income above the median were included in the model fitting. Increases in income had a negative linear effect on risk of disability for both men and women living on middle-to-high incomes. Evidence for diminishing returns on the effect of income on risk of disability was seen at the top end of the income distribution (> 90th percentile; edf for smoothed income term in model fitted to individuals with incomes below 95th percentile, edf = 1.001 for men and 1.002 for women; edf for smoothed income term in model fitted to individuals with incomes below 98th percentile, edf = 1.700 for men and 2.1 for women).

FIGURE 1
Odds ratios for difficulty walking 100 m by per-person monthly family income for (a) men and (b) women: Brazilian National Household Surveys, 1998 and 2003.

Modeling of the income gradient in disability within each 10-year age category revealed some evidence of diminishing association at the greatest ages for men, but not for women (significance of smoothed income term: P = .580 for men, P = .007 for women aged 80 years or older; P < .01 in all other age–gender subgroups). This may reflect a selective survival effect among older men, consistent with findings in the US population.2

An alternative way of examining the income disability relationship is to divide the population into percentile groups by income (as in Table 2) and use conventional logistic regression models (Table 3). As with the spline plots, these logistic models were adjusted for education, age, region of residence, place of residence, education, and study year. The OR for disability in the 85th to 90th percentile of income was 0.53 (95% CI = 0.44, 0.63) for men compared with the bottom half of the income distribution, but in the 95th percentile, the OR was markedly lower (OR = 0.35; 95% CI = 0.28, 0.44). A similar pattern of sustained improvement across the highest income percentiles was present for women.

TABLE 3
Logistic Regression Model of Difficulty or Inability to Walk 100 m: Brazilian National Household Surveys, 1998 and 2003

Sensitivity Analysis

To assess whether the result reported was specific to mobility disability, we undertook the same generalized additive logistic analysis for a broader measure of disability. The alternative outcome recorded inability to carry out a wide range of everyday tasks, including lifting a weight, doing housework, or bending down (see “Methods” section). The shape of the income–disability relationship in this case (data not shown) was little changed from the results presented above. Accounting for proxy respondents in the models revealed that they were more likely to report walking difficulties (OR = 1.55; 95% CI = 1.46, 1.65) than were those who did not respond by proxy, but this did not influence the estimated income–disability association. Reported ethnicity was not significantly associated with walking difficulty before or after adjusting for other variables, including income (adjusted P = .10).

DISCUSSION

In this study, we examined income–disability relationships in a sample of over 63 000 older people across Brazil. In all analyses, there was a strong and predominantly linear relationship between higher income and lower prevalence of disability. Even for those elderly people in the highest 3 income groups (the top 15% of income), disability prevalence rates decreased when the values of per capita family income increased. Although we have found some evidence for a curvilinear relationship, this was only across the very highest income levels, with modestly diminishing disability gains present only across the top 10% of income. However, because absolute amounts of income at the high end rise very rapidly, the top 5% of the population still had markedly lower ORs for disability than those 5% or 10% lower on the income scale.

Limitations

In evaluating these results we should consider the limitations of the data. Socioeconomic differences in health can be measured by a range of markers, but there is evidence that income is a better discriminator of health status than are education or occupation.26 The self-reporting of income is always likely to be approximate, but reported values are thought to be lower than the actual values, so this bias would tend to reduce rather than increase the effect of income on health. In addition, there may be a tendency in the PNAD surveys for low-wage workers to be recorded as receiving the official minimum-wage value, and because of this factor, we concentrated here on the upper half of the income distribution. Income is highly skewed, with a few families receiving very large amounts, so we examined relationships both by actual income and by grouping the upper half of the income distribution into percentile groups, ranking the sample by increasing income, and reducing the effects of underreporting and the skew of the data. Our analysis primarily used mobility disability (i.e., difficulty walking 100 m) as the disability marker. This may not reflect overall disability, but when we repeated the analysis with a broader marker, the pattern of association with income was little changed.

The data presented are cross-sectional and provide no direct evidence of the causal direction between family income and disability prevalence. However, Davey-Smith and Lynch4 and others27 have argued that there is a great deal of evidence supporting a causal explanation for the effect of income on health.

Strengths

Although the available data have limitations, they also have enormous strengths. These include the very large sample size and nearly national coverage. This study provides the only national data (with the exception of the sparsely populated rural areas of the northern region) on disability among older people in Brazil. The data set includes key demographic and socioeconomic variables, allowing for the adjustment of models for potential confounders. The disability data include information on widely used markers of the disablement process, including mobility disability.

We are not aware of any previously published, comparable results with a large-scale data set from a developing country. For the United States, Minkler et al.2 found results similar to ours in older people. A number of theorists have argued that the shape of the income–health relationship has fundamental importance28: if the income–health relationship were attributable to poverty or material disadvantage, then a strongly curvilinear relationship would be expected. If, instead, the relationship is more linear, then the mechanisms are likely to be more subtle and include behavioral and psychosocial factors. These “subtle” mechanisms of disability in old age are likely to include medical care interventions, such as hip and knee joint replacements,29,30 cataract operations, and the benefits of many other forms of research derived from prevention and treatment technology.31 Unfortunately, no data are available in the PNAD surveys on access to such care.

An interesting finding in our analysis is the apparent reduction in the prevalence of reported difficulty walking 100 m in 2003 compared with 1998 (Table 3). For example, the relevant OR for men for 2003 was 0.86 (95% CI = 0.80, 0.92) compared with 1998, with a very similar OR for women (OR = 0.85; 95% CI = 0.81, 0.90). Parahyba and Da32 found this reduction in prevalence was more marked in the oldest age group (aged 80 years or older) and among poorer respondents (i.e., those with family incomes less than 3 times the minimum wage per person [US $248])—groups with relatively high prevalence rates. Reductions had occurred for both genders and in all the major regions of residence. A third wave of PNAD data will be collected shortly, and this will allow a more stable estimate of change to be calculated.

Lower rates of disability in rural areas were found in the 1998 survey.18 This may reflect real differences in health, but may also be caused by shorter survival times after onset of disabling disease (with relative lack of medical input), or migration of ill seniors into urban areas. Unfortunately, these hypotheses could not be explored in the available cross-sectional data. As noted in the “Sensitivity Analysis” section, a term for self-reported race was not significant in our models, perhaps reflecting the overriding importance of income differences, which outweigh race-specific effects in these data.

Overall, our analysis clearly shows that, in Brazil, increasing income remains a very strong factor associated with reduced mobility disability risks, even across the highest income groups. To maximize the gains in functioning in the older population over the coming decades, it will clearly be critical to address issues of poverty. In addition to this, however, more work is needed to understand health inequalities at levels of income well above that of poverty.1

Conclusions

Disability prevalence among older people decreases across the income gradient in Brazil, including across the highest income groups. There is evidence for modestly diminishing returns from rising income above the 90th percentile of income. In addition to poverty, the other factors involved in the progressive reduction in disability associated with increasing income need to be identified.

Human Participant Protection

This study involved secondary analysis of fully anonymized and publicly available data, and no additional institutional review board approval was necessary.

References

1. Baum F. Wealth and health: the need for more strategic public health research. J Epidemiol Community Health 2005;59:542–545 [PMC free article] [PubMed]
2. Minkler M, Fuller-Thomson E, Guralnik JM. Gradient of disability across the socioeconomic spectrum in the United States. N Engl J Med 2006;355:695–703 [PubMed]
3. Lynch G, Kaplan G. Socioeconomic position. : Berkman L, Kawachi I, editors. , Social Epidemiology New York, NY: Oxford University Press; 2000:13–35
4. Davey-Smith G, Lynch G. Socioeconomic differentials. : Kuh D, Ben-Shlomo Y, editors. , A Lifecourse Approach to Chronic Disease Epidemiology Oxford, England: Oxford University Press; 2004
5. Marmot M. The influence of income on health: views of an epidemiologist. Health Aff 2002;21:31–46 [PubMed]
6. Subramanian SV, Kawachi I. Being well and doing well: on the importance of income for health. Int J Soc Welfare 2006;15(suppl 1):S13–S22
7. Lynch J, Smith GD, Harper S, et al. Is income inequality a determinant of population health? Part 1. A systematic review. Milbank Q 2004;82:5–99 [PMC free article] [PubMed]
8. Fried LP, Guralnik JM. Disability in older adults: evidence regarding significance, etiology, and risk. J Am Geriatr Soc 1997;45:92–100 [PubMed]
9. Guralnik JM, Fried LP, Salive ME. Disability as a public health outcome in the aging population. Annu Rev Public Health 1996;17:25–46 [PubMed]
10. Kington RS, Smith JP. Socioeconomic status and racial and ethnic differences in functional status associated with chronic diseases. Am J Public Health 1997;87:805–810 [PMC free article] [PubMed]
11. Wolfson M, Rowe G, Gentleman JF, Tomiak M. Career earnings and death: a longitudinal analysis of older Canadian men. J Gerontol 1993;48(4):S167–S179 [PubMed]
12. Bassuk SS, Berkman LF, Amick BC., III Socioeconomic status and mortality among the elderly: findings from four US communities. Am J Epidemiol 2002;155:520–533 [PubMed]
13. Marmot MG, Shipley MJ. Do socioeconomic differences in mortality persist after retirement? 25 year follow up of civil servants from the first Whitehall study. BMJ 1996;313:1177–1180 [PMC free article] [PubMed]
14. Schoeni RF, Martin LG, Andreski PM, Freedman VA. Persistent and growing socioeconomic disparities in disability among the elderly: 1982–2002. Am J Public Health 2005;95:2065–2070 [PMC free article] [PubMed]
15. Melzer D, McWilliams B, Brayne C, Johnson T, Bond J. Socioeconomic status and the expectation of disability in old age: estimates for England. J Epidemiol Community Health 2000;54:286–292 [PMC free article] [PubMed]
16. Breeze E, Fletcher AE, Leon DA, Marmot MG, Clarke RJ, Shipley MJ. Do socioeconomic disadvantages persist into old age? Self-reported morbidity in a 29-year follow-up of the Whitehall Study. Am J Public Health 2001;91:277–283 [PMC free article] [PubMed]
17. Jagger C, Matthews R, Melzer D, Matthews F, Brayne C. Educational differences in the dynamics of disability incidence, recovery and mortality: findings from the MRC Cognitive Function and Ageing Study (MRC CFAS). Int J Epidemiol 2007;36:358–365 [PubMed]
18. Melzer D, Parahyba MI. Socio-demographic correlates of mobility disability in older Brazilians: results of the first national survey. Age Ageing 2004;33:253–259 [PubMed]
19. Parahyba MI, Veras R, Melzer D. Disability among elderly women in Brazil [in Portuguese]. Rev Saude Publica 2005;39:383–390 [PubMed]
20. Bianchini ZM, Albieri S. Principais Aspectos da Amostragem das Pesquisas Domiciliares do IBGE—Revisão 2002 Rio de Janeiro, Brazil: IBGE; 2003
21. Guralnik JM, Ferrucci L, Simonsick EM, Salive ME, Wallace RB. Lower-extremity function in persons over the age of 70 years as a predictor of subsequent disability. N Engl J Med 1995;332:556–561 [PubMed]
22. Lan T, Melzer D, Tom B, Guralnik J. Performance tests and disability: developing an objective index of mobility-related limitation in older populations. J Gerontol A Biol Sci Med Sci 2002;57:M294–M301 [PubMed]
23. Melzer D, Izmirlian G, Leveille SG, Guralnik JM. Educational differences in the prevalence of mobility disability in old age: the dynamics of incidence, mortality, and recovery. J Gerontol B Psychol Sci Soc Sci 2001;56:S294–S301 [PubMed]
24. Melzer D, Lan TY, Guralnik JM. The predictive validity for mortality of the index of mobility-related limitation—results from the EPESE study. Age Ageing 2003;32:619–625 [PubMed]
25. Wood S. Generalized Additive Models: An Introduction with R Boca Raton, FL: Chapman & Hall/CRC; 2006
26. Benzeval M, Judge K, Shouls S. Understanding the relationship between income and health: how much can be gleaned from cross-sectional data? Soc Policy Admin 2001;35:376–396
27. Benzeval M, Judge K. Income and health: the time dimension. Soc Sci Med 2001;52:1371–1390 [PubMed]
28. Mackenbach JP, Martikainen P, Looman CW, Dalstra JA, Kunst AE, Lahelma E. The shape of the relationship between income and self-assessed health: an international study. Int J Epidemiol 2005;34:286–293 [PubMed]
29. Melzer D, Guralnik J, Brock D. Prevalence and distribution of hip and knee joint replacements and hip implants in older Americans by the end of life. Aging Clin Exp Res 2003;15:60–66 [PubMed]
30. Steel N, Melzer D, Gardener E, McWilliams B. Need for and receipt of hip and knee replacement—a national population survey. Rheumatology (Oxford) 2006;45:1437–1441 [PubMed]
31. Pardes H, Manton KC, Lander ES, Tolley HD, Ulllan AD, Palmer H. Effects of medical research on health care and the economy. Science 1999;283:36–37 [PubMed]
32. Parahyba MI, Da SS. Disability prevalence among the elderly in Brazil. Cien Saude Colet 2006;11:967–974

Articles from American Journal of Public Health are provided here courtesy of American Public Health Association
PubReader format: click here to try

Formats:

Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...

Links

  • MedGen
    MedGen
    Related information in MedGen
  • PubMed
    PubMed
    PubMed citations for these articles

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...