Background
Dengue, a mosquito-borne viral disease of humans, is becoming a major global public health concern and socioeconomic burden, especially in tropical and subtropical countries because of the increasingly high incidence of infection [
1]. This includes the Southeast Asian nation of Malaysia that occupies parts of the Malay Peninsula and the island of Borneo. In this country of 32 million people there has been a recent steep rise in the number of dengue cases reported annually, from 46,171 in 2010 to 83,849 in 2017 [
2,
3]. The largest ever outbreak was reported in 2014, with 108,698 confirmed cases, which equated to an incidence rate of 361 per 100,000, and 215 deaths [
4,
5]. The incidence of dengue varies between the 13 states and three federal territories that comprise Malaysia due to several associated risk factors. According to current national statistics on dengue, Sabah, located in East Malaysia on the northern portion of Borneo, recorded a cumulative 3615 dengue cases between December 2019 and June 2020, making it the state with the fourth highest incidence, ranking only behind the more developed states of Kuala Lumpur, Selangor and Johor [
4]. Patterns of dengue transmission in Sabah are associated with a rapid rate of urbanization in close proximity to disturbed forest environments, thereby providing a risk of spill-over of sylvatic pathogens to human populations [
6].
In order to combat the escalation of dengue in Malaysia considerable measures have been undertaken to prevent or limit the risk factors of infection among the population. The most powerful prevention strategy is through health education and community participation. It is crucial to assess local communities on their knowledge, attitude and practice (KAP) towards dengue prevention practices so as to achieve an effective health intervention. While the current investigation is the first of its kind to be performed in East Malaysia several questionnaire-based studies conducted recently in Peninsular Malaysia (separated from Borneo by the South China Sea) have indicated a significant association between KAP and dengue infection. For instance, persons who possess a high level of knowledge of dengue demonstrate a significantly better attitude and practice towards prevention measures [
7]. There is also contradictory evidence to indicate that having knowledge of dengue is not certain to make an individual adopt recommended preventive behaviours [
8]. Such a discrepancy may be explained by several factors including a person’s lack of exposure to a dengue awareness program [
9], their misconception due to inadequate knowledge [
10] or traditional beliefs [
11], as well as unreliable/improper methodology of data collection or questionnaire use [
12].
It has been shown that measurement properties are population-specific and vary depending upon the setting and context of the questionnaire used [
13]. An existing questionnaire prepared for one population may not be applicable to another in a different country or region with a distinct set of socio-cultural, health system, economic and political contexts. For instance, Sabah located in east Malaysia has a diverse multiethnicity of more than 40 ethnic groups, within which there are over 200 sub-ethnic groups each having its own language, culture and belief system. Additionally, according to the most recent national census (2015), Sabah has the highest non-Malaysian population (25.6%) [
14]. Interestingly, the hotspots for highest prevalence of dengue in the Sabah area are occupied predominantly by the non-Malaysian population. This unique combination of Malaysian and non-Malaysian ethnic groups makes Sabah an advantageous site for dengue investigation. There is no gold standard questionnaire for the assessment of KAP due to the heterogeneity that exists in each population. Thus, the development and validation of a KAP questionnaire that is appropriate for the Sabah population would be extremely beneficial. However, a tried and tested survey validation process for Malaysian populations is currently lacking, particularly for use among the Sabah population, and so this also requires evaluation and documenting.
The majority of surveys that are conducted by questionnaire are reported either as a piloted, pre-tested questionnaire or a Cronbach’s alpha single-administration test score reliability coefficient as a measure of internal consistency. Thus, a rigorous methodology is required to examine the degree an instrument is affected by measurement error (reliability) and by the construct it intends to measure (validity) [
15]. The types of validity include concurrent validity (measuring the degree that it purports to measure) and criterion validity (measuring the degree to which the results of the questionnaire are an adequate reflection of a gold standard). Hence, with the aim to assist in planning an effective health intervention towards dengue infection, an accurate measurement of psychometric properties of the KAP questionnaire that is specific to the Sabah population is needed. The present survey describes the procedures involved in the development and assessment of the reliability and validity of this KAP questionnaire.
Discussion
To the best of our knowledge this is the first report to describe the development and assessment of content validity, face validity, constructs validity and reliability of a KAP questionnaire for the resident population of Sabah, Malaysia, in order to inform the planning of an effective local health intervention for dengue infection. Overall, the results of the IRT analysis for knowledge and the EFA for attitude and practice indicated that the measurement model for each construct should undergo modification to improve the model fit. The final version of the KAP questionnaire consisted of 25 knowledge items, eight attitude items and 11 practice items, each of which is reliable and valid for the Sabah population.
The IRT analysis of the knowledge section found several items to be problematic in terms of their relevance to measuring knowledge of dengue infection. As such, those items were excluded from further analysis. The ideal parameter range for discrimination values is from minus infinity to plus infinity; however, questions with negative figures of discrimination are recognized as problematic because they infer that participants with a high score are less expected to support more stringent response alternatives [
28]. The items retained in the final model were based on unidimensionality, where the P-value is more than α (> 0.05). All 25 retained items (out of 44) in the knowledge section had acceptable difficulty and discrimination values. Moreover, the average Cronbach’s alpha score was 0.75, indicating that the knowledge section of the questionnaire may be considered as reliable [
29]. This value is comparable to those obtained from previously published validation studies on KAP in Malaysia [
30,
31].
For the attitude section, eight out of 15 items were retained, revealing close relations between factors and items [
32]. Six items were excluded from further analysis as very low h2 values indicated that they are not correlated with other items that explain attitudes towards dengue prevention. In addition, one item was excluded in spite of having the minimally accepted factor loading value because the overall Cronbach’s alpha score increased when it was removed from analysis. The reliability analysis of items retained in the attitude section indicated an acceptable Cronbach’s alpha value of 0.83, demonstrating internal consistency. This internal consistency for attitude was higher than that obtained in a prior study performed in Perak state, Malaysia (Cronbach's alpha = 0.638) [
33].
In regard to practice, 11 out of 18 items were retained as each having an acceptable factor loading value (> 0.4) and corrected item-total correlation value (> 0.5). In some KAP survey guidelines it is recommended that developers start with a large number of items and apply item reduction techniques to select a small number of final items [
13]. The 11 items retained in the practice section were reliable with acceptable a Cronbach’s alpha score of 0.86 demonstrating internal consistency. This value is higher than that attained in the one previous study reported in Malaysia, undertaken in Perak state (0.79) [
33].
Lastly, the treatment-seeking behaviour section showed poor psychometric properties. The factor analysis was not suitable, while the reliability analysis indicated an unacceptable Cronbach’s alpha score, demonstrating low internal consistency. Hence, this section was removed from analysis.
The present study has several limitations. Firstly, the participants were recruited only from Sabah, Malaysia, as the questionnaire was intended to be used specifically for the Sabahan population. Thus, cross-validation studies are needed in other parts of East Malaysia, as well as extended to Peninsular Malaysia, for application of the questionnaire among other communities. Secondly, IRT and EFA were used in this study to assess reliability and validity; however, it is recommended that in future confirmatory factor analysis should be conducted to validate the EFA model. Thirdly, the initial design of the questionnaire allowed respondents to choose a “don’t know” response for any item. In a number of cases, this action may have excluded them from being taking into account when calculating a total score, which in turn prevented a better determination of the KAP. In retrospect, a “best guess” response option is more helpful than a non-answer to diagnosis, and therefore we have made this minor modification to the questionnaire.
Although this is a convenience sampling, in order to ensure that the data collection was representative respondents were recruited from across the entire state. Ethnicity is the foremost sample characteristic that may not be representative. Sabah is multiethnic, with ethnicity distributions that are not equal among the state's five divisions. For example, Rungus ethnics are more prevalent in Kudat division, Kadazandusun, Brunei, Bajau, and Sabah Malay in West Coast division, while Bugis and Suluk ethnics are located predominantly in Tawau & Sandakan division. In addition, there are smaller minority groups in Sabah that are defined simply as “others” in Malaysian census data. Hence, it cannot be excluded that the study's ethnicity distribution is not representative of the Sabah population as a whole. Such non-representativeness may have biased the outcomes. We acknowledge the particular importance of representativeness when estimating item discrimination and difficulty as both may depend on the specific sample.
Conclusion
In this study a new KAP questionnaire on dengue was developed and validated for the Sabah population in Malaysia, which may serve as a survey reference for researchers wishing to study Sabahan communities. In our validated KAP questionnaire, the final model has four sections and comprises 50 items, consisting of six items on sociodemographic information, 25 items on knowledge, eight items on attitude and 11 items on preventive practice. Evidence from IRT analysis and EFA indicates that the knowledge, attitude and practice sections are psychometrically valid, while each also demonstrates good reliability. However, the psychometric properties of the treatment-seeking behaviour section are unsatisfactory, suggesting that further development of this section is warranted in future studies. Moreover, further investigation, such as confirmatory factor analysis, is warranted to confirm the KAP scales.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.