IJE Advance Access published online on May 24, 2008
International Journal of Epidemiology, doi:10.1093/ije/dyn088
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Morbidity and mortality in Brazilian municipalities: a multilevel study of the association between socioeconomic and healthcare indicators
1Department of Epidemiology and Biostatistics, Institute of Community Health, Fluminense Federal University, Niterói, RJ—Brazil 24033-900.
2Department of Epidemiology, Institute of Social Medicine, Rio de Janeiro State University, Rua São Francisco Xavier 524—7th floor, Rio de Janeiro, RJ—Brazil 20550-900.
* Corresponding author. Department of Epidemiology and Biostatistics, Institute of Community Health, Fluminense Federal University, Rua Marquês de Paraná 303—Annex HUAP 3rd floor, Niterói, RJ—Brazil, 24033-900. E-mail: lutricav{at}vm.uff.br
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResultsDiscussionAcknowledgementsReferences |
|---|
Background Socioeconomic and healthcare indicators are major determinants of health outcomes. The impact of social and healthcare inequalities on Brazilian morbidity and mortality indicators is of concern but it is not well studied.
Methods A multilevel ecological study was performed in order to investigate the association between a set of socioeconomic and healthcare indicators and five morbidity and mortality outcomes. Datasets were presented at three hierarchical levels: local (lower level), regional (intermediate level) and state (higher level). A Poisson regression model was estimated for each outcome with random intercept and fixed regression coefficients for independent variables at the three levels. The magnitude of outcome variability at intermediate and higher levels was assessed for all models.
Results All outcomes were associated with both socioeconomic and healthcare variables, with predominance of associations at the local level. General and high-complexity healthcare infrastructures were directly associated with indicators related to later stages of the demographic and epidemiological transition process. A mild effect on morbidity and mortality related to political voting patterns was found at the local level.
Conclusions Healthcare conditions and socioeconomic indicators are associated with health outcomes in a complex way at the local level in Brazil, but part of the variability of health outcomes is related to factors operating at higher levels. Some possible interaction effects and cross-sectional design limitations of this study must be considered.
Keywords Morbidity, mortality, inequalities, healthcare, cross-sectional studies
Accepted 21 April 2008
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionAcknowledgementsReferences |
|---|
The association between socioeconomic status and health indicators has been widely studied, and it is recognized as a cause-effect relationship. Access to social services such as education, housing, food, as well as equitable income distribution is important determinants of individual health conditions. Gaps in the distribution of these goods have contributed to health inequalities worldwide.1–6
The same type of association, as described earlier, is applicable for healthcare, though less prevalent, in the literature. Marmot7 stated that universal access to effective healthcare is a condition for a civilized society, and that those social goods prevent illness and suffering, but he considers this matter as being ‘not problematic’ and requiring ‘no further elaboration’. Probably the case of central capitalist countries, where healthcare systems have less heterogeneous access, quality and technology,8 in which those findings are based, are not easily comparable with the complex Brazilian health situation, both in terms of its epidemiological profile as well as the well-established, persistent inequality in allocation of public health resources, precludes such assumptions.9
In Brazil, in the light of evident achievements for the public health sector since the 1990's, the analysis of different healthcare profiles should be included in analytic and predictive models, taking into account the impact of socioeconomic conditions. For instance, the purpose of the present study was to assess the association between socioeconomic and healthcare variables and morbidity and mortality indicators for Brazilian municipalities in 2001.
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionAcknowledgementsReferences |
|---|
An ecological cross-sectional study was carried out using multilevel Poisson regression models. These models consisted of three levels: local (lower level), regional (intermediate level), a subdivision of Brazilian states and state (higher level). The numbers of units for each level were, respectively: 5505 municipalities, 136 regions and 26 states (the Federal District, composed by one municipality, was analysed together with its neighbour region and state).
The regions composing the intermediate level, unlike municipalities and states, are not administrative units. They are defined by the federal government in order to promote regional development policies.10 In 2000, their minimum population was 36 144 inhabitants, the maximum was 19 196 979 inhabitants and the median 756 354 inhabitants. Their minimum area was 2785.44 km2, the maximum was 482 748.80 km2 and the median 26 182.23 km2. The minimum number of regions per state is two, the maximum is 15 and the median 4.5.
The health indicators analysed were selected based on their ability to provide markers for the extremes of the epidemiological transition process, thus identifying differences constituting health inequities.11 We used mortality rates due to infectious diseases as indicators of health related to the initial phase of the epidemiological transition, while mortality rates due to neoplasms and circulatory diseases were chosen to represent its current phase. Numerators of the mortality rates were counts of death certificates whose underlying cause pertained, respectively, to the Chapters I, II and IX of the 10th Revision of International Statistical Classification of Diseases and Related Health Problems (ICD-10). The source of mortality data was the Brazilian Mortality Information System. Morbidity indicators were selected as regards to their role as indicators of ambulatory care sensitive conditions (ACSC). These conditions are traditionally used as health management indicators, but there is evidence they could express morbidity severity at a population level.12 Infectious bowel disease and diabetes were selected from the list of ACSC because they produced a large number of cases to provide stable rates. For hospitalization rates, the numerator was the count of hospital admissions whose causes were among ICD-10 codes A00–A09 (infectious bowel diseases) and E10–E14 (diabetes). The source of hospital data was the Brazilian Hospital Information System. Morbidity and mortality indicators were studied separately given the essential differences within these two health outcome subsets.13 Three-level Poisson mixed models with random intercept and otherwise fixed coefficients14 were fitted for each one of five outcomes. The number of deaths was corrected for under-reporting (whose estimated value is 20%, depending on the region, age group and cause of death) while deaths and hospital admissions due to ill-defined causes were non-proportionally redistributed.15 Under-reporting was corrected by modified James-Stein empirical Bayes estimators for events in delimited geographic areas were applied in order to increase the number of deaths in municipalities were under-reporting and high proportion of ill-defined events were identified.
At the local level, the independent variables covered the main dimensions of basic social policies other than healthcare, and all levels of complexity for the healthcare system. They represented the following dimensions: population (dependency ratios for those aged <15 and 65 years and older); education (proportion of children aged 7–14 attending school); income level (per capita income) and income distribution (Theil index); infrastructure (proportion of households served by water supply, sewage system and garbage collection); health financial autonomy (proportion of municipality budget allocated for healthcare); primary prevention (vaccine coverage in the first year of life); primary care (Family Health Program visits per inhabitant and basic medical procedures per inhabitant—the Family Health Program coverage was not regarded since it still showed low levels in 2000, the reference year); medium and high-complexity outpatient care services (specialized and high-complexity outpatient procedures per inhabitant); hospital services (hospital admission rate, ‘hospital export rate’: proportion of admissions per inhabitant in other municipalities and high-complexity hospital procedures per inhabitant) and health human resources (number of physicians and nurses per inhabitant).
At the second and third levels, the selection of variables aimed to express upper level decision-making instances. At the regional level, independent variables were the proportion of urban population and ‘hospital export rate’ (proportion of hospital admission of patients living in the region, which occurred in hospitals located in different regions). At the state level, a health policy variable was proposed: the proportion of state and federal representatives as well as municipality mayors belonging to left-winged (pro public healthcare) or right-winged (pro private healthcare) parties. Political parties were thus dichotomized based on the voting patterns of their members over the approval of the Healthcare Chapter of the current Brazilian Constitution.16
The source of data for socioeconomic variables was the 2000 Brazilian census official databases managed by the Brazilian Institute for Geography and Statistics (Instituto Brasileiro de Geografia e Estatística—IBGE). The source of data for (dependent and independent) variables related to healthcare were health information systems managed by the Ministry of Health's Informatics Department (Departamento de Informática do SUS—DATASUS), and for variables related to political parties, databases from the Superior Electoral Tribunal (Tribunal Superior Eleitoral—TSE) were used. All these databases can be freely downloaded from those institutions’ websites.
Independent variables refer to year 2000, since most of them came from the latest Brazilian census. The outcomes were calculated for 2001 in order to introduce a time gap between the independent and dependent variables but keeping the features of a cross-sectional design.
To prevent model collinearity, variance inflation factors were calculated. Values of this factor up to 10 and an average value not considerably >1 were regarded as acceptable.17
Given this was an exploratory study, the functional shapes between independent variables and investigated outcomes were not known. Thus, to prevent lack of adjustment, the original (numeric-type) independent variables were categorized in quartiles and the first quartile was the reference category. Alternatively, healthcare variables were dichotomized (availability or not of each procedure) in models when variables could not be included in the quartile format.
As the study comprised all Brazilian municipalities, regression models were weighed by municipality population size in order to stabilize variances in outcomes. The weighting procedure consisted of adding to the regression model a parameter known as offset, which is the natural logarithm of the expected number of events, obtained by outcome rate standardization.18
For parameter estimation, a second-order penalized quasi-likelihood (PQL) algorithm, which consists of second order terms for Taylor series expansion, was adopted. After this linearization process, models were estimated through restricted iterative generalized least squares, which provides less biased estimates of the variance components and, when the outcomes are continuous, it coincides with estimation through restricted maximum likelihood.19 Standard errors of regression coefficients were adjusted to allow extra-Poisson variance. Variable selection was based on Wald tests for the hypothesis that all coefficients related to that variable were simultaneously equal to zero. This test was used in this study, even though it produces only approximate estimates for non-linear regression models.20,21 As for keeping the second or the third levels in the analysis, we assessed the magnitude of the random intercept variance at each level, also based on Wald tests. CIs of 95% were calculated for all fixed effects, whilst point estimates and standard errors are presented for the variance of random effects. All the calculations were made in MLwiN version 2.01.22
|
Results |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionAcknowledgementsReferences |
|---|
Some findings were constant in all models: the presence of regional and state variability for the outcomes, extra-Poisson under-dispersion (outcomes showed variance lower than the mean) and inclusion of socioeconomic and healthcare variables. For instance, for infectious disease mortality, the Chi-square test result for keeping the regional level in the model was 38.26 (P-value = 0.0013), and for keeping the state level in the model was 10.92 (P = 0.0089); the variance for the outcome at the regional level was 0.0040 (SE = 0.0007) and the variance for the outcome at the state level was 0.0149 (SE = 0.0045); the measurement of the extra-Poisson variability resulted in a factor equal to 0.1719 (SE = 0.0033).
The largest value of the variance inflation factor was 8.2 and the average was 2.7; that was considered an acceptable level of collinearity, hence all independent variables could be selected in all models. However, each model showed some specific features, as mentioned subsequently.
For infectious diseases mortality rates, per capita income, Theil index and high-complexity outpatient procedures showed a positive and non-linear association. An inverse dose–response effect was observed for household sanitation services. Household water supply and basic medical procedures showed a inverse non-linear association. Intermediate-complexity outpatient procedures showed an inverse association in the 3rd quartile, and hospital export rate in the 3rd and 4th quartiles. High-complexity hospital admission rate was another variable inversely associated to this indicator. Otherwise, at the state level, the 2nd quartile of the proportion of mayors belonging to liberal political parties was associated to higher rates (Table 1).
|
Neoplasms mortality rates showed positive dose–response effect for dependency ratio at 65 years and older, per capita income and Theil index. The proportion of children aged 7–14 attending school was directly associated in the 3rd and 4th quartiles. The variable Family Health Program visits was inversely associated with this indicator. Only the 4th quartile of hospital fractions of high-complexity care was positively associated (Table 2).
|
The circulatory system disease mortality rate model evidenced the same effects as observed in neoplasms mortality for the following variables: dependency ratio at 65 years and older, per capita income, Theil index, proportion of children aged 7–14 attending school and Family Health Program visits. High-complexity outpatient procedures were positively associated in the 3rd and 4th quartiles. The 2nd quartile of the proportion of mayors belonging to liberal political parties was associated to higher rates at the state level (Table 3).
|
Hospital admission rates due to infectious bowel diseases showed a positive dose–response effect for dependency ratio at <15-years-old, and a similar, but inverse, effect for the number of nurses per inhabitant. The proportion of households served by sewage system and high-complexity outpatient care procedures were inversely associated, and the variables Health Family Program visits and hospital export rate were directly associated to this indicator (Table 4).
|
As seen in the neoplasm and circulatory system disease mortality rate models, the diabetes mellitus model produced similar results to dependency ratio at 65 years and older, per capita income, Theil index, proportion of children aged 7–14 attending school and Family Health Program visits. The 3rd and 4th quartiles of the proportion of households served by any sanitary installation, and the availability of high-complexity outpatient and hospital care were directly associated to this indicator. At the state level, the direct effect was observed for the 2nd quartile of the proportion of mayors belonging to liberal political parties (Table 5).
|
|
Discussion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionAcknowledgementsReferences |
|---|
This study intended to assess the association between morbidity and mortality indicators and socioeconomic and healthcare variables in Brazilian municipalities. Despite limitations discussed subsequently, some findings offer an opportunity for reflecting on the structure of the Brazilian epidemiological profile at the local level given the social and healthcare scenario in recent years.
The direct association found between income and income inequality with all mortality rates and diabetes mellitus hospital admission rates outlines a likely structuring of vital processes at the municipality level in Brazil, where high per capita income and income inequality together might have an adverse impact on some health outcomes. Recent studies have demonstrated existing mortality differentials related to income inequality in more affluent areas.23–27 Morbidity in higher per capita income Brazilian municipalities may be affected by the same process, as showed in several studies.28–39
As for aspects of the Brazilian health system structure evaluated in this study, it was found that the Family Health Program shows lower utilization rates where there are higher neoplasm and circulatory system mortality rates, as well as diabetes mellitus hospitalization. This finding may be a sign of sensible allocation of resources in accordance with the Programme's leading principles. The Programme aims at increasing healthcare access at less covered regions, which are at the same time the poorer and less served by infrastructure.40 However, this supposedly adequate allocation for the Family Health Program was not accompanied by a positive impact on one of its target outcomes, since infectious bowel disease hospitalization was directly associated to the Programme productivity.
High-complexity level of healthcare services and human resources allocation are in turn strongly structured in advanced demographic and epidemiological transition areas. This is evidenced by the positive association found between high-complexity outpatient care and all non-transmissible disease outcomes and between high-complexity hospital care and neoplasm mortality rates and diabetes mellitus hospitalization (the latter was also positively associated with the number of nurses per inhabitant). Previous studies demonstrated that Brazilian population ageing is more advanced in more developed areas of the country. The same has probably occurred with high-complexity healthcare supply and human resources allocation.41–43 Further analyses, using a longitudinal design, are needed to understand the apparent lack of impact of the higher level of healthcare on health indicators related to non-transmissible diseases.
The regression results for infectious diseases mortality rates showed a typically expected inverse relationship with basic sanitation coverage,44 the same occurring for hospitalization due to infectious bowel diseases, which evidences how susceptible morbidity and mortality rates due to infectious diseases are to basic sanitation, recognizably poorly and unequally implemented in Brazil.45 The contradictory scenario concerning the role of healthcare on mortality risk due to infectious diseases suggests some explanatory hypotheses: (i) a reducing effect of basic and intermediate-complexity outpatient procedures; (ii) lower infectious diseases mortality in areas with high-complexity hospital coverage corroborate other findings of this study, e.g. the positive relationship between outcomes related to advanced stages of the epidemiological transition process and the presence of a high-complexity hospital network; (iii) heterogeneity of infectious diseases subgroups, acting at different phases of the epidemiological transition process, constructing the ‘epidemiological complexity’ landscape, evidenced in Brazil by the coexistence of traditional, emerging and re-emerging infectious and parasitic diseases.46 These hypotheses, however, should be tested in further analyses.
An effect of the political context was found at the state level. Inclusion of the variable related to the proportion of mayors belonging to liberal political parties in three models, whose outcomes operate at different stages of the demographic and epidemiological transition process, has three major implications. First, is the identification of political instances as strategic players for health policy definition, which is a principle of the Brazilian health system. Second, it shows an association between political decision making and health outcomes, even though, due to the study design, it cannot be considered a cause-effect relationship. Third, it shows the extent of application of Brazilian public health system principles involving the ideological affiliation of its politicians.
The effect of the variable related to the proportion of mayors belonging to liberal political parties was similar in the models for mortality rates by infectious diseases and neoplasms, and hospitalization due to diabetes mellitus: a positive association for the 2nd quartile and no association for the 3rd and 4th quartiles. A likely interpretation for these results is that there is a negative impact on these health indicators in states with low proportions of mayors belonging to liberal political parties (pro public healthcare), and that this negative effect is no longer found when this proportion increases. Thus, non-compliance with the Brazilian health system principles by local policy-makers may result in negative health outcomes, showing the potential and actual impact of local health decision making in Brazil.47 But this hypothesis can only be properly tested in more complex models.
This study presented methodological limitations that should be addressed. The most important of all are related to its cross-sectional design, which limits assumptions of causal relationships between exposures and outcomes. Thus, the confirmation of the time relationship between the independent variables and outcomes demands additional studies with longitudinal design.
The quality of data was a concern in this study. In order to minimize problems deriving from the quality of information, only governmental databases, used to provide official demographic and socioeconomic indicators, were utilized. Regarding the outcomes, acknowledgment that mortality information has under-registration and high proportion of ill-defined causes (the latter occurring to hospitalization information as well) led to utilization of a correction approach to the mortality and morbidity data. Since the quality of health information is directly related to socioeconomic indicators,48 not performing this correction would introduce a measurement bias in the analysis.
This study indicates a need for incorporating complexity into the analysis of the Brazilian epidemiological structure in recent years. This complexity is not only evidenced by the number and variety of independent variables that were associated with health indicators but also showed the need for considering higher hierarchical levels. Results of multilevel regression models revealed how the variance breaks up between and within regions and states concerning the health outcomes studied. Nevertheless, further compositional and contextual variables responsible for this variability are yet to be uncovered.
The analysis described here, using Brazilian health information data, can be useful to support research on health inequalities in other regions of the world, including developed countries, since the fast demographic ageing process experienced over the second half of the last century in many developing regions is promoting a convergence among population structures worldwide.49 On the other hand, the implementation of neo-liberal healthcare reforms in developed countries over the last three decades is increasing the vulnerability of their social welfare system, and thus promoting an increasing similarity between the core and the periphery of capitalism.50 Thus, there is a wide field to be investigated, in which old and new health determinants can be identified, integrated, analysed, and moreover, included in innovative proposals against persisting health inequalities.
|
Acknowledgements |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionAcknowledgementsReferences |
|---|
This study was supported by Carlos Chagas Foundation for Research Support at Rio de Janeiro State, Brazil (E-26/151.246/2002 to L.T.C.).
Conflict of interest: None declared.
KEY MESSAGES
|
|
References |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionAcknowledgementsReferences |
|---|
1 Wilkinson RG, Pickett KE. Income inequality and population health: a review and explanation of the evidence. Soc Sci Med (2006) 62:1768–84.[CrossRef][Web of Science][Medline]
2 Lubetkin EI, Jia H, Franks P, Gold MR. Relationship among sociodemographic factors, clinical conditions, and health-related quality of life: examining the EQ-5D in the U.S. general population. Qual Life Res (2005) 14:2187–96.[CrossRef][Web of Science][Medline]
3 Jones AM, Nicolas AL. Measurement and explanation of socioeconomic inequality in health with longitudinal data. Health Econ (2004) 13:1015–30.[CrossRef][Web of Science][Medline]
4 Poerwanto S, Stevenson M, de Klerk N. Infant mortality and family welfare: policy implications for Indonesia. J Epidemiol Community Health (2003) 57:493–98.
5 Lindholm C, Burstrom B, Diderichsen F. Class differences in the social consequences of illness? J Epidemiol Community Health (2002) 56:188–92.
6 Harding S, Balarajan R. Longitudinal study of socio-economic differences in mortality among South Asian and West Indian migrants. Ethn Health (2001) 6:121–28.[CrossRef][Web of Science][Medline]
7 Marmot M. Social and economic environment and health. In: Social Determinants of Health—Marmot M, Wilkinson RG, eds. (1999) New York: Oxford University Press. 2–7.
8 Saltman RB. Equity and distributive justice in European healthcare reform. Int J Health Serv (1997) 27:443–53.[Web of Science][Medline]
9 Elias PE, Cohn A. Health reform in Brazil: lessons to consider. Am J Public Health (2003) 93:44–48.
10 Paasi A. The institutionalisation of regions: a theoretical framework for understanding the emergence of regions and the constitution of regional identity. Fennia (1986) 164:105–46.
11 Sen A. Why health equity? Health Econ (2002) 11:659–66.[CrossRef][Web of Science][Medline]
12 Basu J, Friedman B, Burstin H. Primary care, HMO enrollment, and hospitalization for ambulatory care sensitive conditions: a new approach. Med Care (2002) 40:1260–69.[CrossRef][Web of Science][Medline]
13 Wagstaff A, Paci P, Doorslaer EV. On the measurement of inequalities in health. Soc Sci Med (1991) 33:545–57.[CrossRef][Web of Science][Medline]
14 Hox JJ. Multilevel regression models. In: Applied Multilevel Annalysis—Hox JJ, ed. (1995) Amsterdam: TT-Publikaties. 11–30.
15 Cavalini LT, Ponce de Leon ACM. Correction approach for underreporting of deaths and hospital admissions due to ill-defined causes. Rev Saude Publica (2007) 41:85–93.[Web of Science][Medline]
16 Viana ALA, Queiroz MS, Ibañez N. Implementation of a single system: New relationships between the public and private sectors in Brazil. In: Health Policy Reform, Nacional Variations and Globalization—Altenstetter C, Bjorkman JWorg, eds. (1997) London: Macmillan Press Ltd. 331–47.
17 Chatterjee S, Price B. Multicolinearity and variable selection. In: Regression Analysis by Example—Chatterjee S, Price B, eds. (1991) New York: John Wiley & Sons. 191–93.
18 Draper NR, Smith H. The matrix approach to linear regression. In: Applied Regression Analysis—Draper NR, Smith H, eds. (1981) New York: John Wiley & Sons. 70–140.
19 Browne WJ, Draper D. A comparison of Bayesian and likelihood-based methods for fitting multilevel models. Bayesian Anal (2006) 1:473–514.[CrossRef]
20 Snijders T, Bosker R. Discret dependent variables. In: Multilevel Analysis: An Introduction to Basic and Advanced Multilevel Modelling—Snijders T, Bosker R, eds. (1999) London: Sage Publications. 207–38.
21 Goldstein H. The basic two-level model. In: Multilevel Statistical Models—Goldstein H, ed. (2003) London: Edward Arnold. 33–34.
22 Rashbash J, Browne W, Healy M, Cameron B, Charlson C. MLwiN version 2.01 (2004) London: Institute of Education/Multilevel Models Project.
23 Santos SM, Barcellos C, Sá Carvalho M. Ecological analysis of the distribution and socio-spatial context of homicides in Porto Alegre, Brazil. Health Place (2006) 12:38–47.[CrossRef][Web of Science][Medline]
24 Iyer S, Monteiro MF. The risk of child and adolescent mortality among vulnerable populations in Rio de Janeiro, Brazil. J Biosoc Sci (2004) 36:523–46.[CrossRef][Web of Science][Medline]
25 Szwarcwald CL, Andrade CL, Bastos FI. Income inequality, residential poverty clustering and infant mortality: a study in Rio de Janeiro, Brazil. Soc Sci Med (2002) 55:2083–92.[CrossRef][Web of Science][Medline]
26 Goldani MZ, Barbieri MA, Bettiol H, Barbieri MR, Tomkins A. Infant mortality rates according to socioeconomic status in a Brazilian city. Rev Saude Publica (2001) 35:256–61.[Web of Science][Medline]
27 Szwarcwald CL, Bastos FI, Barcellos C, Pina MF, Esteves MA. Health conditions and residential concentration of poverty: a study in Rio de Janeiro, Brazil. J Epidemiol Community Health (2000) 54:530–36.
28 Amaral MA, Nakama L, Conrado CA, Matsuo T. Dental caries in young male adults: prevalence, severity and associated factors. Braz Oral Res (2005) 19:249–55.[Medline]
29 Abbade LP, Lastoria S, Almeida Rollo H, Stolf HO. A sociodemographic, clinical study of patients with venous ulcer. Int J Dermatol (2005) 44:989–92.[CrossRef][Web of Science][Medline]
30 Santos IS, Boccio J, Santos AS, et al. Prevalence of Helicobacter pylori infection and associated factors among adults in Southern Brazil: a population-based cross-sectional study. BMC Public Health (2005) 5:118.[CrossRef][Medline]
31 Zago-Gomes MP, Stantolin GC, Perazzio S, Aikawa KH, Gonçalves CS, Pereira FEL. Prevalence of anti-hepatits A antibodies in children of different socioeconomic conditions in Vila Velha, ES. Rev Soc Bras Med Trop (2005) 38:285–89.[Web of Science][Medline]
32 Figueira-Silva CM, Pereira FE. Prevalence of Epstein-Barr virus antibodies in healthy children and adolescents in Vitoria, State of Espirito Santo, Brazil. Rev Soc Bras Med Trop (2004) 37:409–12.[Medline]
33 Melzer D, Parahyba MI. Socio-demographic correlates of mobility disability in older Brazilians: results of the first national survey. Age Ageing (2004) 33:253–59.
34 Martins MC, Fatigati FL, Vespoli TC, et al. Influence of socioeconomic conditions on air pollution adverse health effects in elderly people: an analysis of six regions in Sao Paulo, Brazil. J Epidemiol Community Health (2004) 58:41–46.
35 Peres MA, Peres KG, Antunes JL, Junqueira SR, Frazão P, Narvai PC. The association between socioeconomic development at the town level and the distribution of dental caries in Brazilian children. Rev Panam Salud Publica (2003) 14:149–57.[Medline]
36 Souza LJ, Souto Filho JT, Souza TF, et al. Prevalence of dyslipidemia and risk factors in Campos dos Goytacazes, in the Brazilian state of Rio de Janeiro. Arq Bras Cardiol (2003) 81:249–64.[Medline]
37 Pattussi MP, Marcenes W, Croucher R, Sheiham A. Social deprivation, income inequality, social cohesion and dental caries in Brazilian school children. Soc Sci Med (2001) 53:915–25.[CrossRef][Web of Science][Medline]
38 Freitas OC, Resende de Carvalho F, Marques Neves J, et al. Prevalence of hypertension in the urban population of Catanduva, in the State of Sao Paulo, Brazil. Arq Bras Cardiol (2001) 77:9–21.[Medline]
39 Viana MB, Fernandes RA, Carvalho RI, Murao M. Low socioeconomic status is a strong independent predictor of relapse in childhood acute lymphoblastic leukemia. Int J Cancer (1998) 11(Suppl):56–61.
40 Macinko J, Guanais FC, de Fátima M, Souza M. Evaluation of the impact of the Family Health Program on infant mortality in Brazil, 1990–2002. J Epidemiol Community Health (2006) 60:13–19.
41 Medici AC. Main features of Brazilian health system. In: Financing Health Policies in Brazil: Achievements, Challenges and Proposals. Sustainable Development Technical Department Papers Series—Medici AC, ed. (2002) Washington: Inter-American Development Bank. 2–4.
42 Berman P. Organization of ambulatory care provision: a critical determinant of health system performance in developing countries. Bull World Health Organ (2000) 78:791–802.[Web of Science][Medline]
43 Bossert T. Analyzing the decentralization of health systems in developing countries. Decision space, innovation and performance. Soc Sci Med (1998) 47:1513–27.[CrossRef][Web of Science][Medline]
44 Corvalán C, Briggs D, Kjellstrom T. Development of enviromental heath indicators. In: Linkage Methods for Environment and Heatlh Analysis—Briggs C, Corvalán C, Nurminen M, eds. (1996) Geneva: World Health Organization: United Nations Environment Programme. 19–53.
45 Soares SR, Parkinson J, Bernardes RS. Analysis of scenarios for wastewater and urban drainage systems in Brazil based on an integrated modeling approach. Water Sci Technol (2005) 52:53–60.[Web of Science][Medline]
46 Possas CA, Marques MB. Health transitions and complex systems: a challenge to prediction? Ann NY Acad Sci (1994) 740:285–96.[CrossRef][Web of Science][Medline]
47 Samuels D. Reinventing local government? Municipalities and intergovernmental relations in democratic Brazil. In: Democratic Brazil: Actors, Institutions, and Processes—Kingstone P, Power TJ, eds. (2000) Pittsburgh: University of Pittsburg Press. 77–100.
48 Bryce J, Boschi-Pinto C, Shibuya K, Black RE. The WHO Child Health Epidemiology Reference Group. WHO estimates of the causes of death in children. Lancet (2005) 365:1147–52.[CrossRef][Web of Science][Medline]
49 Restrepo HE, Rozental M. The social impact of aging populations: some major issues. Soc Sci Med (1994) 39:1323–38.[CrossRef][Web of Science][Medline]
50 Hunter D. The challenges of healthcare restructuring. Nurs Times (1997) 93:67–70.[Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  W. T. Caiaffa, C. da Silva Cardoso, and F. A. Proietti Commentary: Governance: does it matter in shaping health in urban settings? How in-depth can we go? Int. J. Epidemiol., August 1, 2008; 37(4): 784 - 785. [Full Text] [PDF]
|
|
|
|
|
|
S. Ebrahim Latin America: old and new challenges Int. J. Epidemiol., August 1, 2008; 37(4): 689 - 691. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||