Applying GIS to Nutrition Surveys

By Filippo Dibari, Andrew Seal and Paolo Paron

Filippo Dibari is a Food Technologist with long term community based experience in developing countries and consultancies with NGOs and UN agencies in Africa, Latin America and South East Asia. Having recently completed an MSc in Public Health Nutrition at the London School of Hygiene and Tropical Medicine. He is currently working for Valid International.

Andrew Seal is a Lecturer in International Nutrition at the Centre for International Child Health, ICH, London. He has worked as a researcher/nutritionist on refugee and emergency nutrition projects in Africa, Asia and Eastern Europe. His main research interests are in micronutrient deficiencies and emergency nutrition programmes.

Paolo Paron is a Geomorphologist and GIS Specialist with research and teaching experience in the fields of GIS and remote sensing analysis and in environmental geology. Currently a visitor at the Oxford University Centre for the Environment, he is soon joining an FAO Project based in Nairobi as a long term consultant.

We would like to thank the Ministry of Health, Angola and WFP for their support, and are grateful to the Canadian International Development Agency for funding the survey via WFP. Thanks also to MSF-B for providing the Kuito town plan used in the case study example, and for their collaboration during the survey.

This article highlights the potential benefit, resource needs and constraints of applying Geographical Information Systems (GIS) analysis to a conventional nutritional survey dataset in a developing country or an emergency context.

The article is organised on two levels. Those completely new to GIS applications should find the body of the article useful in decision making and in designing the terms of reference for initiatives in this area. For nutritionists who are already familiar with datasets management and with the basics of GIS and who have already tried to apply GIS to nutrition surveys, the step-by-step approach outlined here is supported by substantial technical detail online at http://www.ennonline.net/articles/gis/index.html

 

 

A Geographical Information System (GIS) comprises a set of hardware and software tools that help to visualise and to locate, rather than analyse, the patterns of a phenomenon. Nutrition is one of the many areas of possible application of GIS methodologies and Public Health Nutrition in emergencies has only recently discovered the potential. Almost any nutrition survey aiming to define the nutritional status in a certain area (at district, town, province, region, state, nation or continent level) can be enhanced by a GIS presentation. No major changes are needed in conducting the nutrition survey or in identifying its best epidemiological design.

UN Agencies and NGOs are already including greater use of GIS in their work. Early Warning Systems Mapping, Poverty Mapping and Vulnerability forecasts are three of the largest applications among such institutions. The skills required in data management and digital mapping are still limiting the use of such powerful tools among smaller organisations. The demand has stimulated the creation of not-for-profit companies dedicated specifically to GIS applications in humanitarian relief¹.

Even EP-INFO, probably the most popular software among public health nutritionists, has evolved in the last few years to include a GIS analysis component (Epi Map) plus a large free source variety of maps of administrative boundaries, online atlas, health risk exposure data, and others².

Why field programmes should consider using GIS

Most humanitarian workers will find the following questions familiar:

"Where is the best site to set up the next Therapeutic Feeding Centres in accordance with the prevalence of registered acute malnutrition and the distance from the available health posts?"

"We have nutrition data of this area, plus an old non digitalized (papercopy) map of the region, and a very basic GPS. How is it possible to create a map on my computer with all the information in order to show it to both the donors and the local community authorities and then use it to help make important decisions about future initiatives?"

"In order to better target the beneficiaries, how do we set up a quick decisionmaking set of criteria based on the available data coming from nutri- tional status but also administrative boundaries, traditional family clan areas, water availability, transport routes and security considerations?"

"I have driven for hours up this increasingly narrow track in search of a non - existent village in the middle of nowhere. Where exactly am I right now"

While GIS is not the "magic bullet" for such questions, the use of GIS certainly contributes tools to provide better answers. Under the mantra 'keep it simple' and equipped with basic data management skills, there are simple GIS steps which can add significant value to a nutritional analysis report. More specifically, GIS helps in decision making because it provides maps of the phenomenon of interest. The map is like the 'tip of an iceberg', visually representing the bulk of the information and datasets 'underneath'. This key concept is reflected in figure 1, where the yellow area in the map graphically represents the record and all its data, also highlighted in yellow in the table.

How data are presented can have a number of practical implications. For example, there might be the need to share a decision regarding the outcomes of a recent nutrition survey with the leaderships of a few villages in a rural area. How these data are represented will affect how well they can be shared and how much the programme can benefit from the stakeholders' contributions, independent of their education level. With this scenario in mind, figure 2 presents three different ways of representing the same dataset: a table, a histogram and a map. In this situation, the GIS map may well be the best way to communicate the available data.

Applied to nutrition, GIS analysis is potentially an extremely powerful tool for monitoring, evaluation and targeting. For example, it can reflect situations where the overall nutritional status does not change in quantitative terms, but marked changes in the spatial distribution of malnutrition occur (see figure 3). This is reflected in the theoretical scenario reflected in figure 4, which compares the distribution of child global acute malnutrition (GAM) at time 0 and 1. The prevalence of GAM is roughly the same at the two moments in time, however the spatial distribution has completely changed and therefore the appropriate areas to target have/should also. In figure 4, the areas in which the socio-economic indicators have reported a high prevalence of "very poor" people have been coloured blue. Compared with figure 3, it can be clearly seen that these blue sections do not overlap with the sections in which the highest level of acute severe malnutrition is actually located. This suggests that using socio-economic status to target geographic areas would not be a robust means to target severe acute malnutrition.

Resource implications of applying GIS

To apply GIS to a nutrition survey, extra time is demanded in two phases of the survey. A few minutes will be enough to collect geographical data using a Global Positioning System (GPS) during the field phase of the survey. However, extra time is required for cleaning and analysis of the dataset to prepare for presentation in a GIS-map. As a whole, the process does not require a high level of 'personnel time', as long as the person working with GIS is already experienced or has received handson training. The steps for conducting a nutrition survey, together with GIS-related actions, are listed in table 1.

What equipment is needed?

When applying GIS analysis in a nutritional survey, a GPS device and GIS software are needed. The GPS allows users to collect data on both locations and tracks, while the GIS software allows the display and analysis of these data combined with data derived from a nutrition survey.

GPS device

There are different kinds of GPS that can be categorised in order of complexity and cost as basic GPS, mapping/cartographic GPS, car GPS and differential GPS. The first two are quite enough for survey purposes. Three brands of GPS handsets are the most common on the market, and are considered the most reliable by the authors³. For the purposes described in this article, the models Garmin GPS or Garmin e-trex family (or similar other brand) are recommended. The choice of the model is related to the storage capacity of the waypoints and how data are downloaded into the computer.

A commercial GPS has a standard error of about 10 metres in positioning, which normally is well below the precision that a nutritional survey requires. Some GPS can measure also the altitude, which can be useful in adjusting haemoglobin cut-offs to calculate the prevalence of anaemia in a survey.

 

GPS/GIS software

Once the data have been collected by a GPS, there are several possible methods, solutions and software options for proceeding with the analysis. Here only one of them is considered.

In order to run the GIS analysis, at least three pieces of software are required. The first is necessary to download the data from GPS into the PC with the correct geographical coordinates (e.g. Mapsource⁴). The second facilitates combining the GPS data with the nutrition dataset and available maps (e.g. Microsoft Excel or Access). The third is the GIS software, which is used for spatial analysis (e.g. ArcView, EpiMap, ArcGIS). In this article, ArcGIS (ESRI) is chosen to illustrate a case study.

Guide to applying GIS to a nutrition survey

Three kinds of data are necessary for GIS analysis in the context of a nutrition survey (see figure 5):

1. Data coming from anthropometry, questionnaires, biochemical analysis, etc.
2. Waypoints collected with a GPS
3. Maps of the area obtained from different sources⁵.

The sets of data are combined within the GIS software which allows for their integration and analysis. The way to introduce waypoints and maps into the GIS can appear complex and therefore a step-by-step approach is summarised here. Reference is made to a series of boxes (1- 10), which go into a high level of technical detail and are available online at http://www.ennonline.net/articles/gis/index.html In order to highlight critical points in the procedure described in the boxes, a case study has been taken into consideration. It consists of a nutrition survey undertaken by the Institute of Child Health, London in collaboration with WFP and MSF-Belgium in Bie Province, Angola, in November 2004.

Table 1 Actions needed when including GIS in a nutrition survey
Steps for conducting a Nutrition Survey*	GIS
	Action	Specific equipment
1. Define survey objectives
2. Budget for the survey	Include a GPS, ensure software is available
3. Choose the survey design	Decide where to collect the waypoints or the tracks: i.e. centre points of the clusters, wells, roads, administrative boundaries, etc.
4. Plan for personnel, facilities and equipment
5. Select the sample
6. Develop the questionnaire	Include an area in the questionnaire for the collected GPS data
7. Pre-test the questionnaire
8. Train the personnel	Train the personnel in collecting the GPS data and filling in the questionnaire correctly	GPS
9. Standardize the anthropometric technique
10. Interview/data collection		GPS
11. Edit and code the answers	Data entry of the GPS data on relation with the answers and/or anthropometric data	GPS, PC, software
12. Tabulate the data	The prepared tables include also the GPS data	PC
13. Analyse and report the survey results	Queries regarding geographical aspects are answered, and then interpreted	PC

* Adapted from FANTA (2003)

Step 1: How to introduce the waypoints into the GIS software

The GIS analysis of nutritional data consists of four phases:

Phase 1 Geographical collection of the waypoints using a GPS in the field (see box 1 online)
Phase 2 Downloading the waypoints from the GPS into the computer (see box 2 online)
Phase 3 Adjustment of the waypoints/tracks dataset (see box 3 online)
Phase 4 Combining the geographical data with the nutritional survey data to answer specific queries (see box 4 online)

An example of the final result can be seen in figure 6.

Step 2: How to introduce a map into the GIS software

Once a map has been obtained, it must be made compatible with the GIS software. The steps to follow are:

1. Preparation of a digitalised map of your area of interest (see box 5 online)
2. Digitalisation and cleaning of the map (see box 6 online)
3. Importing the map and the GPS waypoints and tracks into the GIS (see box 7 online)
4. Georeferencing the map (see box 8 online)
5. Joining the nutritional survey data and cluster points (see box 9 online)
6. Visualising the nutritional survey data on top of a map (see box 10 online).

Key considerations when employing GIS analysis for the first time

Skills in GIS software are required as GIS software are not always user friendly and training in their use can be expensive, according to the level of knowledge. Supervision of data management by an experienced user is recommended for the first time.

Skills in managing data using different software are required. For example, it is necessary to know how to export data from EPINFO into ArcMap, passing through Microsoft Excel or Word, in order to achieve the correct format of the files or of the tables to insert into the GIS software. These operations are not complicated or time demanding per-se whenever the procedure is known. They can be so, however, if the operator has to find the way to do it on his /her own.

Choosing which GPS to use is important. Certain GPS do not allow the user to download their coordinates into the computer, or require additional equipment/manual data copying which has resource implications and introduces greater room for error.

Which GIS software to use is also a key consideration. There is a large choice of GIS data software, recently even open-source ones, th have minimum computer operating requirements⁶. The three leading software houses are ESRI, MAPINFO and PCI. ESRI is the most common source of the relatively old, but still reliable and useful, ArcView software (recently updated) and their new GIS platform, ArcGIS. The latest version of EPINFO (2005) includes a component to apply epidemiological data on to digitalized maps and an instruction manual accompanies the software (MAPINFO). The authors are not familiar with this system so cannot provide further comment. However, EPINFO 2005 is downloadable for free from the Centres for Disease Control, Atlanta⁷.

From the point of view of inferential statistical analysis, GIS is of little value for comparisons within single cluster-sampled surveys, since such survey can only provide a reliable estimate for the entire survey area. It is not valid to analyse such surveys by cluster to try and prove a relationship between a risk factor and an outcome. If comparisons between sections are required, then an independent representative sampling method is needed for each geographical section that you wish to compare.

There are plenty of good GIS manuals that can help the reader gain more in-depth knowledge, and two recommended reads are included at the end of this article. More specific public health oriented resources are available on the web, although they tend to consider 'western contexts' rather than developing countries ones. Good online resources include WHO (which includes a Global Atlas)⁸ and the National Centre for Health Statistics of USA⁹.

Final conclusions and considerations

Statistical analysis of GIS nutrition datasets is an area still in its infancy and requires further research and development. Where stratified cluster surveys are used or several different discrete surveys are available, then GIS provides a clear and powerful means to compare differences between areas. As mentioned earlier, cluster survey data is, as a result of the sampling design, only statistically representative of the whole survey area so care must be taken in ascribing significance to geographical variations between clusters within the same survey. When comparing areas within the same cluster survey, then differences that are observed between clusters may be suggestive of real geographical differences, but such differences should not be accepted as statistical fact. This is an important issue to remember if GIS is used for targeting. When systematic (interval) sampling is used, then true geographical differences are easier to identify using spatial statistic tools that are becoming more widely available within GIS software packages.

An investment in GIS in a nutritional survey today, may allow users to benefit from the opportunity to link their data with other databases sources at a relatively low cost in the future. For instance, data on climate, rainfalls, soil erosion, but also food items prices and most of the typical food security parameters, could be subsequently linked. Those data are becoming more and more readily available on the internet thanks to governmental and non governmental agencies, institutions and academic bodies. All this should greatly expand the range of potential cross-referencing of this tool.

It should mot be forgotten that, in some situations, GPS handsets may be considered as military equipment and not appropriate for humanitarian organisations. The potential risks involved in using GPS should be assessed in each operational context.

For further information, contact: Paolo Paron, email: paoloparon@yahoo.it or Andy Seal, email: a.seal@ich.ucl.ac.uk

Recommended manuals P.A. Longley, M.F. Goodchild, D.J. Maguire and D.W. Rhind (2001) Geographic Information Systems and Science, Wiley. (http://www.wiley.com/legacy/wileychi/gis/volumes.html) Ellen K. Cromley & Sara L. McLafferty (2002) GIS and Public Health, The Guilford Press. This explores in depth the nature of spatial data, the mapping of health information and it presents also the use of GIS in different contexts of public health (e.g. vector-borne diseases and access to health services).

---

¹See section later for references to GIS application sources.

²See link to key website sites at http://www.ennonline.net/articles/gis/index.html

³Garmin (http://www.garmin.com/) - manuals of recommended models can be downloaded from this site, Magellan (http://www.magellangps.com/en/) and Trimble (http://www.trimble. com)

⁴Mapsource is the software used with the Garmin model GPS, typically sourced from where the GPS has been purchased.

⁵See online link http://www.ennonline.net/articles/gis/index.html for map sources.

⁶They normally require a PC with a minimum of : Windows operating system, Pentium processor or AMD equivalent, 256 or higher RAM, and a good graphic memory (at least 64 MB).

⁷http://www.cdc.gov

⁸http://www.who.int/csr/mapping/en/

⁹http://www.cdc.gov/nchs/about/otheract/gis/gis_publichealthinfo.htm