Journal of Biostatistics and Epidemiology

J Biostat Epidemiol. 2018;4(4): 222-231

Original Article

Partial Least Square Path analysis of knowledge, attitude and practice regarding dengue

Lamidi-Sarumoh Alaba Ajibola1,3, Shamarina Shohaimi1,2* Mohd. Bakri Adam2 Mohd. Noor Hisham Mohd. Nadzir1,
Oguntade Emmanuel Segun2
1
Department of Biology, Faculty of Science, Universiti Putra Malaysia, 43400 Serdang, Selangor, Malaysia
2
Institute for Mathematical Research, Universiti Putra Malaysia, 43400 Serdang, Selangor,Malaysia
3
Department of Mathematics, Faculty of Science, Gombe State University, P.M.B 127, Tundun Wada, Gombe, Nigeria

ARTICLE INFO ABSTRACT

Received 25.04.2018
Revised 27.10.2018 Introduction: Knowledge, attitude and practices regarding dengue are latent variables which are substantiated
Accepted 02.12.2018 through manifest variables. The manifest variables that form the indicative construct of knowledge, attitude
Published 20.12.2018 and practice can be factored into sub-constructs such that the impact of each indicative variable can be verified.
Method: Evaluation of the sub-constructs of knowledge, attitude and practices regarding dengue using a
Partial least square path models with R programming language.
Result: The measurement model revealed the sub-constructs that are negatively affecting the latent variables
Key words:
and the ones that are having low impact.
Knowledge,
Conclusion: This analysis gives the possibility of observing the exact knowledge, attitude and practices
Attitude,
regarding dengue that are inadequate among respondents. The result from this methodological approach can
Practices,
be used as an aid for the community health programs and campaigns on how to enlighten the populace of
Dengue,
interest on the required awareness about dengue, attitude towards dengue and the preventive practices that are
Path analysis,
deficient among them.

It is also sometimes used to establish the impact

Introduction
of health and vector control programs on a
Dengue fever (DF) is a viral infection spreading particular population. Presently, there is no
among humans through the bite of infected licensed drug for DF, suppressing the dengue
female Aedes mosquitoes (1). The specific virus episodes after infection remains vague; it is
vectors are Aedes aegypti and Aedes albopictus virtually depending on the severity of the
(2). Cartographic techniques used in exploring infection on the viraemic subject (4,5) The newly
spatial information on the prevalence and introduced dengue vaccine known as dengvaxia
incidence of DF had revealed that the spread of has many limitations such as cost-effectiveness
dengue could be as high as approximately 400 (6), the age limit of 9-45 years of age can only be
million per annum across the globe (3). vaccinated, the vaccine takes a period of one year
Knowledge, attitude and practice (KAP) to be completed with an interval of six months
regarding DF and its vectors had been known to and being vaccinated is not a substitute for
be an advantageous way of reducing the vectors protection against mosquito bite. Dengvaxia is
population and curtail the spread of the infection. also accompanied by some side effects such as

*
Corresponding author: Shamarina Shohaimi,Email: shamarina@upm.edu.my Phone number: +60192747525

Please cite this article in press as: Alaba. L.S, Ajibola, Shohaimi S, Bakri Adam M, et al. Partial Least Square Path
analysis of knowledge, attitude and practice regarding dengue. J Biostat Epidemiol. 2018; 4(4): 222-231
Partial Least Square Path analysis of knowledge, attitude and practice regarding dengue
Vol 4 No 4 (2018)

difficulty in breathing, low blood pressure or Materials and method

collapse in people who had no previous history of
DF (7). Aiming mostly at the dengue vectors to 2.1 Data and instrument
curtail the menace of dengue infection spread, Data used in this study were sampled from a
KAP regarding dengue vectors and infection real-life data collected from an intercept sampling
remain the cheapest and safest way of curtailing survey decisively to explain the how the Partial
possible dengue outbreak compare to the disease least square path analysis (PLS-path analysis)
risk, economic burden and possible eventual methodological approach can be used to improve
mortality after the incidence of the infection. KAP regarding dengue studies. The validated
During the 1950s and 1960s in Americas (8), structured bilingual questionnaire (Malay and
dengue vectors were effectively eliminated English language) used during the survey has 30
through an eradication program named
questions on knowledge which were factored into
“eradication of open source breeding site” and
6 sub-constructs, 13 questions on the attitude
another study from northern Thailand had also
were factored into 4 sub-constructs and 14
shown that the appropriate knowledge and
awareness, the right attitude and habitual questions on practices were also factored into 4
preventive practices against possible breeding sub-constructs. All the sub-constructs were
sites of dengue vectors had reduced the dengue formed based on factor analysis. Partial least
vectors and subsequent infection (9). square path analysis was used to verify the
weights of each sub-construct on the main
Presently, there is a break by the World Health construct via R programming language.
Organization (WHO) on the first dengue vaccine
2.2 Latent constructs and definition of sub-
because of the severe reaction caused by the
constructs
vaccine on people who had never had DF (10,11).
Curtailing the morbidity and mortality attributed Let x1 , x2 , …, xn be the manifest variables
to DF still lies within KAP regarding dengue which are used to indirectly measure approximate
although the prospect of KAP regarding dengue representation of the sub-constructs such that
studies may depend on the methodological each question xi is 1 or 0 and
approach used in analyzing the data collected. n = number of questions. The latent constructs
were considered to be caused by the sub-
constructs (cause-effect relationship).

𝑥1 … 𝑥3 𝑥1 … 𝑥5 𝑥1 … 𝑥5 𝑥1 … 𝑥5 𝑥1 … 𝑥7 𝑥1 … 𝑥5
Figure 1: Knowledge was measured by were also measured by varying manifest variables
formative sub-constructs and the sub-constructs of knowledge on dengue. The sub-constructs of

http://jbe.tums.ac.ir 223
Partial Least Square Path analysis of knowledge, attitude and practice regarding dengue
Vol 4 No 4 (2018)

knowledge are staidness of dengue infection symptoms (SAS), elimination and biting time
(SD1), the severity of dengue fever and vectors (EBT).
(SD2), possible breeding site (PBS), primary and
secondary transmission (PST), signs and

𝑥1 … 𝑥3 𝑥1 … 𝑥4 𝑥1 … 𝑥3 𝑥1 … 𝑥3

Figure 2: Attitude was measured by constructs of attitude are staidness of dengue

formative sub-constructs and the sub- infection and prevention (SD3), elimination
constructs were also measured by manifest of dengue vectors (EDV), infection and re -
variables of attitude towards dengue. The sub - infection (IAR) and consci ousness (CON).

Practice
s

E E P P
LM AM MB PT

𝑥1 … 𝑥3 𝑥1 … 𝑥4 𝑥1 … 𝑥3 𝑥1 … 𝑥4

Figure 3: Practices is measured by formative are the elimination of larval mosquitoes (ELM),
sub-constructs and the sub-constructs are also elimination of adult mosquitoes (EAM),
measured by the manifest variable of practices prevention from mosquitoes bite (PMB), and
regarding dengue. The sub-constructs of practices protective practices (PPT)

http://jbe.tums.ac.ir 224
Partial Least Square Path analysis of knowledge, attitude and practice regarding dengue
Vol 4 No 4 (2018)

S
D1
S
D2
P Knowledg
BS e suppressing
T
ST EL
S M
AS EA
E Practic M
BT es
P
MB
S
D3 P
PT
E Attitude
DV suppressin
I

Figure 4: Path diagram depicting the knowledge, attitude and practices regarding dengue model.

Figure 4 showed that the ellipses are the inner The arrows represent the relationship between the
models or structure showing the relationship variables.
between the latent constructs. The rectangles are The two simple KAP model was formulated
the outer model showing the relationship between based on the concept of KAP studies:1). The
each construct and its indicative sub-constructs better the knowledge and attitude, the better
which form blocks for each of the latent variables. practices and 2). The better the knowledge, the
better the attitude.

𝑃𝑅𝐴𝐶𝑇𝐼𝐶𝐸𝑆1 = 𝑓( 𝐾𝑁𝑂𝑊𝐿𝐸𝐷𝐺𝐸1, 𝐴𝑇𝑇𝐼𝑇𝑈𝐷𝐸1 ) (2.2.1)

𝐴𝑇𝑇𝐼𝑇𝑈𝐷𝐸1 = 𝑓(𝐾𝑁𝑂𝑊𝐿𝐸𝐷𝐺𝐸1) (2.2.2)

The algorithm of PLS on KAP regarding dengue

Start: Answered validated structured KAP questionnaires regarding dengue
Step 1: Factor the indicative sub-construct for each latent variable
Weight =1 for the correct answer
Weight = 0 for the wrong answer
Step 2: Sum each of the correct sub-construct independently
Step 3: Standardize the indicative sub-constructs scores
Step 4: Calculate the outer approximation of latent variables
Step 5: Obtain the inner weights

http://jbe.tums.ac.ir 225
Partial Least Square Path analysis of knowledge, attitude and practice regarding dengue
Vol 4 No 4 (2018)

Step 6: Calculate the internal approximation of the latent variables

Step 7: Obtain new outer weights
Repeat the last four steps until convergence of the outer weights
End.

2.3.1 The inner model matrix

2.3 Framework for PLS-PM for KAP The inner model is mostly referred to as
regarding dengue with R structural model; it represents the connection of
Step 1: Install plspm and colortools the latent variables somehow depicting a causal
Install.package (“plspm”) process or a network that can be represented in a
Install.package (“colortools”) path matrix format. The path matrix must be a
After installation, the library can be used to square matrix (number of rows is equal to
load the package library (“plspm”) columns) and must also be a lower triangular
Step 2: Import your data Boolean matrix. The elements in the diagonal and
data (KAP) above it must be zero but the element below the
Figure (1), (2), (3) elucidates the variables and diagonal can either choose the value of zeros or
their descriptions ones.
In order to accomplish a PLS-path model with Step 3: Definition of the inner matrix
the functions of plspm(), 3 elements had to > KNOWLEDGE1 = c (0,0,0)
be in place 1) data set, 2) the inner model and 3) > ATTITUDE1 = c (1,0,0)
the outer model. > PRACTICES1 = c (1,1,0)

The matrix is created by row binding

> KAP_path = rbind (KNOWLEDGE1, ATTITUDE1, PRACTICES1)
> colnames(KAP_path) = rownames(KAP_path)
> KAP_path
KNOWLEDGE1 ATTITUDE1 PRACTICES1
KNOWLEDGE1 0 0 0
ATTITUDE1 1 0 0
PRACTICES1 1 1 0
2.3.2 The outer model
The matrix is read by columns affecting rows. The outer model specifies the set of indicative
In this case “Knowledge affect Attitude and sub-constructs that form each blocks of the latent
Practices” and “Attitude affects Practices” the variables (KAP)
zeros in the diagonal means that no latent variable Step 4:
can affect itself (PLS-path models is solely meant Definition of the indicative sub-constructs
for recursive models). associated with the latent variable to form the
> innerplot(KAP_path) to outer model.
visualize the inner model. > KAP_blocks = list (1:6,
7:10, 11:14)

http://jbe.tums.ac.ir 226
Partial Least Square Path analysis of knowledge, attitude and practice regarding dengue
Vol 4 No 4 (2018)

The list above has three blocks; one per each Step 5: Definition of the vector mode
block represents the indicative sub-constructs.
There are two possible modes in
The first block (1:6) corresponds to the latent
plspm()namely Mode “A” and Mode “B”.
variable knowledge (Figure 1) which is
Mode A means that the indicative sub-constructs
associated with the first six column of the data
are reflective (cause-effect from the latent
set. The second block (7:10) is associated with
variable). Mode B means the indicative sub-
attitude (Figure 2) formed by column 7 to 10 of
constructs are formative (cause-effect to the
the data set and lastly, the third block (11:14) is
latent variable).
associated with practices (Figure 3) formed by
In this case, all are latent variables are
column 11 to 14 of the same data set.
formative, thus the required vector mode is;
KAP_modes = c ("B", "B", "B")
Step 6: Run plspm analysis
>KAP_pls =plspm(KAP,KAP_path, KAP_block,modes = KAP_modes)
>KAP_pls

Results and Discussion a formative indicator requires comparison of the

outer weights in order to determine which of the
Examining a PLS- Path model involves two
indicative sub-constructs contributes mostly to
phases: Assessment of the outer model
the latent constructs, in other words, correlation
(measurement model) and the inner model
of the “loadings” are not required (10).
(structural model). Assessing the outer model of
3.1 Plot the measurement model
>plot(KAP_pls, what = "weights")

Figure 5: Measurement model showing the weights of each indicative sub -construct on the latent
variables (red arrows signify negative influence while blue arrows signify positive influence).

From Figure 5, weights on KNOWLEDGE1 latent variable which implies that the knowledge
showed the two of indicative sub-constructs (PST on the primary and secondary transmission (PST)
and SD2) that are having a negative impact on the and knowledge on the severity of dengue fever

http://jbe.tums.ac.ir 227
Partial Least Square Path analysis of knowledge, attitude and practice regarding dengue
Vol 4 No 4 (2018)

and vectors (SD2) are inadequate among the construct that have positive contribution to the
respondents. SD2 has the weight of -0.8684 latent variable is the attitude towards infection
which means knowledge on primary and and re-infection (IAR) although positive but very
secondary transmission of dengue (PST = -0.087) low compare to attitude about the elimination of
is inadequate among the respondents but not as dengue vectors (EDV).
much the knowledge on the severity of dengue
Weights on PRACTICES1 indicated that
fever and vectors (SD2). In descending order,
practices on the elimination of the larval
SD1, PBS, EBT and SAS had a positive impact
mosquitoes (ELM) and protective practices
on the latent variable of KNOWLEDGE1.
(PPT) are inadequate among the respondents
Weights on ATTITUDE1 also revealed the two because of their negative impact on the latent
indicative sub-constructs (SD3 and CON) are variables. The positive indicative sub-constructs
having negative impact on the latent variable are protection from mosquitoes bite (PMB) and
which means that attitude towards how serious the elimination of adult mosquitoes (EAM). The
dengue infection could be and prevention (SD3) two positive sub-constructs (PMB and EAM) are
and consciousness of being infected with dengue already known among the respondents although
(CON) is inadequate among the respondents. EAM was lower among the respondents than
Comparing both negative sub-constructs, CON PMB.
has less weight than SD3. The indicative sub-

3.2 Plot the structural model

>plot(KAP_pls)

Figure 6: Structural model sho wing the relationship between the latent variables.

>KAP_pls$inner_model to extract the (2.2.2), the model ATTITUDE1 as a function of

estimates which can be used to calculate the KNOWLEDGE1 can be estimated with the
scores of the latent variables. From equation equation below

http://jbe.tums.ac.ir 228
Partial Least Square Path analysis of knowledge, attitude and practice regarding dengue
Vol 4 No 4 (2018)

ATTITUDE1 = 4.761375𝑒 −16 + 6.391013𝑒 −01 (KNOWLEDGE1 ) + 𝑒𝑟𝑟𝑜𝑟 (4.2.1)

$ATTITUDE1
Estimate Std. Error t value Pr(>|t|)
Intercept 4.761375e-16 0.1812839 2.626474e-15 1.000000000
KNOWLEDGE1 6.391013e-01 0.1812839 3.525416e+00 0.002416401

From equation (2.2.1), the model and ATTITUDE1 can be estimated with the
PRACTICES1 as a function of KNOWLEDGE1 equation below:

PRACTICES1 = −1.111388𝑒 −15 + 1.402388𝑒 −01 (KNOWLEDGE1)

+ 8.074686𝑒 −01 (ATTITUDE1)
+ 𝑒𝑟𝑟𝑜𝑟 (4.2.2)

$PRACTICES1
Estimate Std. Error t value Pr(>|t|)
Intercept -1.111388e-15 0.1039191 -1.069475e-14 1.000000e+00
KNOWLEDGE1 1.402388e-01 0.1351138 1.037931e+00 3.138408e-01
ATTITUDE1 8.074686e-01 0.1351138 5.976212e+00 1.502835e-05

>KAP_pls$inner_summary[, "R2", drop = FALSE] to extract the coefficient of determination 𝑅 2

R2
KNOWLEDGE1 0.0000000
ATTITUDE1 0.4084505
PRACTICES1 0.8164141

The estimated model from equation (4.2.1) means that the amount of variability in of
has a coefficient of determination 𝑅 2of 0.41 PRACTICES1 can be explained by 82% of
which means that the amount of variability in KNOWLEDGE1 and ATTITUDE1 of the
ATTITUDE1 can be explained by 41% of respondents but ATTITUDE1 has higher
KNOWLEDGE1 of the respondents.The influence than KNOWLEDGE1 (Figure 6).
estimated model from equation (4.2.2) has a >KAP_pls$effects to extract the direct, indirect
coefficient of determination 𝑅 2 of 0.82which and total effects.The direct effects are the path

http://jbe.tums.ac.ir 229
Partial Least Square Path analysis of knowledge, attitude and practice regarding dengue
Vol 4 No 4 (2018)

coefficient (figure 7), the above extraction will

also give indirect effects and total effects.

relationships direct indirect total

1 KNOWLEDGE1 -> ATTITUDE1 0.6391013 0.0000000 0.6391013
2 KNOWLEDGE1 -> PRACTICES1 0.1402388 0.5160542 0.6562930
3 ATTITUDE1 -> PRACTICES1 0.8074686 0.0000000 0.8074686

Among the latent variables of the structural considering the effects and weights of the
model, it is was shown that the relationship indicative sub-constructs on the latent variable
between KNOWLEDGE1 and PRACTICES1 has which may assist to project into the future of
indirect effects which make the total effect of community health and vector reduction programs
their relationship to be 0.6562930. The total on dengue.
effect of KNOWLEDGE1 to ATTITUDE1 and
What was known before?
ATTITUDE1 to PRACTICES1 remains the same
Majority of KAP regarding dengue studies (14)
as the direct effects because there were no
based the scores of the latent variables on the
indirect effects existing between both
overall composite score, analysis and conclusions
relationships. Furthermore, the relationship
were made without verifying the impacts of the
between KNOWLEDGE1 to ATTITUDE1 and
sub-constructs.
ATTITUDE1 to PRACTICES1 were significant.
The novelty of the study
Conclusion
Indicative sub-constructs were factored and
scored independently; their impacts on the main
The conceptual motive for assuming all the latent
latent variables were verified by using Partial
variables was formed by their indicative sub- least square path model analysis in a formative
constructs was based on the fact that there are mode
some KAP regarding dengue that is already
famous among some populace while some are References
still lacking. With the formative latent variable,
1.Gubler DJ. Epidemic dengue / dengue
the extractions of the weight of each indicative hemorrhagic fever as a public health , social and
sub-constructs on the latent variables were economic problem in the 21st century. Trends
verified. In conclusion, it is not always Microbiol. 2002;10(2):100–3.
2.Ponlawat A, Scott JG, Harrington LC.
appropriate to add up all the test scores of latent
Insecticide susceptibility of Aedes aegypti and
variables together and make conclusions without Aedes albopictus across Thailand. J Med

http://jbe.tums.ac.ir 230
Partial Least Square Path analysis of knowledge, attitude and practice regarding dengue
Vol 4 No 4 (2018)

Entomol [Internet]. 2005;42(5):821–5. Available from:

from: http://dx.doi.org/10.1080/14760584.2017.12768
http://www.ncbi.nlm.nih.gov/pubmed/16363166 31
3.Murray NEA, Quam MB, Wilder-Smith A. 11.Robinson ML, Durbin AP. Dengue vaccines:
Epidemiology of dengue: Past, present and future Implications for dengue control. Curr Opin Infect
prospects. Clin Epidemiol. 2013;5(1):299–309. Dis. 2017;30(5):449–54.
4.Malavige GN. Dengue viral infections. 12.Kaliyaperuma K. Guideline for Conducting a
Postgrad Med J [Internet]. 2004;80(948):588– Knowledge, Attitude and Practice (KAP) Study.
601. Available from: Community Ophthalmol [Internet].
http://pmj.bmj.com/cgi/doi/10.1136/pgmj.2004. 2004;IV(1):7–9. Available from:
019638 http://v2020eresource.org/content/files/guideline
5.Rothman AL. Immunity to dengue virus: A tale _kap_Jan_mar04.pdf
of original antigenic sin and tropical cytokine 13.Sanchez G. PLS Path Modeling with R
storms. Nat Rev Immunol [Internet]. [Internet]. R Package Notes. 2013. 235 p.
2011;11(8):532–43. Available from: Available from:
http://dx.doi.org/10.1038/nri3014 http://gastonsanchez.com/PLS_Path_Modeling_
6.Flasche S, Jit M, Rodríguez-Barraquer I, with_R.pdf
Coudeville L, Recker M, Koelle K, et al. The 14.Lamidi-Sarumoh AA, Shohaimi S, Bakri
Long-Term Safety, Public Health Impact, and Adam M, Noor Hisham Mohd Nadzir M,
Cost-Effectiveness of Routine Vaccination with a Emmanuel Segun O. Systematic review of
Recombinant, Live-Attenuated Dengue Vaccine knowledge, attitude, and practices regarding
(Dengvaxia): A Model Comparison Study. PLoS dengue in Malaysia. J Appl Pharm Sci [Internet].
Med. 2016;13(11):1–19. 2018;8(12):80-091. Available from:
7.Khunsha Fatima, Syed NI. Dengvaxia http://www.japsonline.com
controversy: impact on vaccine hesitancy. J Glob
Health. 2018;8(2):8–10.
8.Schliessmann DJ. Aedes aegypti eradication
program of the United States--progress report
1965. Am J Public Health Nations Health
[Internet]. 1967;57(3):460–5. Available from:
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC
1227178/pdf/amjphnation00071-
0078.pdf%0Ahttp://www.pubmedcentral.nih.go
v/articlerender.fcgi?artid=1227178&tool=pmcen
trez&rendertype=abstract
9.Benthem BHB Van, Khantikul N, Panart K,
Kessels PJ, Somboon P, Oskam L. Knowledge
and use of prevention measures related to dengue
in northern Thailand. Trop Med Int Heal.
2002;7(11):993–1000.
10.Aguiar M, Halstead SB, Stollenwerk N.
Consider stopping dengvaxia administration
without immunological screening. Expert Rev
Vaccines [Internet]. 2017;16(4):301–2. Available

http://jbe.tums.ac.ir 231