SPATIAL VARIABILITY OF CHILDHOOD OBESITY IN RESPONSE TO SOCIOECONOMIC HETEROGENEITY. THE CASE OF ATHENS METROPOLITAN AREA, GREECE

605

SPATIAL VARIABILITY OF CHILDHOOD OBESITY IN RESPONSE TO SOCIOECONOMIC HETEROGENEITY.

THE CASE OF ATHENS METROPOLITAN AREA, GREECE

Christos CHALKIAS1,*

, Apostolos G. PAPADOPOULOS2, Giorgios MPENEKOS

3,

Kostas TAMBALIS4, Glykeria PSARRA

5, Labros SIDOSSIS

6

1Assistant Professor, Department of Geography, Harokopio University of Athens, El. Venizelou 70,

Athens, 17671, Greece. Tel. +30 210 9549347, Email: [email protected] 2Associate Professor, Department of Geography, Harokopio University of Athens, El. Venizelou 70,

Athens, 17671, Greece. Tel. +30 210 9549318, Email: [email protected] 3Geographer, Msc Student, Department of Geography, Harokopio University of Athens, El. Venizelou

70, Athens, 17671, Greece, Email: [email protected] 4Researcer, Department of Science of Dietetics-Nutrition, Harokopio University of Athens, El.

Venizelou 70, Athens, 17671, Greece, Email: [email protected] 5Phd Student, Department of Geography, Harokopio University of Athens, El. Venizelou 70, Athens,

17671, Greece, Email: [email protected] 6Professor, Department of Science of Dietetics-Nutrition, Harokopio University of Athens, El.

Venizelou 70, Athens, 17671, Greece. Tel. +30 210 9549154, Email: [email protected].

ABSTRACT:

Childhood obesity (ChO) is a major health issue especially in developed countries. Spatial analysis

approaches of this issue focus on identifying spatial patterns of obesity and investigating the

association between ChO and geographic conditions. The work of many researchers worldwide

provides strong evidence of association between obesity and various environmental and

socioeconomic factors. In this study we assess the relationship between ChO & overweight with

socioeconomic factors in Athens metropolitan area; an area where severe socioeconomic segregations

exists. At the first stage of this analysis, the identification of spatial clusters of high ChO rates in

Athens metropolitan area is implemented. For this purpose, geospatial research for the

“obesity/overweight epidemic” is implemented with the use of GIS technology in a spatial analysis

research context. The first findings show clustering of ChO and overweight rates across Athens

metropolitan area and some spatial differences especially between obesity – overweight and/or male –

female rates. Moreover, it seems that the spatial clusters of ChO are concentrated in areas with

degrated socioeconomic conditions. The results of this analysis could be beneficial for ChO prevention

and reduction actions and policies.

KEYWORDS

ChO, Spatial Clustering, socioeconomic heterogeneity, Athens Metropolitan Area, Geography of

Health

1. INTRODUCTION

Childhood obesity (ChO) has been recognized as an epidemic in many developed and developing

countries (Martorell et al 2000, James, 2008). Specifically, in Greece, the prevalence of ChO has

increased by approximately 50% in the last decade, and nowadays over 40% of children are exceeding

normal weight, raising serious concerns for their health (Georgiadis & Nassis 2007). Obesity in

children is associated with immediate health risk factors (e.g. diabetes, hypertension e.t.c.), a higher risk of being obese in adulthood and also influences social and psychological functioning

(Williams et al 2005). Moreover, the food habits and culture of the children will influence their future

wellbeing and happiness, reflecting at the same time their familty socio-economic status and living

conditions. The study of spatial clustering in ChO/overweight could be a useful tool to develop and

evaluate targeted public health policies. Therefore, the aim of this study is to estimate the spatial

clustering of ChO/overweight, in almost all 8- to 9-year old children in Athens metropolitan area,

Greece. For this purpose, geospatial research for the “obesity/overweight epidemic” was implemented

with the use of GIS technology in a spatial analysis research context.

606

Geospatial research in health science is divided into the following general categories: a) Thematic

mapping of the spatial distribution of diseases (among others Rushton and Lolonis, 1996, Langford et

al., 1999), b) Spatial pattern detection methods as Openshaw’s Geographical Analysis Machine (GAM

- Openshaw, 1995), Kulldorff’s Spatial Scan Statistic (Kulldorff, 1997, Kulldorff, 2001),

geographically weighted regression (GWR Brunsdon et al., 1998, Fotheringham et al., 2002) and

Exploratory Spatial Data Analysis (Anselin, 2003, Jacquez et al., 2005) and c) Investigation of the

relationships between health problems and various socioeconomic and environmental variables (e.g.,

Chaikaew et al, 2009, Lim et al, 2010, Cao et al, 2010). Here, we consider the spatial patterns of ChO

focusing our analysis on Athens metropolitan area, Greece. This analysis addresses the following

questions: a) Are there spatially persistent areas of high obesity rates across the area under

investigation? and b) is socioeconomic heterogeneity of the study area related to ChO spatial

clustering?

The discussion is organized as follows. Section 2 presents both obesity and geographical data. Section

3 presents a methodological framework for estimating obesity clustering. Section 4 shows the results

of the proposed analysis. Finally, Section 5 concludes the discussion.

2. DATA

2.1. Study design and participants

ChO data derived from a yearly, national school-based health survey (Tambalis et al. 2010).

Specifically, anthropometric data and information on age, gender, city and area were collected,

between every May from 1997 to 2007, in almost all schools of Primary Education (<85%). In this

study we analyse data from 1997 and 2007 surveys concerning almost the total amount of students

(24,085 students for the 2007 survey) 8 to 9- year-old (51.05% males) of nearly all the primary

schools of Attica prefecture. Athens metropolitan area has been chosen for our study as it consists of a

wide heterogeneity according to areas of children’s setting (different types of residential areas) and

ethnicity (e.g. natives, immigrants) and it represents about the 35% of the whole student population of

Greece. BMI cut-off points were used by age and sex category, according to IOTF standard definitions

for underweight, normal weight, overweight, and obese (Cole et al., 2000, 2007), as the most proper

for large epidemiological studies.

607

Figure1. Study area – Athens metropolitan area and administrative units

2.2. Anthropometric Measurements

Anthropometric measurements were performed by trained physical education professionals. The

students were measured in the morning, without their shoes and lightly dressed, using a standardized

procedure. Weight was measured with electronic scales with a precision of 100 gr. Standing height

was determined to the nearest 0.1 kg with the child’s weight being equally distributed on the 2 feet and

head, back and buttock on the vertical land of the height gauge. BMI was calculated as the ratio of the

body weight in kg to the square of body weight in meters (kg/m2).

2.3. Geographical data

The geographical database of this study consists of various layers representing the administrative

districts of Greece and of background – environmental and socioeconomic information. Furthermore,

weight condition data from the survey were added in the GIS database in order to perform the

geocoding of ChO data.

3. METHODOLOGY

In order to analyse and visualize spatial patterns of ChO GIS technology was adopted and the

following general procedures were implemented: a) Spatial Database creation, b) Geocoding –

Aggregation of data c) Assessment of spatial clustering and d) preliminary investigation of the

relationship between ChO and socioeconomic heterogeneity.

The spatial database was created in GIS context with the use of ArcGIS software package. This

geodatabase consists of thematic layers representing the administrative districts (area topology)

according to the “Kapodistrias” administration plan (Figure 1) and the primary schools of the study

area. Next, we proceed to the geocoding and the aggregation of ChO data. For this task the address of

each school as well as the thematic layer of administrative units (municipalities) were used. With this

procedure the allocation of each primary school in the spatial background of the study area was

implemented. The next step was the calculation of the total number of ChO cases for each

administrative unit. These counts were expressed next as proportion of the population size (incidences

per 1000 children) in order to estimate the prevalence of the phenomenon under investigation. This

action is typical for spatial epidemiological analysis.

In order to reveal the aforementioned local associations of ChO the freeware SatScan package (version

9.0) was implemented. The SatScan method is used in many applications in the field of epidemiology

(Kulldorff, 2010). The first step of this method is the placement of circles of various sizes at every

municipality of the study area. For each circle a likelihood ration statistic is computed according to the

number of observed and expected cases of ChO within and outside the circle and compared to the

likelihood under the null hypothesis. The likelihood function under the alternative hypothesis

assuming Poisson distribution is proportional to

f

cc

Ic

c

c

c

)()(

Where c and E(c) are the observed and the expected number of cases in a circle, (N-c) and (N-E(c))

are the observed and expected number of cases outside the circle, N is the total number of cases and If

is the indicator function with value equal to 1 if γ is larger than the expected number of cases given the

null hypothesis and O otherwise. The circles with the highest likelihood ratio are signed as potential

clusters. Furthermore, the p-value, based on Monte Carlo simulation, is computed and used as an

evaluator of the spatial random distribution of the cases. For each simulation, the actual value is

compared with the set of simulated values to find the significance probability. The outputs of the

SatScan method are: a set of clusters, the relative risk of ChO for the different clusters and a

corresponding p-value for each cluster after Monte Carlo simulations. To accommodate different at-

risk groups, such as gender we proceed to the split of the population into the proper strata. In this

608

study we assumed primary and secondary clusters as long as their corresponding p-values were less

than 0.05 after 999 Monte Carlo simulations for each dataset.

Finally, an analysis of selected socioeconomic variables for ChO clusters was carried out. At this stage

a limited number of such variables (e.g. educational level, sectors of economic activity, occupations)

were taken into account for each municipality were ChO clustering occurs.

4. RESULTS AND DISCUSSION

As mentioned previously, spatial scan analysis quantifies the similarity of ChO rates among spatially

related geographical units. After applying the spatial scan method for overall obesity for 2007 (map

fig.2 /obese/total) a geographical area located in the western part of the study area was identified as the

most significant hot spot cluster (Chaidari, Keratsini, Korydallos, Nikaia, Ag. Varvara, Egaleo

municipalities). Furthermore, spatial scan analysis of childhood overweight rates identified a hotspot

(maps fig.2/overweight/total/male/female) in the south part of the study area (Glyfada, Argyroupolis,

Elliniko municipalities).

The same analysis for 1997 reflected similar – though more solid - clusters compared to 2007 analysis.

The obesity cluster is extended covering all the western part of the study area. Moreover, the

overweight cluster has a more scattered geographical form which consists of two pockets in the west

and east part of the study area. The spatial analysis of ChO rates across Athens metropolitan area

shows minor differences between male and female rates.

Figure2. Spatial patterns of ChO and overweight (male, female, total) for 1997 and 2007 according to

Kulldorff’s spatial scan analysis.

609

The results of the proposed analysis show significant high obesity clustering in Egaleo, Ag.Varvara

Chaidari, Korydallos, Keratsini and Nikaia municipalities (in the south-west part of the study area).

Moreover significant results with high overweight clustering depicted in Glyfada, Elliniko and

Argyroupolis municipalities (in the south part of Athens metropolitan area).

Focusing on these clusters and looking for socioeconomic characteristics such as the employment

status and the educational level we realized that there is a significant although complex relationship

between these features. The employment status was structured along two classes, the A class

occupations included jobs with medium to high income, and people engaged in scientific professions

and managerial positions. On the other hand the B class occupations included jobs with low to

medium income and people employed in technical occupations and services. In the municipalities

where high obesity clustering occurs (Egaleo, Ag.Varvara Chaidari, Korydallos, Keratsini and Nikaia)

the analysis revealed high proportions of people employed in the B class occupations which implies a

lower socioeconomic status and low educational level (figure 3). The socioeconomic situation of the

overweight cluster (municipalities of Glyfada, Elliniko and Argyroupolis) is similar to the average of

the total area under investigation with relatively high proportions of A class occupations and

significantly improved educational level.

Figure3. Spatial patterns of total ChO and overweight in 2007and socioeconomic characteristics

The findings concerning the significant clustering of ChO may be related among others to the

relatively low level of sports and recreation infrastructure in these areas, where specific employment

categories such as construction workers, technical assistants, service workers and salesmen, and office

clerks are concentrated. The low capacity of these municipalities to cover the sports and recreation

needs of the children in combination with the low education and socioeconomic status of their families

seems to have created an “obesogenic environment” in the late 90s.

The relatively better situation (i.e. lower concentration of obesity) in 2007 maybe related to the

improvement of infrastructures supporting physical activities in those municipalities in the decade of

Obese Cluster

Overweight Cluster

610

2000, due to the economic growth observed in the wider city. At the same time the significant

presence of foreign population (i.e. economic immigrants) in those areas may have smoothened the

ChO pattern which was formulated a decade earlier.

As for the south newly formulated cluster of overweight children in the 2007 this may be related to the

relocation of low to middle income strata to those areas due to their search for better housing

conditions in an already higher status area.

In the future, ChO data might be analysed within GIS context in cooperation with other geographical

variables in order to distinguish important correlations with additional environmental and

socioeconomic data. Moreover, the exploratory geospatial research of ChO/overweight in more detail

level could contribute to investigate potential spatial clustering in micro scale (neighbourhood level).

5. CONCLUSIONS

The analysis of ChO spatial clustering was carried out with the application of spatial scan analysis.

This analysis reveals persistent clustering of ChO across Athens metropolitan area. The examination

of the ChO spatial patterns for the study area shows that there is a coherent spatial cluster of obesity in

the southwest part while the most significant overweight cluster is located in the south part of Attica.

The first findings showed that there is a significant relationship between ChO and socioeconomic

heterogeneity in Athens metropolitan area. Therefore, it seems that ChO clustering is evident in areas

with high concentration of people with low socioeconomic status and education level. Further research

is required to analyse how the socioeconomic and environmental characteristics interact as driving

mechanisms for the clustering of ChO.

REFERENCES

Anselin, L. (2003), GeoDa 0.9 User’s Guide, Urbana-Champaign, IL, Spatial Analysis Laboratory

(SAL), Department of Agricultural and Consumer Economics, University of Illinois

Brunsdon, C., Fotheringham, S., and Charlton, M., (1998), Geographically weighted regression.

Journal of the Royal Statistical Society: Series D (The Statistician), 47, 431–443.

Cao Z D, Zeng D J, Zheng X L, et al., (2010), Spatio-temporal evolution of Beijing 2003 SARS

epidemic. Sci China Earth Sci, 53, 1017–1028.

Chaikaew N., Tripathi N.K., and Souris M., (2009), Exploring spatial patterns and hotspots of diarrhea

in Chiang Mai, Thailand, International Journal of Health Geographics, 8:36.

Cole, T., Bellizzi, M., Flegal, K., and Dietz, W., (2000), Establishing a standard definition for child

overweight and obesity worldwide: international survey. BMJ, 320, 1240-1243.

Cole, T., Flegal, K., Nicholls, D., and Jackson, A., (2007), Body mass index cut offs to define thinness

in children and adolescents: international survey. BMJ, 335, 194-198.

Fotheringham, A.S., Brunsdon, C., and Charlton, M., (2002), Geographically Weighted Regression:

the analysis of spatially varying relationships, (Chichester: John Wiley and Sons).

Georgiadis, G., and Nassis, G., (2007), Prevalence of overweight and obesity in a national

representative sample of Greek children and adolescents. Eur J Clin Nutr, 61, 1072-1074.

Jacquez, G.M., Greiling, D.A., Kaufmann, A.M., (2005). Design and implementation of a space–time

intelligence system for disease surveillance. Journal of Geographical Systems, 7, 7–23.

James, W.P.T., (2008), The epidemiology of obesity: the size of the problem. J Intern Med, 263, 336-

352.

Kulldorff, M., (1997). A spatial scan statistic. Communications in Statistics—Theory and Methods 26,

1481–1496.

Kulldorff, M., (2001), Prospective time periodic geographical disease surveillance using a scan

statistic. J R Stat Soc Ser A, 164, 61-72.

611

Langford, I.H., Leyland, A.H., Rasbash, J., and Goldstein, H., (1999), Multilevel modelling of the

geographical distributions of diseases. Journal of Royal Statistical Society Series C, 48, 253–268.

Lim S., Hellinski B., Lee H., (2010), Geocorrelation of Contributing Factors to Asthma in New South

Wales, International Journal of Geoinformatics, 6(4), 41-47.

Martorell, R., Kettel, K.L., Hughes, M.L., and Grummer-Strawn L.M., (2000), Overweight and

obesity in preschool children from developing countries. Int J Obes Relat Metab Disord, 24, 959-967.

Openshaw, S., (1995), Developing automated and smart spatial pattern exploration tools for

geographical information systems applications. Statistician, 44, 3–16.

Rushton, G., and Lolonis, P., (1996), Exploratory spatial analysis of birth defect rates in an urban

population. Statistics in Medicine, 7, 717–726.

Tambalis K.D., Panagiotakos D.B., Kavouras S.A., Kallistratos A.A., Moraiti I.P., Douvis S.J.,

Toutouzas P.K., & Sidossis L.S., (2010), Eleven-year Prevalence Trends of Obesity in Greek

Children: First Evidence that Prevalence of Obesity Is Leveling Off, Obesity, 18-1, 161–166.

Williams J, Wake M, Hesketh K, Maher E, Waters E., Health-related quality of life of overweight and

obese children. JAMA 2005; 293:70-76.