—from Epidemiological Bulletin, Vol. 23 No. 4, Diciembre 2002—

Development of the Healthy Condition Index Using
Geographic Information Systems in Health

One of the imperatives in health is the reduction of inequities between population groups or geographical areas within a country or region.(1) The identification of those population groups that present greater unmet health needs is an essential public health function and its systematic fulfillment will make it possible to monitor the health situation and orient policies and health programs aimed at eliminating or reducing such inequalities in health.

Health needs in a geographical area or population group are usually characterized by variables and indicators that represent different dimensions. Unmet health needs are not only expressed through morbidity and mortality indicators; because they are related to health determinants, socioeconomic indicators and indicators of resources, access, and health services coverage are also an important source of information for the evaluation of such needs (social indicators approach).(2,3) In this regard, the need is expressed as an injury or risk for health, or as a deficiency. In contrast, when looking at health in terms of quality of life,(4) where more weight is given to the determinants than to the consequences of the disease, a healthy situation will arise where better living conditions, availability of resources and greater well-being exist.

In order to facilitate the allocation of resources, it is sometimes necessary to synthesize the information into a single index. That index should take into account the different aspects of the health situation, be simple to calculate and consider the distribution of all groups, including geographical patterns. Following these requirements, this article presents a procedure to calculate a simple healthy conditions index that may be used to guide decision-making in health, including the use of a geographic information system developed at PAHO.

Materials and methods
National-level health indicators from PAHO’s Core Health Data System were used for these analyses.(5) For examples at the subnational level presented here, data from the first subnational level (e.g. states, provinces or departments) were used, as published, following the recommendations on standards, by the countries in their Basic Indicators brochures and mentioned in PAHO’s Annual Report of the Director for 2000.(6)

The cartographic databases of the national and first subnational level used to display thematic maps of the Americas come from the Digital Atlas of the World(7) and were edited by PAHO;(1) these maps are presented at a scale of 1:100,000,000. The georeferenced data were processed and analyzed with the computer package SIG-Epi.(8) Thematic maps using choropleth ranges were prepared to describe the geographical distribution of the variables, including quantiles as a method of classification of groups (quintiles). Different data layers were superposed to show their spatial distribution and relation. SIG-Epi’s Composite Health Index calculation tool was used for processing the georeferenced data. The results were grouped in accordance with an ordinal scale showing favorable, average and poor health conditions.

The selection of indicators for the healthy conditions index was based on the following criteria: 1) availability for all the countries of the Region of the Americas; 2) representativeness of various dimensions of health; 3) accepted validity; 4) generated by routine information systems; and 5) with sufficient variability to discriminate between situations. Following these criteria, indicators were included that reflect a healthy environment and infrastructure (population coverage of water and of sanitary disposal of excreta), community development (percentage of urban population), availability of health resources (physicians per population), and access to health services (coverage of vaccination in infant population). Other indicators of human capital development (literacy) and well-being (life expectancy at birth) were included to complement the index. Some of the indicators included in the index were not available at the subnational level; consequently, alternative indicators representing the appropriate dimensions of each area of analysis were used to exemplify the procedure, including population growth rate, total fertility rate, frequency of low birthweight and infant mortality as “proxy” indicators for community development, access to health services and for well-being, respectively.
Once the indicators are identified, it is necessary to define how to combine indicators with different units of measure to calculate a unique standard index. Different procedures can be applied, but a simple and statistically robust one consists of standardizing all the units into a single scale. To this end, Z-scores2 were calculated. Z-scores represent one of the most commonly applied methods used in measuring and characterizing individuals with regard to their populations. They measure the distance between the value of an observed unit and the distribution average, which would represent an expected achievable level.(9) This approach is best known for its use in the evaluation of the nutritional status of preschool children.(10)

The average and the standard deviation of a frequency distribution of the population are required to obtain the Z-scores. In the present study, the Z-score of each geographical unit for each indicator was calculated as the difference between the observed and mean values, divided by the standard deviation, as follows: Z=(Xi-X)/S, where Xi is the observed value, the average and S the standard deviation. The healthy condition index (HCI) for each geographical unit was obtained by calculating the algebraic sum of the Z-scores for each indicator, as follows: HCI = Z1 + Z2 + Z3 + ..... + Zn.

Finally, the results of the sums were ordered and re-classified in quantiles in order to identify the geographical or population groups with the highest total scores, that is, those with the best health conditions. The results of the indices were presented in thematic maps.

Results
Thirty-nine of the 48 countries and territories of the Region were included in the analysis, corresponding to those with complete information on basic indicators for the period 1995-2000. The majority of unavailable data came from countries and small territories of the Caribbean.

As shown in Table 1, there exist large differences in the distribution of the values of the indicators composing the Healthy Condition Index (HCI) in countries of the Region. For example, the weighted average of the number of available physicians per 10,000 population in the Americas was 19.8; however, the ratio between the maximum and minimum values was 32 times greater in the most favorable situation than in the least favorable. In contrast, the weighted average of measles vaccination coverage in children under 1 was 92.5%, and the same ratio 1.3 times greater in the most favorable situation than in the least favorable. Important differences were also detected in the indicators at the subregional level. The highest values, except for the availability of physicians by population, are consistently found in North America, while the lowest ones tend to occur in Central America.

In the maps showing the distribution of indicators, some countries of the subregions consistently present the most favorable conditions. For example, life expectancy at birth exceeds 75 years (higher two quintiles) in 19 of the 48 countries (Figure 1). This group includes the United States and Canada in North America, half of the countries of the Caribbean, while in Central America and the different subregions of South America, only Costa Rica and Chile, respectively, have reached that level. The countries furthest from this level, which still do not reach 66 years of age, are Haiti, Guyana, Bolivia, Grenada, Honduras, and Guatemala. With regard to literacy, the percentage of literate population exceeds 95.5% in 20 countries in the Americas (Figure 2). The greatest percentages are found in Canada and the United States in North America, Costa Rica in Central America, Argentina, Chile, and Uruguay in the Southern Cone, Guyana, Cuba, and other countries in the Caribbean. In that same subregion, the situation is more precarious in Haiti, while in Central America, El Salvador, Guatemala, Honduras, and Nicaragua present the smallest fraction of literate population with less than 75%. A similar distribution is seen for the coverage of drinking water. While 20 countries of the Region show levels higher than 93%, 7 have not exceeded 75% (Figure 3). The availability of physicians is higher than 14 per 10,000 population in 18 of the countries (Figure 4), while in 18 other countries the coverage does not reach 10 physicians per 10,000. Unlike other indicators, the least favorable situation is found in little-populated countries of the Caribbean, in addition to Belize, Guatemala, Honduras, and Nicaragua in Central America and Bolivia and Paraguay in South America.

The HCI synthesizes all the indicators for countries of the Region, their ranking and the distance with respect to the regional values (Figure 5). The most favorable conditions, with higher values of the HCI, are found in Canada, the United States and Mexico in North America; Costa Rica and Panama in Central America; Barbados, Cuba, French Guiana, Trinidad and Tobago in the Caribbean; Venezuela in the Andean Area; and Argentina, Chile and Uruguay in the Southern Cone. In contrast, the countries with lower values, suggesting more important needs, include Haiti, Guyana and Suriname in the Caribbean; Guatemala, Honduras, El Salvador, Nicaragua in Central America; and Bolivia and Paraguay in South America.

Within countries, the determination of healthy conditions follows the same heterogeneous mosaic pattern observed in the countries of the Region. The situation of Mexico is presented here to illustrate this aspect (Figure 6). Although some states in the south of the country generally show the lowest levels of all the indicators presented, this low level of the HCI is also found in other geographical units in the center of the country, potentially indicating low levels of some indicators and a more important deviation of the Z-score with respect to the rest of the states.

Finally, although values were not available for all the indicators to calculate the subnational level HCI, a map was prepared showing healthy condition indices for the countries of the Region, calculated with alternative indicators. This allowed identification of areas with less favorable health conditions (Figure 7). In the Region, various areas present less favorable conditions, that form “foci” or clusters of geographical units. In these groups, the probability of having a low HCI increases when neighbors have similar values. This is an important aspect for the stratification of areas according to their determinant factors and for planning of activities, particularly because many of the less favorable areas are in the borders between countries. This trend towards aggregation is also observed for the most favorable health conditions, but is less frequent in the border areas. It is important to note that the observed differences are based on the calculation of the HCI at the country and not the regional level. This means that low values are not comparable in different countries; however it makes it possible to identify areas where the countries themselves should intervene.

Discussion and general comments
The characterization of healthy places and the monitoring of their conditions are considered essential elements to orient efforts to reduce inequalities in health through additional care and health promotion. The use of indices that summarize information on various dimensions of health development has been suggested in order to make the process of decision-making more efficient. Moreover, it has been recommended to include computerized technological tools, such as geographic information systems (GIS), that facilitate the management, visualization, query, and analysis of data. This article presents the use of some positive indicators (i.e. not based on health impairments) and a procedure for standardizing the calculation of indicators, using the SIG-Epi system.

The results displayed in thematic maps showed that there exists a great variability in the distribution of the health conditions in the Region of the Americas at the country level and within the same. Greater consistency was also achieved in the identification of areas of more or less favorable health conditions through the use of the HCI, which would not have been achieved using a single indicator (see for example Figures 1, 4 and 5). Using the same group of indicators for the HCI, presented at the subnational level, illustrates that it is possible to identify with greater precision areas in which to target health efforts. The presence of contiguous unfavorable areas on the borders suggests the possibility of promoting efforts in such areas, using binational resources. This has been occurring in different subregional initiatives of health promotion, control and surveillance of health problems in the Southern Cone, the Andean Area, and Central America.

Specific numbers and types of basic health indicators routinely collected were used here, representing different dimensions of the health process. These indicators include important health determinants, and their specific analysis allows identification of areas of intersectoral work. The process of production and collection of basic data at the national level was initiated in at least 23 of the 48 countries and territories of the Americas, and efforts have been put in place to standardize their contents, as in the case of basic indicators of Central America (available as of 2003). At the subnational level, where the information did not exist, other alternative indicators were used, that represent the same dimensions of health. This implies that it is not indispensable (although recommendable) to dispose of the same basic indicators to calculate a healthy condition index. However, it is suggested to keep the set of indicators used in the calculation to the minimum necessary. It is also important to mention that in general, the use of different indicators provides results with a similar hierarchy, which indicates the consistency of the method. The selection of indicators for the Healthy Condition Index was based, among other reasons, on the availability of basic indicators for all the countries of the Region that represent various health dimensions, that were accepted for their validity, and generated by routine information systems. Before generating such an index and interpreting its results, it is recommended to carry out an exploratory analysis of the distributions of the indicators to be used and of its correlations, in order to select those that meet the minimum requirements mentioned previously.

There exist different procedures to assign scores, from the use of cut off points to methods based on regression. In 1996, PAHO utilized a procedure to identify healthy spaces.(11) The approach consisted of assigning a score of zero or one to the units of analysis in relation to the fulfillment of a criterion, which was to belong to the 3 superior quintiles of the frequency distribution of the indicator. This approach has the limitation that it does not account for the magnitude of the deviation of the group value with respect to the reference used, but only if this deviation exists or not. It can be mentioned that other more complex and precise statistical methods based on regression or principal component analysis also exist. However, both are computationally more complex and the former in particular requires the definition of weights for the variables. In contrast, the Z-score used in the present approach presents the advantage of utilizing the full distribution of values for the analysis and not only the extreme values or defined criteria. The Z-score also measures the relative distance between a unit of analysis and the average of the distribution, which would represent an achievable level in the absence of inequality. Other additional advantages of the procedure are the simplicity of its calculation and its additive and comparison properties.

Different computer packages may be used to carry out this analytical process. The epidemiological package for tabulated data EpiDat(12) makes it possible to calculate a composite health index. However, in order to incorporate the dimension of the spatial distribution of the indicators, the geographic information system SIGEpi(8) was used on this occasion, in particular the Composite Health Index calculation tool.

In short, from the perspective of health promotion, the detection and evaluation of healthy conditions is a critical step for the definition of priorities and intersectoral work in health, including the adjustment of health services. The use of the synthetic index proposed here, based on basic health indicators in the context of a GIS, facilitates the analytical process, allows for the identification of “foci” or population groups in less favorable conditions and, through this, orients the formulation of adequate health plans and programs. When the units of analysis are smaller, the results are more specific and useful for decision-making. For this reason, it is recommended to promote the collection and use of disaggregated information at the level of municipalities, taking into account the fact that in most countries, they correspond to the operational units for health services.

References:
(1) Pan American Health Organization. Annual Report of the Director 1998. Health Situation in the Region of the Americas. Washington, DC. PAHO; 1998.
(2) Castillo-Salgado C. Los servicios de Salud en las Américas: Análisis de indicadores Básicos. Cuaderno Técnico no 14. Organización Panamericana de la Salud. Washington DC, 1988: 147-152, 221-230.
(3) Siegel LM, Attkinson CC, Carson LG. Need identification and program planning in the community context. Attkinson CC et al. (eds). Evaluation of Human Services Programs. New York:Academic Press, Inc;1978:226-227.
(4) Organización Panamericana de la Salud. Municipios y comunidades saludables. Guía de los alcaldes para promover calidad de vida. Washington, DC. OPS; 2002.
(5) Pan American Health Organization, Special Program for Health Analysis. Core Health Data Regional Initiative; Technical Health Information System [Internet site]. Available at: http://www.paho.org/English/SHA/coredata/tabulator/newTabulator.htm. Accessed on 15 July 2002.
(6) Pan American Health Organization. Advancing the People’s Health. Annual Report of the Director - 2000. Washington, DC: PAHO, 2000.
(7) Environmental Systems Research Institute. ESRI Data and Maps. Redlands: ESRI, Inc. 2000.
(8) Martínez R, Vidaurre M, Nájera P, Loyola E, Castillo-Salgado C. SIGEpi: Geographic Information in Epidemiology and Public Health. PAHO’s Epidemiological Bulletin 2001; 22:3.
(9) Sentís J, Ascaso C, Valles A, Canela J. Bioestadística. Serie de Manuales Básicos para Licenciatura y Residencia. Barcelona: Salvat Medicina. 241 p.
(10) WHO Working Group. Use and interpretation of anthropometric indicators of nutritional status. Bull WHO 1986; 64:929-941.
(11) Pan American Health Organization. Health Situation.Healthy People, Health Spaces. Annual Report of the Director - 1996. Washington, DC: PAHO, 1996.
(12) Dirección Xeral de Saude Publica, Xunta de Galicia and Special Program for Health Analysis, PAHO/WHO. EpiDat. Paquete de Análisis de Datos Tabulados. Versión 3.0. Santiago de Compostela: Xunta de Galicia, 2003.

Source: Prepared by Drs. Carlos Castillo-Salgado and Enrique Loyola of PAHO’s Special Program for Health Analysis (SHA), from a presentation at the Health Promotion Forum of the Americas organized by PAHO and the Ministry of Health of Chile, Santiago, Chile, 20-24 October 2002.

Return to Index
Epidemiological Bulletin, Vol. 23 No. 4, Diciembre 2002