Modeling and predicting the spread of Filaria and development of a Decision...

Modeling and predicting the spread of Filaria and development of a Decision Support System


Atma Bharathi M*#., Ananth R.*, Mohan Raj K.*, Arun Kumar P.*
*Institute of Remote Sensing, College of Engineering Guindy, Anna University, Chennai.
#Corresponding author: [email protected]


Filariasis is a vector borne disease with mosquitoes as the vector. Our model considered the environmental parameters such as temperature, relative humidity, rainfall, water stagnation spots, water quality, forest type, and soil type to map the potential mosquito breeding zones. A SRTM scene was used to derive DEM, and subsequently slope and sink vectors to determine possible water stagnation zones. Forest type and soil types were derived from landuse and geomorphology maps. The water quality and rainfall data was collected as point data from TWAD board and DMS and an IDW map was derived. The layers were combined and a Cumulative Suitability Index representing high, moderate, marginal and low risk zones was developed using the analytical hierarchical procedure. The results obtained were compared with filaria incidence data from DPH. We found that while there existed topological consistency in prediction, the model had largely exaggerated the risk zones. The reasons and methods to fine tune the model are discussed.

In addition, a site suitability map for setting up new Health Sub Centers (HSC) under each DPH was developed and the method is discussed.


The World Health Organization reports that a global health disaster where 750 million could be at risk with 120 million already infected, would plague the world. Filariasis is becoming a serious health problem in East Asia since China, India, Indonesia alone account for two-thirds of world’s infected population [1], [2]. India is the largest Lymphatic Filariasis endemic country in the world and has targeted elimination by 2015 . In a health survey conducted in 1995 India had 420 infections. Among the 7 heavily infected regions in Tamil Nadu, Cuddalore stands in an alarming state. [3]
Reports from earlier studies reflect upon the importance of climatic and environmental parameters in developing ‘early warning systems’ for epidemic diseases [4], [5].

In India the state level health administrative structure has a Directorate of Medical Services (DMS) at the top, the Government Hospitals (GH) and Public Health Centers (PHC) are its two sub divisions. While the primary aim of Government Hospitals is to control the morbidity and mortality, the primary aim of Public Health Centers is prevention of diseases, especially endemic and epidemic ones. The PHCs in turn have Health Sub Centers (HSCs) under them, which constitute the last level of service. [6] This work was conducted to aid the DPH by being a decision support system. Cuddalore, which spans from 11 ? 9′ 16” to 11 ? 54′ 13” in latitude and 79 ? 48′ 34” to 78 ? 52′ 56” in longitude and is a district along the east coast of Tamil Nadu, was our study area.

About Filariasis

Filariasis is a vector borne disease with kinds of mosquitoes being its vector. It is a harboring disease wherein the infected person can remain normal and in many cases without the knowledge of being infected for up to 15 years. Microfilaria is the parasite and it infects the lymphatic system in humans. It is passed from human to human by mosquitoes, which must survive at least ten days for transmission of microfilaria [7].

Lymphatic Filariasis (LF) is primarily a disease of the poor [3] because of its frequent prevalence in remote rural areas, disfavoured urban, and fringe areas. India loses about 1.2 billion person-days due to LF. About one hundred and twenty million people in at least 80 countries of the world are infected with lymphatic filarial parasites, and it is estimated that 1 billion (20% of the world’s population) are at risk of acquiring infection [2], [7].

Ninety percent of these infections are caused by the parasite named Wuchereria bancrofti. The major vectors for W. bancrofti are

• Culicine mosquitoes in most urban and semi-urban areas,
• Anophelines in the more rural areas
• Aedes species in many of the endemic regions.
Figure 1 shows the mode of transmission of disease from the larvae to the human system and the various stages in between.
Mode of transmission of filaria to humans.

The breeding, longevity and development of the vector depends on certain Socio Economic factors such as Urbanization which in turn causes breakdown of sanitation, unhygienic and unhealthy conditions, improper waste and solid management, rapid Industrialization and its byproducts such as plastic bags, disposable cups, tires which are dumped into drainage thereby blocking drain. The other prime factor is the Environmental condition and the following is the list of favourable conditions for the breeding and development of the vector.

• Temperature : 22 to 38 ? C
• Relative Humidity : 70 %
• Drainage : Chocked drains with stagnant water.
• Common breeding spots: Cesspools, soakage pits, ill-maintained drains, septic tanks, open ditches and burrow pits.
• Aquatic plants like Pistia stratiotes in ponds.
• Evergreen forests
• Stagnated water parameters
o Ponds that lack enough oxygen do not contain fishes – the potential predators for the eggs of mosquitoes
o Ponds with submerged aquatic vegetation and with algal blooms aid in increased egg laying.
• Soils like alluvial, clay aid in water stagnation and hence are favorable factors [7].
Our work was centered on mapping and modeling these environmental parameters and using the model to predict the potential zones for filaria and validating the model with the real incidence data.


The following geospatial layers were collected.

1. Cuddalore administrative boundary map.(district, block, village panchayat)
2. Cuddalore road network map.
3. Habitation points.
4. Land use map.
5. Rivers, streams, tanks map.
6. Geomorphology map.
7. Forest map.
8. Landmarks (containing PHC, Govt. Hospitals).
9. SRTM imagery and the following non spatial data were collected
10. TWAD board water quality data of hand and power pumps
11. Rainfall data from DMS
12. Medical service jurisdiction of PHC, HSC
13. Census data for every HSC
14. Filarial incidence, HSC wise count.

Data Preparation

The constituent villages of each HSC and PHC were merged to create administrative boundary maps. The water quality index table and the rainfall tables were linked to the habitation points. The disease incidence table was linked to the newly created HSC administrative boundary map to create the ground truth map.

Preparation of Slope and Sink maps

The SRTM imagery was reprojected to TM with WGS 84 datum to match the rest of the datasets. A subset of the study area was made with the district boundary map. Using ERDAS Imagine 8.6, surfacing was performed and the raster was converted to a point coverage. Using 3D analyst in ESRI ArcGIS 9.1 a Triangular Irregular Network (TIN) was built. Later, a TIN grid was made and subsequently slope and sink rasters were prepared. Finally the sink raster was reconverted to vector format to aid in overlay analysis. Buffers of size 300m were built around the sinks to represent the regions which would be affected by these sinks.

Preparation of rainfall and water quality index themes
From the newly created water quality index point dataset, only the non potable pumps were chosen. An Inverse Distance Weight (IDW) was generated using the pH values as Z field. pH value is related with the Biological Oxygen Demand (BOD). A higher BOD reduces the suitability of the water body for fishes (which feed on the mosquito eggs and larvae) and thus increases the mosquito population [8]. Subsequently the IDW raster was converted to vector polygons using the raster to vector conversion tool of Spatial Analyst.
Similarly, with the rainfall observation stations were plotted, an IDW was created and it was converted to vector.

Mapping of filarial risk zones

The themes which represent the influence of environmental factors like Slope, water quality, rainfall, sinks, geomorphology, land use and forest were combined using the Union operation. The respective theme, its sub elements with ranks and the weight for each theme is given in Table 1. As per the analytical hierarchical procedure, the ranks in each theme sum up to 1 and the weights of all themes sum up to 1. The value of rank or weight represents the sub element / theme’s impact on the model.

Ranks and weights for factors influencing spread of filariasis.Finally CSI was calculated using the formula Sum[ weights * ranks ] for every factor. CSI = [(0.3085 * slope_rank ) + (0.2639 * waterquality_rank) + (0.1744 * rainfall_rank) + (0.122 * sink_rank) + (0.080 * geom._rank) + (0.049 * luse_rank) + (0.028 * forest_rank)]

The zones had a mean of 0.2612 and Standard Deviation = 0.1028. The zones were classified using 1% Standard Deviation as high (greater than mean + 1SD), moderate (between mean and mean + 1SD), marginal (between mean – 1SD and mean) and low (lesser than mean – SD) risk zones and a map was made as shown in Fig 2.
Filarial risk zones from the environmental model.Validation with field incidence

The field incidence table had a mean incidence of 0.3791 and a SD of 0.9169. Similarly another map was created using 1SD as shown in figure 3.
Filarial risk zones from field incidence data

Suitable sites for setting up new HSCs

As each village grows in population, it becomes necessary for the DPH to allot new HSCs. The DMS considers villages or village groups with population greater than 5000, accessibility to roads, closeness to habitation and presence of a government or vacant land as suitable factors for setting up a new HSC [7].

To map the sites suitable for setting up a HSC, the following layers were prepared. Buffers were created for Habitation points for a distance of 500m. Similarly, buffers of 500, 250, 150 and 50m were created around NH, SH, district and village roads respectively to represent their accessibility. An analytical hierarchical procedure based table was developed containing the union of habitation buffer, road buffer, HSC population theme and landuse themes. A cumulative suitability index was calculated with the ranks and weights. Finally a suitability map was generated with 1SD values as shown in Fig 4.
Suitable sites for setting up new HSCs

Comparing the model’s prediction of risk zones and the field reality, we find topological consistency. The high risk zones, though it appears larger in the model, does not change in location. Similarly, the moderate risk zones correlate topologically, but it has swelled in the model. The marginal risk zone has encroached into no-incidence zone of the field data. Further, the boundary of the field incidence map goes with the HSC boundary while the model maps the exact boundary of the risk zone. We find that the model has exaggerated the effect.


The model was constructed by considering only the environmental parameters. Socio-economic factors like poverty, illiteracy, urbanization play a key role. The model if expanded to consider socio-economic factors would score well. This lays stress on the availability of geographic data.

The SRTM imagery contains an error of 30 meters. With accurate digital surface models, the runoff, water stagnation zones that play a dominant role, can be predicted accurately. There was a gross temporal mismatch in the data used. While the census data was 2001, the field incidence data was from 2004 and the geospatial layers had different preparation dates. The model can be enhanced by interpreting a satellite imagery to determine the tanks in which “Pistia stratiotes” grows.

This study warrants further fine tuning of the weights and ranks allotted in the model. A pilot study could be conducted on a village group; the parameters, their ranks and weights could be reverse engineered. Newer and hitherto unknown factors might also be discovered. This model could then be applied on district level and the results can be validated.


We express our deep gratitude Dr. S. Kaliappan, Former Director, Institute of Remote Sensing, Anna University, Dr. Ramalingam Project Head NRIS, Institute of Remote Sensing and Mr. M. V. HEMADRI, Scientist, IRS, Anna University, for permitting us to take up this work.
We express our gratitude to Dr. Chandra Mohan Joint Director of Public Health, Cuddalore for providing us with the required health data of Cuddalore District.
We express our sincere thanks to Dr. Haffiza and Mr. C. Palaniswamy District Malaria Officer of Directorate of Public Health, Cuddalore for lecturing on various aspects of the disease filariasis and on the overall structure of the Health Dept. in Tamil Nadu.
With great pleasure, we thank Dr. Jagadeesh Additional Director, Vector Control, Directorate of Medical Service and other faculties who helped us in guiding our project to its consummation. We are greatly indebted to Mr. R. Narayanan who guided us on various aspects of this project.


[7]. K. Park, “Social and preventive medicine”, pp. 220 – 225.
Journals [3]. Pani SP, Kumaraswami V, Das LK.(2005),”Epidemiology of lymphatic filariasis with special reference to urogenital-manifestations”. Indian Journal of Urology 2005; 21:44-9
[4]. M. C. Thomson, F. J. Doblas-Reyes, S. J. Mason, R. Hagedorn, S. J. Connor, T. Phindela, A. P. Morse & T. N. Palmer, “Malaria early warnings based on seasonal climate forecasts from multi-model ensembles”, Nature 439, 576-579 (2 February 2006) Reports
[2]. “Prospects of Elimination of Lymphatic Filariasis in India”, ICMR Bulletin, Vol 32, No. 4&5, May-June 2002. Interviews
[6]. Dr. Haffiza and Mr. C. Palaniswamy, District Malaria Officer, Directorate of Public Health, Cuddalore. WebPages
[1]. The Global Programme to Eliminate Lymphatic Filariasis (GPELF), World Health Organization, “”
[5]. “Malaria early-warning system shows promise in tackling epidemics”, Innovations Report, “ html/ reports/ life_sciences/ report-54723.html”
[8]. “Management of Ponds, Wetlands, and Other Water Reservoirs to Minimize Mosquitoes”, Brent Ladd and Jane Frankenberger, Agricultural & Biological Engineering Department, Center for Environmental Studies, Purdue University. “”