-from Epidemiological Bulletin, Vol. 22 No. 3, September 2001-

Use of SIGEpi for the Identification of Localities Vulnerable to
Environmental Risks in Mexico

Problem under study
In various areas of Central Mexico, in the state of Querétaro in particular, a group of localities with high levels of marginalization (1) can be found. They are potentially vulnerable to the influence of extreme environmental risks or events derived from land dynamics of the region. These risks fall into several categories, depending on their origin (2): 1) Hydrometeorological risks: disasters derived from severe droughts or high levels of rainfall; 2) Geological risks: proximity to active fault lines, with regard to differential tectonic movements, or terrain subsidence; 3) Geomorphologic risks: Dynamics of different relief types, some prone to erosion and in others, flooding and sedimentation; 4) Chemical risks, resulting from human activity: proximity to electric lines or pipelines transferring hazardous materials, such as gas or oil.

The vulnerability of a locality to environmental risks is understood as the occurrence of emergencies or demands that exceed the capacity of the health sector to respond (3). Lack of resources or distance to infrastructure and services increase the problem. Indeed, a large number of communities do not have health care centers in close proximity, are far from principal transportation routes, and lack sufficient resources in case of environmental disaster, the latter due to the precarious and temporary nature of the materials used to build their dwellings.

In order to enhance the measurement of vulnerability, a five-level socioeconomic “marginalization” index officially used in Mexico, was adopted as a synthetic measure. This index was constructed using the method of principal components, and is used to differentiate states, municipalities and localities according to the global impact of deficiencies in access to elementary education, conditions and size of the dwellings, population distribution and an income insufficient to acquire the basic basket of food (4).

The magnitude and distribution of vulnerability in central Mexico is partially known in different sectors of the public administration. However, given the diversity of information sources and the lack of adequate methodological tools, it is difficult to prepare a comprehensive diagnosis including environmental, social and economic data. The data associated with the health system’s response capacity are particularly difficult to include in the diagnosis.

As a first approach to the problem of identifying, quantifying and locating areas exposed to environmental risks in the state of Querétaro, an application was developed using geo-processing and statistical/spatial analysis tools available in SIGEpi. Production took into account marginalization, access to protective factors, and the presence of environmental risks, and was carried out under the operational model shown in Figure 1. It is important to point out that this framework was the basis for the construction of criteria, classification of variables and analysis of its relations.

Methods
1. Sources of Information
Various layers of digital cartographic information and attributes were compiled from different sources. They were incorporated into the project by their geometric characteristics: point files (localities (5), with or without health services); lines (roads, pipelines (6)) and polygons (municipalities (7), layers of environmental risks (8,9)). Environmental risks were defined in relation to the criteria established by the agencies that provided the information.

Base maps were prepared previously using the Universal Transverse Mercator projection (UTM), zone 14Q, where the state of Querétaro is located, in order to carry out the measurements of areas and distances with greater precision (b). To simplify geo-processing and structuring of the attribute table and data analysis, two separate layers were formed using information on localities in each state: those considered as headquarters for public health services and the remaining localities without public health services. There are a total of 2,112 localities of different sizes in the 18 municipalities of the State of Querétaro, but 28.3% are communities with two dwellings or less, for which no population nor socioeconomic information exists. For purposes of this study, these were eliminated from the analysis.

2. SIGEpi tools used for processing and analysis
Creation of point layers
Through this procedure, the geo-referencing of the clinics and hospitals not belonging to the Ministry of Health of the State of Querétaro (10) (SESEQ) was carried out, based on the geographical coordinates of the headquarter localities (11,12,13). These health service units were added to the services of the SESEQ in a single layer in order to cover all public health services available in the state of Querétaro.

Creation of areas of influence
Influence areas were set up to identify catchment area reach of two types of phenomena: 1) impact areas of environmental risks, such as fault lines, or those of electric lines (200 m) or gas/oil pipelines (one km), and 2) an area 5 km from both health centers and transportation routes available all year, according to PAHO’s definition (14).

Access to services (radial schemes)
With this tool, the shortest linear distance was drawn between a central point (health centers), and satellite or peripheral localities.

Selection by attributes and by layers
These techniques were applied to place localities in relation to environmental risks, or influenced by the protection of services and infrastructure. Through this process and geoprocessing, environmental risk zones were geographically delimited.

Criteria used to select localities with protective factors included the following conditions (c): percentage of the population with social security (median value was 0%); dwellings with some floor covering (the median value was 26.5%) and level of urbanization (more than 10,000 inhabitants). Also, they had to fulfill some categorical value, such as whether they were a municipal government center or not.

In the same way, a selection by layers identified localities within areas of influence or risk, or localities intersecting areas near electric lines or fuel ducts. Once identified, dichotomous values (0,1) were assigned to those localities meeting the conditions established within the operational framework.

Frequency distribution and exploratory data analysis
Exploratory data analysis was carried out for the variables under analysis. Median values were calculated in order to establish cut-off values for the selection of critical localities.

Identification of critical areas (critical localities)
Localities were classified depending on an expected (cut-off) value. The dichotomous score assigned to the communities was added in order to generate an ordinal scale, ranging from 0 to 4, depending on the number of exposure factors.

A similar process was followed for protective factors, with criteria such as the proportion of dwellings built with permanent materials, proximity to roads, or social security coverage; other criteria included the availability of health services located within 5 km and classified by levels of care: hospitals, clinics, and health centers.

Subsequently, we selected localities of greater vulnerability that fulfilled all conditions at the same time, i.e., exposure to risk factors, lowest level of protective factors and greatest marginalization (based on the original official classification).

Spatial autocorrelation
This method is used to determine if an indicator’s value shows a tendency to form geographical clusters or if its distribution is random. Moran’s global autocorrelation index (I) is calculated for a global test to determine the existence of significant clusters in the distribution of data without indicating where it is located.15 With the local test, it is possible to identify the location of the clusters.

For this particular study, Moran’s global I was calculated for the following variables: exposure to environmental factors, access to protective factors, and marginalization, using a radius of 5 km as criterion for vicinity.

Results
The limit, extent, and overlap of environmental risk areas are shown in Map 1. Nine municipalities were found with historical records of drought, the majority located in the Sierra Gorda (Northeast of the state). Small areas with reports of extreme rains and floods were found in the Southern and Northeastern end of the state. Active faults concentrations also were detected in the area of the Sierra Gorda (NE) and in the Neovolcanic Axis range located in the Southwestern extreme of the State. Geological risk zones (areas of subsidence and tectonic movements) also were found to the Southwest of the state. Furthermore, areas of environmental changes due to human activity were identified, such as areas close to high-voltage lines or pipelines of hazardous materials (gas and oil) in the zone known as the industrial corridor of San Juan del Río – Querétaro.

In contrast, the distribution of protective factors, as seen in Map 2, includes both areas with roads and health care units, classified by health care capacity (hospitals, clinics and health care centers). These catchment areas are delineated within a distance of 5 km, equivalent to a hour-long walk on flat terrain. In addition, parts of the state not covered by this infrastructure are shown. The distances between localities and health centers greater than 5km are highlighted by the radial schemes.

Among the 1,447 communities with socioeconomic information, 172 have health units with different levels of capacity - hospitals, clinics, and health centers - and 1,275 do not, the latter having an average population of 254 inhabitants (range from 5 to 3,392 inhabitants).

Of the 1,275 localities under study (without health care units), we find that 1,110 (91.5%) are exposed to at least one risk factor in the surrounding environment, either natural or human-made. In only 7 of the analyzed localities, do all four environmental risk elements appear together (Table 1).

Table 1: Frequency Distribution for a Selection of Variables
Environmental Exposure
Absolute Frequency
Relative Frequency (%)
Cumulated Frequency (%)
0.0 121 9.5 9.5
1.0 506 39.7 49.2
2.0 610 47.8 97.0
3.0 31 2.4 99.5
4.0 7 0.6 100.0
Protective Factors
Absolute Frequency
Relative Frequency (%)
Cumulated Frequency (%)
0.0 3 0.2 0.2
1.0 71 5.6 5.8
2.0 74 5.8 11.6
3.0 233 18.3 29.9
4.0 350 27.5 57.3
5.0 460 36.1 93.4
6.0 37 2.9 96.3
7.0 47 3.7 100.0
Marginalization Values
Absolute Frequency
Relative Frequency (%)
Cumulated Frequency (%)
1.0 44 3.5 3.5
2.0 76 5.9 9.4
3.0 231 18.1 27.5
4.0 298 23.4 50.9
5.0 626 49.1 100.0

 

With regard to protective factors, 849 localities representing 42.7% of those under observation register spatial concurrence in at least four of the seven existing levels. It is notable that only 47 communities (3.7%) have all protective factors.

Finally, we find that 626 localities (49.1% of the total without health services) are classified as communities of high and very high marginalization, defined as values of 4 and 5 according to the official categories use by the Mexican agencies CONAPO-PROGRESA.

The map classifying localities by environmental exposure level, identified several areas in the state where various risk factors coincided (map 3). Areas of interest include the industrial corridor of San Juan del Río Querétaro (to the Southwest of the state), with four factors; Amealco (to the South) with three coincidental risk factors; vicinity of Jalpan in the Sierra Gorda (Northeast of the state) with combinations of two environmental factors; and Peñamiller and Cadereyta, in the center of the state.

Additional thematic maps were constructed to analyze the regional distribution of localities under certain conditions of protection and marginalization (not included).

Based on a classification of protective factors with an ordinal scale, it was possible to recognize groups of localities with high values as determined by the weight of the classification of health care units. The highest values appear to the South Southwest of the state, coinciding again with the industrial corridor of San Juan del Río – Querétaro. The following lower values are found around the other central localities (government headquarter and/or urban).

In addition, the analysis of the regional distribution of localities with high marginalization showed an important concentration in the Northeastern area of the State-corresponding to the Sierra Gorda – however, other concentrations are found toward the periphery, in areas far from the principal lines of communication.

The statistical significance of Moran’s I for risk exposure is showing that its distribution is not random and that the exposure values tend to concentrate in certain places of the state. It also shows that there exists groups of neighboring localities with similar values of exposure (Table 2), within the 5km limit. The levels of protection and marginalization show a clustering of neighboring communities, similar to the environmental exposure.

Table 2: Global Spatial Correlation Indices for Factors Associated to Vulnerability
Moran's I (5km)
Calculated Value of I
Expected Value of I
Standard Deviation
Z score of I
Significance (p)
Environmental Exposure 0.7154 -0.0008 0.0146 49.1277 0.00000
Protection 0.5556 -0.0008 0.0146 38.1662 0.00000
Marginalization 0.4184 -0.0008 0.0146 28.7524 0.00000

 

As a result of applying criteria analysis to synthesize the three groups of factors, a group of 379 critical localities were identified, where 55,083 people (4.4% of the state population) live, under greater environmental exposure, very little protection and high marginalization. Map 4 shows the elevated concentrations of these localities to the Northeast of the state, in the Sierra Gorda zone.

For purposes of estimating population and determining necessary resources to serve the populations in each health jurisdiction, the total of all localities in each administrative unit was calculated. The lowest concentration of critical communities is located in health jurisdiction I (Southeast), with 9 vulnerable localities and 953 inhabitants representing 0.1% of the jurisdiction’s population (Table 3). At the other extreme, jurisdiction IV (Northeast or Sierra Gorda), registers 242 vulnerable localities and 33,993 inhabitants (42.2% of the jurisdiction’s total population).

Table 3: Distribution of Critical Localities in each Jurisdiction
Jurisdiction
Critical Localities
Vulnerable Population
Total Population by Jurisdiction
Vulnerable Pop./Total Pop. (%)
# First-level Units
# Hospital Units
Vulnerable Pop./Unit Level Ratio
I. Querétaro 9 953 706,566 0.1 50 4 19.1
II. San Juan del Río 7 2,579 340,821 0.8 58 4 44.5
III. Cadereyta 121 17,558 122,503 14.3 49 3 358.3
IV. Pinal de Amoles 242 33,993 80,586 42.2 35 1 971.2
State Total 379 55,083 1,250,470 4.4 192 12 286.9

In absolute terms, we observe that the jurisdiction with the highest level of development and largest population (I) has a very small vulnerable population. In contrast, the health jurisdiction with the smallest population, fewest resources, and lowest level of development (IV) shows a highest number of vulnerable localities and population.

The health services response are limited by the availability and specialization of health care facilities. The ratio of vulnerable population per unit of first level care is 50 times higher in Jurisdiction IV that in Jurisdiction I. Also, the number of available care units with high specialization is lower in Jurisdiction IV.

Conclusions
Within a vulnerability analysis framework, tools such as SIGEpi allow the integration of measures and indicators from different sources, and place them in a common space for statistical and geographical analysis. Using this, it is possible to delineate natural hazards in a geographical region, approximate the scale of situations requiring response capacity which exceeds that of the health services and accordingly, evaluate some approaches to mitigate the vulnerability of populations and infrastructures exposed to environmental risks and disasters. This analysis is necessary to support and direct the decision-making process on priorities and interventions. Although not an exhaustive analysis, factors related to risk exposure were weighted according to their potential impact. This allows both the recognition and ability to take advantage of those procedures that determine risk. To this end, traditional univariate and multivariate analytical tools were used, including the value of spatial perspective.

Prospects for SIGEpi
With many issues concerning requirements and needs still facing GIS applications in public health, the solutions presented by SIGEpi, through its analytical tools for epidemiology and public health, open a favorable perspective for this GIS package.

Prior to its launching, SIGEpi, currently in its Beta version, has been tested by various Latin American and Spanish institutions. Their suggestions are being incorporated into the program, and a series of functions still have to be incorporated into the package in the near future. Overall, the design of this SIGEpi has followed a systematic and evolutionary development whereby corrections, suggestions, and observations from internal and external reviewers have been incorporated.

The distribution of SIGEpi will be done by interinstitutional agreements between SHA/PAHO and health/academic institutions interested in its use for diagnosis and evaluation projects, or research in the area of public health and epidemiology. For more information contact Dr. Carlos Castillo-Salgado, Special Program for Health Analysis, PAHO; E-mail sha@paho.org.

Notes:
(b) The projection module has not been incorporated in the Beta version of SIGEpi.
(c) The cutting points correspond to the median values for each variable.

References:
(1) CONAPO. http://www.conapo.gob.mx. October 2001.
(2) Sistema Nacional de Protección Civil. http://www.proteccioncivil.gob.mx. October 2001.
(3) PAHO. Mitigación de desastres naturales en sistemas de agua potable y alcantarillado sanitario. Guías para el análisis de vulnerabilidad. Washington, D.C.; OPS, 1998:110.
(4) CONAPO-PROGESA. CD ROM/La Marginación en México, 1998. INEGI. Conteo de Población y Vivienda, 1995. Actualización de los datos del Censo Nacional de Población y Vivienda, 1990.
(5) INEGI. Conteo de Población y Vivienda, CD ROM / Resultados del Estado de Querétaro, 1995
(6) SEMARNAT. Página electrónica http://www.semarnat.gob.mx / Información estadística y geográfica del medio ambiente / información geográfica y biblioteca digital; http://www.centrogeo.org.mx / biblioteca_dig/. October 2001.
(7) INEGI. Conteo de Población y Vivienda, 1995. Actualización de los datos del Censo Nacional de Población y Vivienda, 1990.
(8) SEMARNAT. http://www.semarnat.gob.mx / Información estadística y geográfica del medio ambiente / información geográfica y biblioteca digital. http://www.centrogeo.org.mx/ biblioteca_dig/. October 2001.
(9) Sistema Nacional de Protección Civil. Vínculos / información y clasificación de desastres. http://www.proteccioncivil.gob.mx/index.html. October 2001.
(10) Secretaría de Salud del Estado de Querétaro. Regionalización Operativa de los Servicios de Salud del Estado de Querétaro. Dirección de Planeación. SESEQ. Documento de Trabajo, 1999.
(11) IMSS. Página Electrónica. Directorio de unidadades médicas, Querétaro. http://www.imss.gob.mx/organiza.htm/. October 2001.
(12) ISSSTE. Página Electrónica. Prestaciones, unidades médicas, distribución geográfica de clínicas y hospitales del ISSSTE en el país, Querétaro. http://www.issste.gob.mx/. October 2001.
(13) Lic. Edna Ruiz. Secretaría de Extensión Universitaria, Universidad Autónoma de Querátaro (UAQ); Octubre, 2001. Comunicación personal (edna@sunserver.dsi.uaq.mx)
(14) Organización Panamericana de la Salud. Extensión de la cobertura de los servicios de salud con las estrategias de atención primaria a la salud y participación de la comunidad. Bol Oficina Sanit Panam 1977; 83 (6):479.
(15) Moran PAP. The interpretation of statistical maps. J R Stat Soc [B] 1948;10:243-51.

Source: Prepared by Geog. Patricia Najera, Eng. Ramón Mártinez, Mr. Manuel Vidaurre, Dr. Enrique Loyola, Dr. Carlos Castillo-Salgado and Mr. Charles Eisner from PAHO’s Special Program for Health Analysis (SHA).

Return to Index
Epidemiological Bulletin, Vol. 22 No. 3, September 2001