Document(s) 11 of 19

Introduction

Malaria is a major public health problem in Sri Lanka, with almost 300,000 infections being reported yearly in a population of 16 million. As much as two-thirds of the entire national public health budget is spent on controlling malaria, and the disease constitutes the fourth highest cause of hospital admission in the country. Malaria has made a major impact on the health, economy, education, and general development of the population (Administration Report 1991). Although in the past P. falciparum malaria was a rarity, its vector is now almost as prevalent as that of P. vivax; further, with chloroquine-resistant strains spreading rapidly, malaria promises to become a more serious problem in the future.

Malaria has always been endemic through much of the country, but especially in the dry zone. The central hill country has remained free of the disease. The malarious areas face perennial transmission with two seasonal rises related to the rainfall pattern: June-July and December-January, associated with the southwestern and the northeastern monsoon rains, respectively. Lately, a significant increase in malarial morbidity has been reported, probably due largely to the increase in P. falciparum infections, many of which are also chloroquine resistant.

A combination of factors contribute to the difficulties associated with trying to achieve a reduction in malaria transmission. These include:

• Environmental and geographic features of the area, such as climate, land-use patterns, and development of irrigation schemes;

• The movement of people and the creation of new settlements;

• Behavioural aspects relating to the population;

• The health care delivery system; and

• The socioeconomic and educational status of the population (Ministry of Health 1987-1991).

In the present study, human malaria infections occurring in an endemic population in Kataragama in southern Sri Lanka were monitored; particular attention was paid to the spatial and geographic features of the area. This enabled an analysis of the malaria risk of this community with respect to three potential risk factors: location of houses in relation to distance from the forest edge and a source of water, and the construction type of houses.

Materials and Methods

Background
This study is part of a research project on the microepidemiology and clinical aspects of malaria, which was conducted in Kataragama, a highly endemic region for both P. falciparum and P. vivax in southern Sri Lanka. The study was carried out over an 18-month period from mid-January 1992 to July 1993. The total population of this study consisted of 1875 residents in 423 houses, in a cluster of eight contiguous villages. The area was geographically and demographically defined by reconnaissance and census.

The Setting
The study site is situated in the country's dry zone, where most people depend on "Chena" (shifting-bush-fallow system) cultivation as a traditional agricultural practice. Most of the farmers live close to the forest edge and/or a source of water. Of the eight villages, one is a small town situated in the middle of the study area. The other villages are surrounded by the forest. A river (the Menik ganga) flows through these villages. The forest is typical of dry zone forests with dry mixed evergreen vegetation. The forest edge consists mainly of thorny bushes, extending gradually to larger trees (scrub to dense forest). The area is characterized by a rainy season from October to February, and a dry season from June to September. In the dry season the river dries up, creating rock pools on the river bed.

Mapping
Aerial photographs of the study area taken in 1993 at an altitude of 10,000 m were obtained from the Survey Department. These photographs were enlarged four times to obtain a map having a scale of 1 : 5000 with real world coordinates. The exact location of the houses, roads, land use, and forest cover, and significant water bodies like rivers, small streams, and reservoirs, were marked on the map. The accuracy of the exact location of such landmarks was confirmed by geographic reconnaissance (GR) carried out during the preparatory stage of the study. The maps were digitized and the GIS package ARC/INFO was used to obtain the nearest distance of a particular house from the forest edge and a source of water. In the case of flowing water bodies, the centre of the flow was used to measure the distance from the water source to the house. All the houses in the study area were numbered, and every house and individual residents in a house were assigned a unique identification number to enable us to monitor malaria infections.

Case Detection
Monitoring of all malaria infections in the study area was carried out by two methods. Passive case detection (PCD) entailed the presentation of fever patients for blood film examination and treatment, either at the Field Research Station which is very close to the study area, or at the Government Hospital, Kataragama, which is about 3 km from the study area. Active case detection (ACD) was done by a series of six mass blood surveys at intervals of 2-3 months. ACD was carried out using three or four teams, with each team doing a house-to-house survey and obtaining thin and thick blood films from all residents. A diagnosis of malaria was established by demonstration of malaria parasites on microscopic examination of thin and/or thick blood films stained with Giemsa stain. It is estimated that a very high proportion of all malaria infections occurring in the population was recorded using these methods.

House Types and Malaria Incidence Rates
All houses in the study area were categorized into two broad classes according to the type of house construction. This categorization was based on the nature of, and building material used for, the walls and roof. Those houses which had incomplete and/or mud walls and roofs made of thatched coconut palms were classified as being of poor construction type, and those with complete plastered brick walls and tiled or corrugated iron roofs were classified as being of good construction type.

The nearest distance from a source of water and from the forest edge to each house was determined using the GIS software package ARC/INFO. In estimating the incidence rates of malaria for each house, those houses with two or fewer residents were excluded from the analysis as they may have given unstable rates, thereby leading to errors in inference.

Definitions
The incidence rate in a house was defined as the total number of malaria infections that occurred in individuals resident in the house during the 18-month period, divided by the number of residents in the house.

The average incidence rate for a group of houses (e.g., of a particular construction type) was estimated as the mean incidence rate of the house in that group.

The incidence rate of a population was defined as the number of infections that occurred in that population divided by the size of the population.

Results

Of a total of 423 houses and 1875 residents in this population, only 343 houses and their resident population of 1744 were considered for analysis, the rest being houses with fewer than three residents.

During the 18-month period, 1579 malaria infections were detected by microscopic examination of blood films in the 343 households; 913 infections were due to P. vivax and 666 to P. falciparum. The incidence rate during the 18 months (number of infections per person) in the selected population was 0.91. Of the total of 343 houses, 182 were of poor construction type and 161 of good type. The distribution of the two types of houses within the area was not uniform, the houses of better construction type tending to cluster around principal roads. Most of the poorly built houses were scattered across the entire area. Residents of this area were almost equally distributed between the good (838) and poor (906) house types.

A significant difference was found in average malaria incidence rates among the populations resident in good and poorly constructed house types, being 0.51 and 1.23 infections per person, respectively. The risk was thus 2.5-fold higher in inhabitants living in the poorly constructed houses (95% CI - 2.19, 2.93).

There was a significant negative correlation between the incidence rates of all houses and the distance of the houses from both the forest edge and a source of water (r = -0.25; p = 0.0001 and r = -0.12; p = 0.0264, respectively). The location of the houses of the two different construction types was not significantly different with respect to their distance from water (p = 0.7251), but it was with respect to distance from the forest edge; the poorly constructed houses were significantly closer to the forest than the good (p = 0.001).

Since house construction type was found to be a determinant of malaria risk and the distribution of the two types of houses were spatially clustered, malaria risk was analyzed in relation to the distance from the forest edge and a source of water separately for the two house types. In neither house type were the incidence rates correlated with the distance from the forest edge (r = -0.09; p = 0.2292 for houses of poor construction type; r = -0.12; p = 0.1431 for houses of good construction type). This implies that the association between malaria risk and the distance of houses from the forest edge was confounded by the type of house construction.

With respect to the malaria risk in relation to the distance from water, a significant correlation was obtained for poorly constructed houses, the risk being greater in houses closer to a source of water (r = -0.31; p = 0.0001). In houses of good construction type, however, a marginally significant correlation was found with distance from the water, but in the reverse direction (r = 0.14; p = 0.0676).

Discussion

Malaria has been prevalent in most parts of the tropical and subtropical world, including parts of Africa, Asia, and Central and South America, for a long time, The disease has been the single most important cause of morbidity in some of these areas. Yet, a desirable level of control has not been achieved during the past two decades. Prospects appear bleak for the near and distant future as limited options for control are currently available. Resources are also becoming scarce. Disease control in the future is likely, therefore, to benefit from concerted efforts, including those based on an understanding of the microepidemiology of the disease in a given situation. This finding motivated this research tem to conduct studies on the microepidemiology and clinical aspects of malaria, paying particular attention to ecological factors.

In a previous study, it was demonstrated that malaria infections in an endemic area were clustered in particular individuals and households (Gamage-Menids et al. 1991). It has also been shown that the incidence of malaria is associated with the construction type of houses. Residents of poorly built houses with incomplete mud walls and thatched roofs were at a higher risk of acquiring malaria than those living in better built houses with complete plastered brick walls and tiled or corrugated iron roofs (Gunawardena 1994).

The present study supports earlier findings of an association between malaria risk and house construction type, this study taking place in a different population and conducted over a different period of time. In addition, this study presents a more detailed examination of two other ecological factors that could determine the malaria risk: the proximity of the house to the forest edge and a source of water. Both these factors could, by their association with the mosquito vector, be expected to influence the malaria risk in an endemic community. The forest edge is thought to provide ideal resting places for the mosquito, thus influencing its survival and the level at which malaria transmission is sustained. The water bodies and rivers examined in this analysis are known to be potential breeding places of the mosquito vector. Thus, proximity to both might be expected to impose an added risk for acquiring malaria.

The findings of this study indicate that the risk of contracting malaria was on average 2.5-fold greater for a resident of a poorly constructed house than for one who lives in a well built house. In the first study which demonstrated an association between house types and malaria risk (Ministry of Health 1991), higher densities of mosquitoes were measured and found resting within poorly constructed houses, than in houses of better construction type.

This finding suggests that the risk of being bitten by a mosquito might have been greater within the poorly constructed houses, given that the construction type of the house itself afforded a better respite for the vector. It is nevertheless believed that several other factors associated with the type of house, such as socioeconomic status, education levels, and occupations of the residents, might independently influence the malaria risk. Thus, for the purpose of the discussion that follows, the house construction type is considered as much a marker of the general standard of living, as of the construction type itself.

Overall, there was a significant negative correlation between the incidence of malaria and the distance of the house from the forest edge and a source of water, when all houses in the community were considered. This was not evident when the analysis was performed separately for each of the two house construction types. In neither house type was the malaria risk correlated significantly with the distance from the forest edge. Houses of good construction type were found to be clustered around the major roads in the area, whereas the houses of poor construction were generally more scattered. Thus, most of the houses that were close to the forest edge were of the poor construction type; this is easily explained by the fact that better houses belong to the more affluent, who have the ability to acquire land in relatively more urban parts of the area. It follows, therefore, that because the better built houses were located away from the forest edge, house construction type either by itself or by the factors that are associated with having a poor house, was confounding the apparent relationship between malaria incidence and the location of the house in relation to the forest edge.

Houses of both construction types were distributed fairly equally in relation to the sources of water. The relationship between malaria risk and proximity to a source of water was different for houses of the two construction types. For the poorly constructed houses, a significant negative correlation was found with the distance from water, demonstrating a decrease in malaria risk with increasing distance from the water. This was not so in the better constructed houses, in which such a correlation was not apparent. In fact, a trend in the reverse direction was evident, i.e., a tendency for the malaria risk to decrease as the house moved closer to the water source, although at a borderline level of significance. The water bodies considered here are potential breeding sites for the vector mosquito. The increasing risk of malaria in houses located close to these water sources is therefore to be expected, for the reason that a higher density of mosquitoes must prevail in the vicinity and the probability exists for a higher man-mosquito contact closer to the water.

This appears to be so in the case of poorly constructed houses, but not in houses of the better construction type. It thus appears that the risk imposed by being close to a source of water is transcended by an improvement in the house type, be it because of the construction type itself or the associated life styles or characteristics of its residents. If the trend of a decreasing malaria risk in residents of good house type closer to water reaches a level of statistical significance, it would suggest that other factors pertaining to its residents and the house which have not been examined here are more important determinants of malaria risk than the distance from water.

The association between malaria risk and the location of the house in relation to a source of water has implications for irrigation schemes and agriculture-based developmental projects, both of which involve manipulation of waterways in which malaria vectors breed. They also result in increasing the number of human settlements located close to water. The finding, however, that the higher risk of malaria imposed by being close to a source of water can be overcome by a better house type and/or higher standard of living is encouraging for such projects, which are expected to eventually increase incomes and improve living standards. It also implies that the initial health hazards for settlers in such schemes could be significantly offset by investing in better housing and a generally higher standard of living.

Relevance of GIS to this Study

This study constitutes an example of the use for GIS in health research. The specific advantages offered for this analysis by GIS were as follows:

• The display of the data pertaining to the study area provided an overview of the malaria incidence in relation to geographically and ecologically important entities. It also provided clues for further analysis, e.g., the fact that the better built houses were seen to be spatially clustered around the major roads suggested that house construction type may have confounded the relationship between malaria risk and the two ecological entities — the distances from the forest edge and water; this prompted an analysis of the correlations with these two ecological risk factors separately in the two different house types. The program offered the obvious advantage of reproducibility of high-quality maps for presentation and publication.

• It was possible to obtain accurate distances between selected landmarks which were essential to this study.

It is obvious that the usefulness of GIS and the value of inferences made from spatial analyses will depend on the quality of the data used. This analysis was based on data of a very high quality both in terms of accuracy and detail, obtained from a research project. It is probably unrealistic to expect data of such high quality from disease control programs. This study and analysis therefore exemplifies the use of GIS primarily as a tool for health research. Our ultimate goal, however, includes the development of a GIS that will be useful as a decision support system for a malaria control program.

In developing a GIS application that can be used effectively in a disease control program, it is essential that the important variables are identified and accurate data obtained. In this regard, it is pertinent to address the following questions:

• Should GIS applications for disease control be scientifically evaluated?

• If these applications are to be evaluated, what criteria should be used?

• At what point should the evaluation be carried out?

Such an activity requires an enormous amount of resources, both in terms of hardware and human resources. Unless there is sufficient scientific evidence to show that GIS applications help in reducing malaria transmission via a better decision support system and implementation of such decisions, it would be an arduous task to convince both policymakers and control personnel of the need to use GIS applications in disease control programs.

The main objective of disease control is to reduce transmission and incidence of disease. Therefore, evaluation of a tool that aids disease control should use the occurrence of disease as the marker. A useful tool should objectively show a reasonable level of reduction of the incidence of disease. What constitutes a reasonable reduction would depend on the program and the perceived threat of the disease.

The application of GIS for health is still in its infancy. Therefore, early evaluations may over- or underestimate the usefulness of GIS as a tool in disease control. However, a mechanism for evaluation of GIS applications should be incorporated in planned stages in the development of such applications in the future. Judicious evaluations would provide useful information that would eventually lead to improvement of the application, increasing the prospects of it being a more useful tool for disease control.

References

• Administration Report. 1991. Anti Malaria Campaign. Ministry of Health, Sri Lanka.

• Gamage-Mendis, A.C.; Carter, R.; Mendis, C.; De Zoysa, A.P.K.; Herath, P.R.J.; Mendis, K.N. 1991. Malaria infections are cloistered within an endemic population: risk of malaria associated with house construction type. Am. J. Trop. Med. Hyg., 45, 77-85.

• Gunawardena, D.M.; Mutuwatta, L.P.; Wickramasingha, A.R.; Weerasingha, S.; Rajakaruna, J.; Carter, R.; Mendis, K. N. 1994. Malaria risk factors in an endemic area of southern Sri Lanka. Sri Lanka Association for the Advancement of Science. (submitted for publication)

• Ministry of Health. 1987-1991. Plan of operation for malaria control in Sri Lanka. Anti-Malaria Campaign, Sri Lanka.

Document(s) 11 of 19