Population vulnerability and broad environmental climatic ...



Population vulnerability and broad environmental climatic exposure

Short version[1] for CD-ROM edition of the paper prepared for presentation at the Open Meeting of the Global Environmental Change Research Community, Rio de Janeiro, October 6-8 2001

Pedro Reginaldo Prata[2], MD, MSc, PhD, FRGS.

ABSTRACT

This paper comprises a geographical study of the spatial distribution of mortality patterns at population level, in Brazil, with a particular analysis of possible explanatory links with local and broad environmental dimensions in contrast with socio-economic conditions. The dependent variable was represented by patterns of ancient and modern mortality causes according to the epidemiological transition theory, as applied to the Brazilian setting. The independent explanatory variables were represented by a general socio-economic, and by local and broad environmental dimensions. The local one was represented by housing conditions water supply and sanitation, and the broad one by climatic characteristics (temperature, UV-B exposure and precipitation). The main method applied in this study was principal components analysis, which revealed associations of inter-related mortality causes and underlying explanatory conditions. The larger geographical areas showed that mortality patterns were associated with two independent underlying explanatory determinants. By contrast, the minor geographical areas, represented by highly urbanised populations, showed the same common mortality patterns but associated with only one underlying explanatory dimension. Multiple regression showed that socio-economic development, despite been important, was not the only dimension explaining the geography of mortality patterns in Brazil. At a local environmental level, housing conditions were found to be a more important environmental explanatory dimension than water supply and sanitation. At a broad environmental level the climatic dimension represented by temperature and UV-B exposure was present in several explanatory models for patterns of modern mortality causes of a chronic sort and for patterns of ancient mortality causes of an infective nature (either as a protective or as a risk factor). The same association with conditions where infections are involved was found with a second climatic dimension, represented by precipitation. The author argues in favour of a general vulnerability dimension, acting spatially upon populations, regardless of specific risk factors to specific diseases.

Introduction

Consistent or common patterns of apparently unrelated causes are often seen in the mortality characteristics of populations (Mackenbach, 1994). Explanation for these patterns may be needed, inasmuch as they indicate that some vulnerability dimension may predispose to illness at population level. Than a web of risk factors may follow precipitating poor health at individual level leading to disease, and enhancing morbidity and mortality at population level.

Hence, the study of the ever-changing disease patterns of populations may contribute to increased knowledge about their determinants at the population level (Evans et al, 1994). This is particularly true when the dimensions of space is taken into account to analyse the geography of mortality patterns in populations. This contrasts with the traditional view of single diseases being explained by single factors or multiple risk factors at individual level.

Differentials in disease pattern distribution spatially may be driven by socio-economic development and local or broad environmental determinants. Indeed the dynamic of environmental changes, which may affect population health are calling the attention of both the lay and scientific communities (McMichael, 1993a, 1993b and 1995; Bentham, 1992, 1993a, 1994).

This paper argues that common patterns of mortality causes do occur in the Brazilian context and verify if any broad dimension at population level can explain these patterns in a consistent way. Brazil, due to its continental size and contrasting environments, is a valuable setting for exploring whether these patterns are driven primarily by socio-economic development or whether there is also a clear local / or broad environmental dimension involved.

Moreover, this approach may contribute by proxy to the understanding of the human dimension effects of global environmental change especially those related to climate change as will be seen.

Methodology

The findings discussed here are based on an ecological epidemiological study characterising a study of aggregates (Susser, 1994.; Rose, 1985 and Mayer, 1983).

The setting

Brazilian larger geographical areas including its 26 Federal States and Brazilian; 27 minor highly urbanised geographical areas (including 26 Capital cities plus 1) have been included in a cross-sectional study covering a 23 years period (1979-1992). This period has been defined as a cross-section to enhance data stability.

The data

Eleven million death certificate records covering that 23 years period (available in CD-ROM) were used to build up a SPSS data-base where the place of residence and cause of death were selected. The population for the period was estimated from the 1980 and 1991 census for each of the geographical areas. Census data (1970, 1980 and 1991) for the socio-economic and local environmental variables were used to derive the indicators needed for this study (see below). The broad environmental variables were derived from a 15 years climatic data (cloud cover, temperature, humidity and rainfall) for the studied geographical areas. Population UV-B radiation exposure has been estimated according to latitude and cloud cover (Prata, 2000).

The analysis

The analytical method used was at first principal components analysis in order to disclose mortality patterns spatially (the dependent variables), and to summarise explanatory dimensions (the independent variables). Principal components analysis is a method used to find combinations of variables, called components, that explain the overall variation, thus reducing the complexity of the data. Moreover, the method permits an empirical approach without requiring in principle any previous assumption about the underlying structure, which might explain the distribution of the variables, as pointed out by Haynes (1986). Secondly, multiple regression was used to disclose the best-fit model to explain the geographical differences of mortality patterns.

The independent variables

The principal components analysis allowed data reduction from seven original socio-economic variables[3] to one general indicator of socio-economic level representing 78% of the geographical variability and one independent dimension represented by inequality (representing 17% of the spatial variability).

Similarly, the method allowed data reduction from four original local environmental variables[4] to only two local environmental indicators each representing almost half of the geographical variability: one expressing water and sanitation and the other housing conditions. Crowding conditions showed related to the sanitary dimension for the capital cities and to housing conditions at larger areas (such as the Federal States).

In the same way, the method allowed data reduction from four original broad (global) environmental variables[5] to only two broad climatic indicators: the main one expressing UV-B radiation exposure and temperature, representing 44% of the geographical variability and second one expressing rainfall and humidity, representing 36% of the spatial variability.

The dependent variables

The principal components analysis allowed data reduction from 47 causes of mortality to six for the Federal States and three for the urban capital cities (see tables 1 and 2). This expressed in summary two patterns of mortality: one called modern and another one called ancient. The modern pattern included some chronic diseases (cardiovascular and chronic respiratory conditions; ischaemic heart disease and multiple sclerosis, and cancer sites such as: breast, lung, prostate, lymphomas) and the ancient pattern included infections, TB, cancer of the liver, cervix and penis. It is important to highlight that at the State geographical level the method disclosed that each of these patterns (the modern and ancient) were related to two independent explanatory dimensions represented by each of the principal components (PC1 and PC2). On the other hand, at the urban capital cities geographical level, each pattern was related to the same explanatory dimensions represented by the first component (PC1).

Table 1 PCs, mortality causes groups, States geographical level

|Dependent |Associated |Associated |Associated |

|components |Main causes |Specific causes |Cancer Sites |

|PC1 |Neoplasms |Ischaemic heart diseases |Breast |

|(modern pattern) |Circulatory diseases |Chronic respiratory |Bladder |

| |Respiratory diseases |Multiple sclerosis |Leukaemias |

| |Musculoskeletal diseases Nervous |Stroke |Ovarian |

| |system diseases |Asthma |Kidney |

| |Congenital diseases | |Lung |

| |Skin diseases | |Lymphomas |

| |Endocrine diseases | |Hodgkin's disease |

| |Mental disorders | |Myeloma multiple |

| | | |Melanoma |

| |Ill-defined conditions (-) | |Colon and rectum |

| | | |Central nervous system |

| | | |Oesophagus |

| | | |Prostate |

|PC2 | | | |

|(ancient pattern) |Infections diseases |TB |Cervical |

| |Pregnancy complications |Appendicites |Penis |

| |Blood diseases |Rheumatic heart diseases |Liver |

Table 2 PC1, mortality causes groups, Capital geographical level

|Dependent |Associated |Associated |Associated |

|PC 1 |Main causes |Specific causes |Cancer Sites |

|Positive |Neoplasms Circulatory |Ischaemic heart diseases |Breast |

|Loadings |Respiratory |Chronic respiratory dis. |Bladder |

| |Musculoskeletal |Multiple sclerosis |Leukaemias |

|(modern pattern) |Nervous system | |Ovarian |

| |Congenital | |Kidney |

| | | |Lung |

| | | |Lymphomas |

| | | |Hodgkin disease |

| | | |Myeloma multiple |

| | | |Melanoma |

| | | |Colon and rectum |

| | | |Central nervous system |

| | | |Oesophagus |

|Negative | | |Prostate |

|Loadings |Ill-defined (-) |TB (-) | |

|(ancient pattern)|Pregnancy (-) |Appendicites (-) |Cervical (-) |

| |Infections | |Penis (-) |

| | | |Liver (-) |

The model fit

Tables 3 and 4 can be better summarised by the square value of the correlation coefficient (R2) that discloses what proportion of the dependent variable geographical variability is explained by the independent ones as they enter the model. The detailed discussion of the model will be found on the full version of this paper.

Table 3 Multiple regression, explanatory variables model-fit according to mortality pattern, States (negative correlations in italic)

|Models |Adjusted |R2 |

|Chr |R2 |% |

|SocEcoCon |0.859 |86 % |

|SocEcoCon + Hous |0.898 |91 % |

|SocEcoCon + Hous + Ineq |0.626 |94 % |

|SocEcoCon + Hous + Ineq + WaSa |0.931 |94 % |

|SocEcoCon + Hous + Ineq + WaSa + Clim-TempUV |0.936 |95 % |

|SocEcoCon + Hous + Ineq + WaSa + Clim-TempUV + Clim-Prec |0.937 |95 % |

|SpMod | | |

|Clim-TempUV |0.718 |73 % |

|Clim-TempUV + SocEcoCon |0.793 |81 % |

|Clim-TempUV + SocEcoCon + Hous |0.834 |85 % |

|Clim-TempUV + SocEcoCon + Hous + Ineq |0.875 |90 % |

|Clim-TempUV + SocEcoCon + Hous + Ineq + Clim-Prec |0.877 |90 % |

|Clim-TempUV + SocEcoCon + Hous + Ineq + Clim-Prec + WaSa |0.872 |90 % |

|CaMod | | |

|SocEcoCon |0.759 |77 % |

|SocEcoCon + Hous |0.862 |87 % |

|SocEcoCon + Hous + Clim-TempUV |0.897 |91 % |

|SocEcoCon + Hous + Clim-TempUV + Clim-Prec |0.897 |91 % |

|SocEcoCon + Hous + Clim-TempUV+ Clim-Prec + WaSa |0.897 |92 % |

|SocEcoCon + Hous + Clim-TempUV + Clim-Prec + WaSa + Ineq |0.893 |92 % |

|InfPreg | | |

|Clim-Prec |0.044 |8 % |

|Clim-Prec + SocEcoCon |0.023 |10 % |

|Clim-Prec + SocEcoCon + Hous |0.004 |12 % |

|Clim-Prec + SocEcoCon + Hous + Ineq |0.004 |16 % |

|Clim-Prec + SocEcoCon + Hous + Ineq + WaSa |0.053 |16 % |

|Clim-Prec + SocEcoCon + Hous + Ineq + WaSa + Clim-TempUV |0.107 |16 % |

|SpAnc | | |

|SocEcoCon |0.127 |16 % |

|SocEcoCon + Hous |0.257 |32 % |

|SocEcoCon + Hous + Ineq |0.411 |48 % |

|SocEcoCon + Hous + Ineq + Clim-Prec |0.499 |58 % |

|SocEcoCon + Hous + Ineq + Clim-Prec + WaSa |0.481 |58 % |

|SocEcoCon + Hous + Ineq + Clim-Prec + WaSa + Clim-TempUV |0.455 |58 % |

|CaAnc | | |

|SocEcoCon |0.364 |41 % |

|SocEcoCon + Hous |0.363 |41 % |

|SocEcoCon + Hous + Ineq |0.355 |43 % |

|SocEcoCon + Hous + Ineq + Clim-TempUV |0.342 |45 % |

|SocEcoCon + Hous + Ineq + Clim-TempUV + Clim-Prec |0.339 |47 % |

|SocEcoCon + Hous + Ineq + Clim-TempUV + Clim-Prec + WaSa |0.320 |48 % |

Table 4 Multiple regression, explanatory variables model-fit according to mortality pattern, Capitals (negative correlations in italic)

|Models |Adjusted |R2 |

|MainGr |R2 |% |

|SocEcoCon |0.676 |69 % |

|SocEcoCon + Ineq |0.675 |70 % |

|SocEcoCon + Ineq + Clim-TempUV |0.673 |71 % |

|SocEcoCon + Ineq + Clim-TempUV + Hous |0.665 |72 % |

|SocEcoCon + Ineq + Clim-TempUV + Hous + Clim-Prec |0.649 |72 % |

|SocEcoCon + Ineq + Clim-TempUV + Hous + Clim-Prec + WaSa |0.631 |72 % |

|SpGr | | |

|Clim-TempUV |0.541 |56 % |

|Clim-TempUV + Clim-Prec |0.650 |71 % |

|Clim-TempUV + Clim-Prec + Ineq |0.667 |72 % |

|Clim-TempUV + Clim-Prec + Ineq + WaSa |0.673 |72 % |

|Clim-TempUV + Clim-Prec + Ineq + WaSa + Hous |0.681 |73 % |

|Clim-TempUV + Clim-Prec + Ineq + WaSa + Hous + SocEcoCon |0.690 |73 % |

|CaGr | | |

|SocEcoCon |0.850 |86 % |

|SocEcoCon + Hous |0.874 |88 % |

|SocEcoCon + Hous + Clim-Prec |0.880 |89 % |

|SocEcoCon + Hous + Clim-Prec + Clim-TempUV |0.901 |92 % |

|SocEcoCon + Hous + Clim-Prec + Clim-TempUV + Ineq |0.910 |93 % |

|SocEcoCon + Hous + Clim-Prec + Clim-TempUV + Ineq +WaSa |0.915 |93 % |

The best fit model

All the relevant findings related to the explanatory power of the independent variables, in the goodness of fit of all possible models can be enhanced by looking at the best-fit models with the significant (at the 5% level, p (0.05) explanatory underlying variables, for each of the mortality pattern groups according to geographical area, as seen on Tables 5 and 6. Once more, the detailed discussion of the best-fit model will be found on the full version of this paper.

Table 5 Stepwise multiple regression, explanatory variables best-fit model according to mortality pattern, Capital

|Explanatory variables |b |s.e. | rp |p |

|MainGr | | | | |

|SocEcoCon |0.83 |0.11 |0.83 |0.00 |

|SpGr | | | | |

|Clim-TempUV |-0.75 |0.11 |-0.81 |0.00 |

|Clim-Prec |-0.39 |0.11 |-0.59 |0.00 |

|CaGr | | | | |

|SocEcoCon |0.83 |0.08 |0.91 |0.00 |

|Hous |0.19 |0.08 |0.44 |0.02 |

Table 6 Stepwise multiple regression, explanatory variables best-fit model according to mortality pattern, States

|Explanatory variables |b |s.e. | rp |p |

|Chr | | | | |

|SocEcoCon |0.73 |0.07 |0.92 |0.00 |

|Hous |0.36 |0.07 |0.71 |0.00 |

|Ineq |-0.19 |0.06 |-0.55 |0.01 |

|SpMod | | | | |

|Clim-TempUV |-0.26 |0.13 |-0.40 |0.05 |

|SocEcoCon |0.31 |0.10 |0.57 |0.00 |

|Hous |0.53 |0.13 |0.67 |0.00 |

|Ineq |-0.28 |0.10 |-0.53 |0.01 |

|CaMod | | | | |

|SocEcoCon |0.54 |0.09 |0.80 |0.00 |

|Hous |0.27 |0.09 |0.56 |0.00 |

|Clim-TempUV |-0.28 |0.10 |-0.53 |0.01 |

|SpAnc | | | | |

|SocEcoCon |0.88 |0.18 |0.73 |0.00 |

|Hous |-0.70 |0.19 |-0.62 |0.00 |

|Ineq |0.58 |0.17 |0.60 |0.00 |

|Clim-Prec |0.35 |0.16 |0.44 |0.04 |

|CaAnc | | | | |

|SocEcoCon |0.70 |0.19 |0.61 |0.00 |

Conclusions

Principal components analysis revealed mortality patterns (one ancient and one modern) associated with underlying explanatory dimensions. The larger geographical areas showed that these patterns were associated with two independent underlying explanatory dimensions. Highly urbanised populations showed these patterns associated with only one underlying explanatory dimension.

Multiple regression showed that socio-economic level, despite been important, was not the only dimension explaining the geographical variability of mortality patterns. At a local environmental level, housing conditions were found to be a more important environmental explanatory dimension than water supply and sanitation. At a broad environmental level the climatic dimension represented by temperature and UV-B exposure was present in several explanatory models for patterns of modern mortality causes of a chronic sort and for patterns of ancient mortality causes of an infective nature (either as a protective or as a risk factor). The same association with conditions where infections are involved was found with a second climatic dimension, represented by precipitation at larger geographical areas.

In regard to the first climatic dimension, the direction of the relationship showed by the regression model, points out to the UV-B exposure effect. The role of solar UV-B exposure on health has been described as having both a protective and a damaging effect depending on the outcome under study. Two summarising papers about these effects are De Gruijl, 1997 and Longstreth et al, 1998. The biological plausibility of the UV-B effect has been described as: direct genetic damage (risk factor for skin cancers), down regulation of the immune system (protective factor for Multiple Sclerosis and immune disorders and risk factor for infectious), acting upon metabolic pathways: Vit-D, Na and Ca equilibrium, hormonal regulation (protective factor for cardiovascular diseases and for some modern cancers).

Furthermore, a few hypothesis can be derived or enhanced by this study:

• Basic and common processes present in the onset of many cancers sites suggests that UV-B can be a protective factor for several modern cancers apart from skin cancers.

• Basic and common processes present in the onset of many cancers sites suggests that UV-B can be a risk factor for several ancient cancers where an infective agent may be involved (liver, penile and cervical).

• UV-B exposure can be a protective factor for several modern mortality causes, such as multiple sclerosis, stroke and ischaemic heart disease.

• UV-B can be a risk factor for ancient mortality causes of an infective nature.

• Rainfall can be a risk factor for ancient mortality causes of an infective nature in the interior.

The findings, which require further studies, strengths the importance of climatic variation upon population health. Moreover, these findings suggest a general vulnerability dimension, acting spatially upon populations, regardless of specific risk factors to specific diseases.

As can be seen, this study is based on a conceptual framework that considers on one hand individual susceptibility and on the other population vulnerability. For the author individual susceptibility to disease (related to the genome, to pre-natal and to life course determinants) has a relationship with specific risk factors operating at individual level, which are associated with specific diseases that requires exposed and not exposed individuals. Differently and complementary, the author considers that population vulnerability to disease (related to the genetic population pool[6], to socio-economic development and inequality, to local and broad or global environmental exposure[7]) has a relationship with disease patterns.

Moreover, appendix A below, presents by map figures (with the aid of the geographical information system software ArcView) the spatial distribution of the summarised variables indexes, in principal components scores, for the Brazilian States geographical level. This is for illustration purpose only. The full version of the paper will also show the figures for the Capital Cities. Notice that the higher positive scores or the darker colour shades indicates higher socio-economic development, higher inequality, better sanitation, better housing conditions, higher UV-B radiation & Temperature or higher levels of rainfall and humidity. Similarly, the higher positive scores or the darker colour shades indicates the higher prevalence of the modern or ancient pattern of mortality[8].

References

Bentham G (1993a). Depletion of the ozone layer: consequences for non-infectious human diseases. Parasitology; 106(S): 39-46.

Bentham G (1993b). Depletion of the Ozone Layer and Change in Incidence of Disease. CSERGE Working Paper; pp 25. UEA. Norwich.

Bentham G (1994). Global Environmental Change and Health. In: Phillips D and Verhasselt Y (eds). Health and Development: 33-49. Routledge. London.

De Gruijl FR (1997). Health effects from solar UV radiation. Radiation Protection Dosimetry; 72 (3-4): 177-96.

Evans RG, Barer ML and Marmor TR (1994). Why are some people health and others not? The determinants of health of populations. Aldine de Gruyter. New York.

Haynes R (1986). Cancer Mortality and Urbanisation in China. International Journal of Epidemiology; 15 (2): 268-71.

Longstreth J, Gruijl F, Kripk M, Abseck S, Arnold F, Slaper H, Velders G, Takizawa Y and Leun J (1998). Health risks. Journal of Photochemistry and Photobiology B: Biology; 46: 20-39.

Mackenbach JP (1994). The epidemiologic transition theory. Journal of Epidemiology and Community Health; 48: 329-32.

Mayer J (1983). The role of spatial analysis and geographic data in the detection of disease causation. Social Science and Medicine; 17 (16):1213-21

McMichael A (1993a). Global Environmental Change and Human Population Health: a Conceptual and Scientific Challenge for Epidemiology. International Journal of Epidemiology; 22(1): 1-8.

McMichael A (1993b). Planetary Overload: global environmental change and the health of the human species. Cambridge University Press. Cambridge.

McMichael A (1995). The health of persons, populations and planets: epidemiology comes full circle. Epidemiology; 6 (6): 633-36.

Prata PR (2000). Epidemiological Transition and the Geography of Mortality in Brazil. PhD Thesis; School of Environmental Sciences, University of East Anglia, UK.

Rose G (1985). Sick Individuals and Sick Populations. International Journal of Epidemiology; 14(1): 32-38.

Susser M (1994). The logic in ecological: I. The Logic of Analysis. American Journal of Public Health; 84 (5): 825-29.

Appendix

Map 1. Socio-economic development index (PC1) spatial distribution, Brazilian, States, 1970, 81,91

Map 2. Inequality index (PC2) spatial distribution, Brazilian, States, 1970, 81,91

Map 3. Water supply & sanitation index (PC1), spatial distribution, Brazilian, States, 1970, 81,91

Map 4. Housing conditions index (PC2) spatial distribution, Brazilian, States, 1970, 81,91

Map 5. Climatic PC1 (UV-B & Temp index) spatial distribution, Brazilian, States.

Map 6. Climatic PC2 (Rainfall & humidity index) spatial distribution, States

Map 7. PC1 Scores geographical modern pattern, main causes of mortality, States, 1979-1992

Map 8. PC2 Scores geographical ancient pattern, main causes of mortality, States, 1979-1992

Map 9. PC1 Scores geographical modern pattern, other specific causes of mortality, States, 1979-1992

Map 10. PC2 Scores geographical ancient pattern, other specific causes of mortality, States, 1979-1992

Map 11. PC1 Scores geographical modern pattern, cancers sites causes of mortality, States, 1979-1992

Map 12. PC2 Scores geographical ancient pattern, cancers sites causes of mortality, States, 1979-1992

-----------------------

[1] The full version will be available on paper to be published elsewhere. For further info contact the author e.mail: pedrorp@ufba.br

[2] Senior Lecturer at the Instituto de Saude Coletiva, UFBA, Brazil.

[3] Family income, low income (%), poverty threshold gap, inequality, illiteracy, schooling in years, and childhood conditions (children at school, child labour).

[4] Adequate water supply, adequate sanitation, adequate housing and over crowding.

[5] UV-B exposure (annual), average temperature, rainfall and relative humidity.

[6] As a result of natural selection at long evolutionary terms and selective survival at short generations terms

[7] In this case it operates at population level, and refers to widespread degrees of exposure where there is no clear cut discrimination between those exposed and those not exposed.

[8] Before deriving the PC scores all rates were adjusted for the age distribution of the period population.

-----------------------

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download