Application of Multiple Linear Regression Model through GIS and ...

HEALTH SCIENCE JOURNAL® Volume 6, Issue 4 (October – December 2012)

_ORIGINAL ARTICLE_

Application of Multiple Linear Regression Model through GIS and Remote Sensing for Malaria Mapping in Varanasi District, INDIA Praveen Kumar Ra1, Mahendra Singh Nathawat2, Mohhamad Onagh1 1.Department of Geography, Banaras Hindu University, Varanasi, INDIA 2.School of Science, Indira Gandhi National Open University, New Delhi, INDIA 1.

ABSTRACT Background: The production of malaria maps relies on modeling to predict the risk for most of the area, with actual observations of malaria prevalence usually only known at a limited number of specific locations. However, the estimation is complicated by the fact that there is often local variation of risk that cannot easily be accounted for by the known variables. An attempt has to be made for Varanasi district to evaluate status of Malaria disease and to develop a model, by which malaria prone zones were predicted by five classes of relative malaria susceptibility i.e. Very Low, Low, Moderate, High, and Very High categories. Methodology: Multiple Linear regression models were built for malaria cases reported in study area, as the dependent variable and various time based groupings of average temperature, rainfall and NDVI data as the independent variables. GIS is be used to investigate associations between such variables and the distribution of the different species responsible for malaria transmission. Accurate prediction of risk is dependent on knowledge of a number of variables i.e Land Use, NDVI, climatic factors, distance to location of existing government health centers, population, distance to ponds, streams and roads etc. that are related to malaria transmission. Climatic factors, particularly rainfall, temperature and relative humidity are known to have a strong influence on the biology of mosquitoes. To produce malaria susceptibility map in this method, the amounts of quantitative and qualitative variables based on sampling of 50×50 networks in form of a 38622×9 matrix have been transferred from GIS software (ILWIS 3.4 and ARC GIS-9.3) into statistical software (SPSS). Results: Percentage of malaria area is very much related to distance to health facilities. It is found that, 4.77% of malaria area is belonging to 0-1000 m buffer distance to health facilities and 24.10% of malaria area comes in 6000-10000 m buffer distance. As the distance to health facilities increases, malaria area is also increasing. Key words: Multivariate linear regression, malaria susceptibility model, GIS, Remote Sensing, Varanasi. CORRESPONDING AUTHOR

Praveen Kumar Rai, E-mail: [email protected]

INTRODUCTION Page | 731 E-ISSN: 1791-809X

Health Science Journal © All rights reserved

www.hsj.gr

Quarterly scientific, online publication by Department of Nursing A’, Technological Educational Institute of Athens

he representation and analysis of

T

importance of these factors has yet to be

maps of disease-incidence data is a

determined.2, 3

basic tool in the analysis of regional

Malaria (marsh fever, periodic fever) is a

variation in public health. Tobler’s first

parasitic disease that involves infection

law of geography, which states that

of the red blood cells (RBCs). Malaria in

“things that are closer are more related,”

humans is caused by the transmission of

is central to core spatial analytical

one or more of four parasitic, class

techniques

Sporozoa.

as

well

as

analytical

The

severity

and

clinical

conceptions of geographic space. In the

presentation of symptoms depend largely

case of disease spread, individuals near

on

the

species

of

Plasmodium

4

or exposed to a contagious person or a

contracted. The four species responsible

tainted

are

for infection are Plasmodium falciparum,

deemed more susceptible to certain types

Plasmodium vivax, Plasmodium ovale

of illnesses.1

and Plasmodium malaria. These species

The rapid urbanization in many parts of

vary widely with respect to geographic

the world is changing the context for

distribution, physical appearance and

human population and their interaction

immunogenic

with

transmission depends on the diverse

environmental

the

natural

setting

ecosystem.

To

factors

malaria mosquito human relationship, it

parasites,

is required to identify the type of human

interactions among them. These factors

migration,

may

economic

status,

environmental This

behavior

aspects

requirement

growth,

socio

and

around

the

them.

underscores

the

human

and

condition

The

may

rainfall,

mosquito

vector

and

the

others,

environmental most

apparent

determinants are the meteorological and environmental

the

vectors,

among

meteorological etc.

the

hosts,

include,

importance of human intervention that affect

influence

Malaria

understand the complex nature of the

population

that

potential.

parameters,

temperature,

such

humidity

as and

population and the intensity of parasitic

vegetation.5,6 That there are so few

transmission in endemic areas, whether

examples of the use of epidemiological

in rural or urban settings. The key

maps

determinants of the outcome of malaria

explained

should be related to the human host,

spatially

parasite,

understanding of how epidemiological

parameters.

vector

or

However,

environmental the

relative

in

malaria

control

by

lack

the

defined

data

of and

may

be

suitable, of

an

variables relate to disease outcome.

Page | 732 Application of Multiple Linear Regression Model through GIS and Remote Sensing for Malaria Mapping in Varanasi District, INDIA


However, recent evidence suggests that

healthcare planning, GIS has shown its

the clinical outcomes of infection are

capability to answer a diverse range of

determined by the intensity of parasite

questions relating to the key goals of

exposure,

efficiency, effectiveness, and equity of

and

developments

in

geographical information systems (GIS) provide

new

ways

epidemiological

data

to

represent

spatially.

GIS

the

provision

services. play

12,13

a

significant

reorganization

climatic

disease

of

the

collection

localities with the presence or absence of the various species.

7,8

This computer

public

of

planning

century,

health

Unquestionably, GIS will

software is being used to correlate the attributes

of

part

public in

especially

the in

in

the

health

and

twenty-first response

to

sweeping changes taking place in the

based technology has been available for a

handling of health information.14

number of years but it is only recently

Vegetation

that it has been widely appreciated as a

vector breeding, feeding, and resting

powerful new tool to augment existing

locations.

monitoring and evaluation methods for

indices have been used in remote sensing

disease mapping.5, 9

and Earth science disciplines. The most

Mathematical and statistical modules

widely used index is the Normalized

embedded in GIS enable the testing of

Difference Vegetation Index (NDVI). It is

hypotheses

estimation,

simply defined as the difference between

explanation, and prediction of spatial

the red and the near infrared bands

and

and

the trends.10

temporal

is A

often

associated

number

of

with

vegetation

Statistical

normalized by twice the mean of these

techniques model the relation between

two bands. For green vegetation, the

parasitaemia

factors

reflectance in the red band is low

(environmental, possible interventions,

because of chlorophyll absorption, and

socio-economic

a

the reflectance in the near infra-red band

model,

is high because of the spongy mesophyll

multivariate

risk

and

risk

factors)

linear

via

regression

which is further used for prediction.11

leave structure.1

GIS plays a variety of roles in the

Because malaria is vector-borne, there

planning

the

are many remotely sensed abiotic and

dynamic and complex healthcare system

biotic environmental variables that are

and disease mapping. Although still at an

relevant

early stage of integration into public

transmission and habitat niches of the

and

management

of

to

the

study

of

malarial

Page | 733 E-ISSN: 1791-809X


www.hsj.gr


vector. For example, the Normalized

indicators (of discrete and continuous

Difference Vegetation Index (NDVI) is a

nature) form remote sensing data and

characterization of vegetative density

incorporate these into the core of the

based on the amount and wavelength of

analysis.

the radiation reflected by the leaves of a

desegregation at which the analysis has

plant.

been undertaken in the present study is

When

vegetation

is

photosynthetically active, it has a high

not

reflectance in the near-infrared region of

methods.

Indeed,

possible

the

through

level

of

conventional

the spectrum and a low reflectance in the red portion of the spectrum. In an

Objectives of the Study

environment where vegetation is healthy

The main aim of this study is to develop

and green, the leaves of the resident

a malaria distribution map and malaria

vegetation

significant

susceptibility models using multivariate

percentage of the visible light produced

linear regression analysis though Remote

by the sun.15 The more vigorous and

Sensing data and GIS techniques.

will

absorb

a

denser the vegetation is, therefore, the higher the NDVI becomes. NDVI has also

Study Area

been used as a surrogate for rainfall

The study is Varanasi district, U.P.,

estimate. It is an effective measure for

extending between the latitude of 25°10’

arid or semi-arid region. For tropical

N to 25°37’ N latitude and longitudes of

regions where ample rainfall is normally

82°39’ E to 83°10’ E, lies in eastern Uttar

received, vegetation index is a less

Pradesh. Its major area is stretched

sensitive measure for estimating rainfall.

towards the west and north of the

The mean vegetation index over a region

Varanasi city spread over an area of

reflects the degree of urbanization or

1454.11 sq. km (Fig.1). Administrative

lack of vegetation. In this sense, NDVI in

the

a grid cell is used as an indicator for the

namely,

mean level of vegetation present in the

which are further sub-divided into eight

cell.

development blocks namely Baragaon,

The present study illustrates how GIS

Pindra, Cholapur, Chiraigaon, Harhua,

allows integration of different data sets

Sevapuri, Araziline and

to

Vidapeeth altogether consisting of 1336

arrive

at

holistic

or

aggregative

solutions. It shows how the technique can

help

in

generating

district

comprises

Pindra

and

two

Varanasi

tahsils Sadar

Kashi

villages.

additional



Data Collection and Assessment of Used

above data sources have been used to

Parameters

generate various thematic data layers.

Successful

prediction

of

malaria

Climate data not only for Varanasi

occurrence and the production of a map

district

of the malaria prone areas call for the

district/places like Patna, Gaya were

collection of the relevant spatial data. A

used for comparative variation and for

number of thematic maps (referred to as

interpolation in trend of rainfall and

data

temperature from Varanasi district.

layers

in

GIS)

on

specific

but

for

various

neighboring

parameters or parameters which are related to the occurrence of malaria,

Methodology

distance to water bodies, distance to

A number of thematic maps (referred to

river,

as data

distance

to

hospital,

rainfall,

layers in

GIS) on

specific

temperature, land use/landcover, NDVI

parameters or parameters which are

etc. have been generated (Fig. 2).

related to the occurrence of malaria, viz.

A malaria susceptibility map (i.e. malaria

land

susceptibility

zone

ponds/tanks, distance to river, distance

susceptibility

index)

and has

malaria also

use,

NDVI,

distance

to

water

been

to road, distance to hospital, rainfall,

prepared. The basic data sources that

temperature and expected population

have been used to generate these layers

density of year 2009 have been generated

are including IRS-1C LISS III data of year

(Fig 2). In this study, Ilwis Version-3.4

2008, SOI topographic maps (1:50,000

and ArcGIS Version-9.3 GIS and ERDAS

scale). Census data of year 2001 was also

Imagine Version-9.1 software were used

used and using this population data of

to produce the layer maps that assist in

year 2001, projected population of year

the

2009 was calculated which was used to

susceptibility maps. Topography map of

calculate population density of year

1:50,000 scale of study area were used to

2009. Besides, field surveys have been

digitize district and development block

carried out for verification and condition

boundary. The coordinates of important

of ponds/water tanks, health facilities in

point for geo reference point like road

PHC’s/CHC’s and government hospitals.

conjunction points and malaria prone

Malaria data of year 2009 was used for

area, existing health care facilities units

this study. These data are taken from

were measured during the field surveys

District Malaria Office, Varanasi. The

using Global Position Systems (GPS)

production

of

the

malaria

Page | 735 E-ISSN: 1791-809X


www.hsj.gr


technology. In the measurement phase,

Furthermore, it will help to make an

one receiver served as a base station,

equation and linear function (model) for

while the other was used to collect GPS

malaria susceptibility in intended study

data at the selected ground control

area. All these used parameters were

points. To establish the relationship

analysed in SPSS statistical software

between object space and image space,

using multiple linear regression model

the ground control points were selected

and crossed to each other and then

in

finally Malaria Susceptibility Index (MSI)

the

model

area

measurements

in

to

conduct

the

all

National

and

Malaria

Susceptibility

Zonation

Coordinate System. The vector maps

(MSZ) were produced.

were produced from the IRS LISS-III

In this study equation of the theoretical

remote

model will be described as follows.

sensing

data

and

SOI

topographical map. Therefore, land use

L  B0  b1 X 1  b2 X 2  b3 X 3  ...  bm X m  

map, NDVI and vector layers of water

Where, L is the occurrence of Malaria in

bodies and other important parameters

each unit, X’s are the input independent

used in this study were delineated in

variables

ERDA Imagine-9.1 and ARC GIS-9.3

observed for each mapping unit, the B’s

software.

geo-

are coefficients estimated from the data

referenced digital map of development

through statistical techniques, and ε

blocks/districts were used. In order to

represents the model error.16 To produce

For

GIS

platform

carry out multivariate analysis of data and to determine the all parameters responsible for malaria in the study area, a multiple linear regression has been used. Multiple Linear regression models were built for malaria cases reported in study area, as the dependent variable

malaria

(or

instability

susceptibility

parameters)

map

in

this

method, the amounts of quantitative and qualitative variables based on sampling of 50×50 networks in form of a 38622×9 matrix have been transferred from GIS software (ILWIS 3.4 and ARC GIS-9.3) into statistical software (SPSS).

and various time based groupings of temperature, rainfall and NDVI data as the independent variables. The multiple linear regression method reveals that how the susceptibility of malaria as the standard variables

deviation and

of

independent

predictors

change.

Discussion Malaria exists in every tropical and subtropical landscape across the globe; sometimes making seasonal excursions into temperate areas as well. 15 The protozoan parasites that cause it have



more complex genomes, metabolisms

disease

and life cycles than almost any other

effective prevention decisions may be

vector-borne

complexity

made by the government and public

target

health

makes

threat.

them

a

This

difficult

for

may

be

uncovered.

institutions

through

More

better

interventions such as drugs and vaccines

allocation of medical resources by using

because

the network analysis models of a GIS.19

the

parasite’s

shape-shifting

ways allow it to evade chemical and immunological defenses. They pose a

Malaria Influencing Data Layers

moving

intentionally

The regression coefficients of this model

changing their outer coating during each

(multiple linear regressions) are given in

phase of their life cycle, and creating a

Table 1.

diverse

Rainfall

target

as

well,

antigenic

and

metabolic

wardrobe through sexual recombination,

Rainfall is considered to be the most

an

creation

important malaria triggering parameter

unavailable to simpler microbes such as

causing soil saturation and a rise in pore-

viruses and bacteria.17

water pressure. However, there are not

The origin and subsequent spread of

many examples of the use of this

malaria diseases have a close relation

parameter in stability zonation, probably

with time and geographic locations. If

due to the difficulty in collecting rainfall

disease

in

data for long periods over large areas.

space/location and time and they contain

After interpolation between amounts of

essential disease attributes, the spatial

annual rainfall in the study area stations,

distribution and temporal characteristics

the isohyets map created. Finally this

of the disease spread may be monitored

map has been grouped into five classes

and visualized for probable intervention.

to prepare the rainfall data layer (Fig.

With the availability of disease spread

2a). It was verified that approximately

models, the contagious process may be

maximum

dynamically simulated and visualized in

occurred in >984 mm rainfall class. In

two or three dimensional spatial scales.18

>984 mm rainfall class, 30.81% of

Consequently;

population

malaria area comes in very low and low

groups may be identified and visually

zones whereas 6.29% of malaria area

located while the spatial distributional

calculated for very high zones but in
10000m buffer distance of stream

categories calculated, which is 0.03%

and in this zone only very low and

(Table 1).

moderate categories are available, which

distance to road increases malaria area

is mainly because of influence of some of

percentage shown decreasing trend.

only

very

low

and

Here also seen that,

low as

the other indicators/variables and at this distance malaria indicators or breeding

Distance to Health Facilities

sources are not very much influence on

Health facilities of the Varanasi district

people. One important thing it was also

are based on mainly modern allopathic

found in this study that as the distance

of treatment. To know the distributional

to rivers/stream increases, percentage of

pattern of health care facilities, data has

malaria effected area in high and very

been collected from CMO office and

high categories are decreasing, which is

government hospital located in rural

8.95% within the 0-1000m and 2.97% in

areas of Varanasi district. The existing

6000-9000m

health facilities both in rural and urban

buffer

zone

respectively

(Table 1).

area were surveyed with the help of Leica

DGPS.

There

are

different

Distance to Road

categories of health centre providing

Similar to the effect of the distance to

infrastructure

streams, distance to road is also one of

district. The PHC’s are dotted in the

important parameters to estimate the

district located at an interval of 10-20

distance of road from existing health

kms and the tahsil hospitals are located

care facilities in the study area. Five

about 50 km apart.

different buffer areas were created on

with the distribution of medical centre’s

the path of the road to determine the

of the district bears a close relationship

effect of the road on the malaria disease

with the hierarchy a central places and

(Fig. 2e). The malaria area percentage in

population

each buffer zone is given in Table 1 and

Besides, the transport network has also

shows that 63.83% of the malaria area

influenced the growth of health care

and

size

treatment

of

in

the

The hierarchical

the

settlement.

Page | 739 E-ISSN: 1791-809X


www.hsj.gr


facilities. Percentage of malaria area is

the area is mainly sourced from heaps of

very much related to distance to health

garbage. The solid and liquid wastes

facilities (Fig. 2f). In Table 1, it is found

generated out of the household and

that, 4.77% of malaria area is belonging

industrial activities are dumped and

to 0-1000 m buffer distance to health

released in uncontrolled sites. These

facilities and 24.10% of malaria area

wastes are disposed of in the low lying

comes in 6000-10000 m buffer distance.

areas where the tanks and ponds are

Table 1 shows that as the distance to

located and due to this malaria vectors

health facilities increases, malaria area

very easily developed and many cases of

are also increasing, except in >10000 m

malaria

buffer zones (7.71% of malaria area

polluted pond water. In this it were

only). Here, in these area may be malaria

found that 44.7% of malaria are occurred

breeding sources are not developed as

within