A Mixed-Methods Approach to Understanding Knowledge of Mosquito-Borne Infections and Barriers for Protection in Hanoi, Vietnam

Dengue is a growing problem in Hanoi, with cyclical epidemics of increasing frequency and magnitude. In June 2019, we conducted a cross-sectional survey using mixed methods to investigate how inhabitants of Hanoi perceive and respond to the risk of mosquito-borne diseases (MBD). A total of 117 participants recruited using a stratified random sampling method were interviewed in three districts of Hanoi. Knowledge and practices (KP) regarding MBDs were assessed using a pre-tested questionnaire. Inferential statistics were used to identify factors associated with KP scores and describe the relationship between variables. Additionally, a “risk-mapping” exercise was conducted in a subsample through semi-structured interviews and analyzed qualitatively and quantitatively using the System Effects platform. Factors significantly associated with knowledge scores were education and family history of MBDs. While knowledge and practice scores were found to be positively correlated in the statistical analysis, this was not corroborated by our observations on the field. The results also revealed gaps in knowledge about MBDs and vectors and highlighted a general feeling of powerlessness which prevented the adoption of protective behaviors. Therefore, educational interventions which provide concrete tools to empower communities should have a positive impact on improving vector control.


Introduction
With 3.97 billion people living in at-risk areas [1], dengue is the most widespread mosquito-borne disease. Recent decades have witnessed a dramatic re-emergence of dengue fever and dengue hemorrhagic fever worldwide-with South East Asia accounting for more than 70% of cases [2], Vietnam is among the countries bearing the highest burden [3,4]. In Hanoi city, rapid urbanization and massive population growth, along with insufficient mosquito control, have contributed to the increase in dengue vector population, leading to more frequent and cyclical epidemics [5,6]. The largest recorded outbreak occurred in 2017, when 36,354 cases and 7 deaths were reported [7].

Mapping Exercise
A mapping exercise was conducted through semi-structured interviews for system effects modelling [22]. Participants were provided a template with the focus question "Why do I get bitten by mosquitoes?". They were asked to write down the factors they thought were related to the risk of getting bitten and indicate the connections between them by arrows. This resulted in directed graphs representing a network of factors centered on the focus question. The exercise was pre-tested using different framing of the focus question in order to assess comprehensibility and refine the formulation.

Statistical Analysis
Descriptive statistics were used to analyze respondents' demographic characteristics. The nature of the data distribution was assessed using the Kolmogorov-Smirnov test. Mann-Whitney U tests, Kruskal Wallis tests and negative binomial regression were used to identify factors associated with KP scores, while Spearman's rho was used to describe the relationship between K and P scores. Categorical variables were expressed as percentages, continuous variables as the mean ± standard deviation and discrete variables as the median ± interquartile range (IQR). A p-value < 0.05 was taken as significant for inferential statistics. Factors with a p-value > 0.1 in univariate analysis were not included in the negative binomial model. All analyses were performed using STATA 15.0.

System Network Analysis
Individual graphs constructed during the mapping exercise were replicated on the System Effects platform (https://systemeffects.com/#/admin) developed by Luke Craven, Northumbria University, and translated into adjacency matrices [22]. Language and concepts of the determinants were homogenized in order to condense individual matrices into one aggregated adjacency matrix which was used to generate a directed network graph in Gephi (https://gephi.org). This graph was visually explored to investigate the relationship between factors and identify potential barriers perceived by the community. Additionally, measures of centrality and ranking of determinants by in-and outdegree were used to identify those with the highest connectivity within the network.

Demographics
A total of 117 questionnaires were completed. Socio-demographic characteristics of the respondents are presented in Table 1. Of the 117 respondents, 57.3% were females and 43.7% were males; the mean age was 52.13 ± 13.95, ranging from 18 to 83. In total, 86.3% had at least secondary education. The most common occupation was farming (34.2%) and 22.2% were retired.

Assessment of Knowledge
The median score for knowledge was 7 ± 4 (med ± IQR) ranging from 0 to 18. The proportions of answers to selected questions are shown in Table 2. The majority of respondents were aware of dengue (82.1%). However, 33.3% did not know any symptoms of MBDs. Symptoms frequently mentioned included fever (61.5%), hemorrhage (45.3%) and rash (14.5%). Most respondents recognized polluted (59.8%) or stagnant water collections (62.4%) as potential mosquitoe-breeding sites but only 1.7% also mentioned clean water collections, which are preferred by Aedes sp. [6]. In total, 76.9% correctly identified either summer or rainy season as high-risk periods. The main source of health information was television (65.8%), followed by local public communication by loudspeaker (36.8%). In contrast, health workers (9.4%) and social media (3.4%) seemed to play a minor role in informing about MBDs.  * more than one answer can be reported.

Assessment of Practices
The median score for practices was 3 ± 1 (med ± IQR), ranging from 0 to 7. The proportions of various preventive methods reported by the participants are shown in Table 3. The most commonly used were bed nets (83.8%) and insecticides (65.8%). When asked about the frequency of elimination of both natural and domestic breeding sites such as water containers, discarded tires and various wastes, 76.1% reported eliminating breeding sites at least weekly and 49.6% daily. Personal protective measures such as coils, repellents and covering clothing were rarely mentioned. * more than one answer can be reported.

Association Between KP Scores and Demographic Variables
Results of the statistical analyses are presented in Table 4. Factors with a p-value < 0.1 in univariate analysis were further assessed using binomial regression to control for confounding. Education (p-value = 0.001) and family history of MBDs (p-value = 0.04) were the two factors found to be significantly associated with knowledge scores. No factor was found to be associated with practices.

Correlation Between Knowledge and Practices
Spearman's rank correlation indicated a strong positive correlation between knowledge and practices (Spearman's rho = 0.6161, p < 0.001).

System Network Analysis
In total, 31 individual graphs were constructed during semi-structured interviews and aggregated into one graph representing the participants' perception of risk factors for getting bitten by mosquitoes ( Figure 1). In addition to its visual exploration, degree and centrality measures presented in Table 5 and Figure 2 showed that the most influential factors identified by the participants were the presence of breeding sites in the environment, the lack of protective measures and the lack of hygiene. Factors related to the lack of protection included misuse of bed nets, poor knowledge about prevention, low risk perception and an attitude of negligence. Overall, our discussions with the participants highlighted a fatalistic attitude regarding disease control: while a majority believed they were at risk of getting infected, they perceived the presence of mosquitoes in the environment as something natural and did not think they could have control over it. Most of them did not feel responsible for eliminating breeding sites and blamed the neighboring households for not performing doing so properly. Additionally, some participants believed mosquitoes could not bite during the day and therefore neglected to use bed nets when sleeping at noontime. not performing doing so properly. Additionally, some participants believed mosquitoes could not bite during the day and therefore neglected to use bed nets when sleeping at noontime.

Discussion
Our study highlighted gaps in knowledge of inhabitants of Hanoi about dengue disease and vectors. Although most of the surveyed population were aware of dengue, one-third of the respondents were unable to give details of any symptoms and most of the remaining respondents only mentioned fever. This is likely to contribute to underestimating the occurrence of dengue fever and result in people not seeking adequate care. Mistaken beliefs also resulted in improper protective behaviors and limited use of personal protection, with a minority of respondents being aware that mosquitoes could bite during the day and reproduce in clean water collections, which could possibly reflect a confusion with the malaria vector Anopheles sp. A previous study conducted in Laos in 2009 revealed that one-third of respondents believed dengue and malaria were transmitted by the same vector and could not differentiate between the two diseases [23]. This misconception could contribute to the implementation of inappropriate Aedes control measures. Our results also suggested limited engagement in spreading health information by health workers.
In agreement with previous studies, we found the level of education and history of MBDs to be

Discussion
Our study highlighted gaps in knowledge of inhabitants of Hanoi about dengue disease and vectors. Although most of the surveyed population were aware of dengue, one-third of the respondents were unable to give details of any symptoms and most of the remaining respondents only mentioned fever. This is likely to contribute to underestimating the occurrence of dengue fever and result in people not seeking adequate care. Mistaken beliefs also resulted in improper protective behaviors and limited use of personal protection, with a minority of respondents being aware that mosquitoes could bite during the day and reproduce in clean water collections, which could possibly reflect a confusion with the malaria vector Anopheles sp. A previous study conducted in Laos in 2009 revealed that one-third of respondents believed dengue and malaria were transmitted by the same vector and could not differentiate between the two diseases [23]. This misconception could contribute to the implementation of inappropriate Aedes control measures. Our results also suggested limited engagement in spreading health information by health workers.
In agreement with previous studies, we found the level of education and history of MBDs to be associated with knowledge [21,24,25]. Unlike most of those studies, we also found a positive correlation between knowledge and practices. However, this was contradicted by our observations in the field: although almost half of the participants reported eliminating breeding sites daily, we were able to collect mosquito larvae on most household premises (results not shown), particularly in rural and peri-urban areas, where the presence of rubbish and tires provided multiple breeding sites. This result is therefore likely to be affected by declaration bias. The influence of knowledge on the adoption of preventive behaviors has been inconsistently described in the scientific literature, with the majority of studies failing to demonstrate any link between a community's knowledge and better practices [25][26][27][28]. Knowledge does not always translate into more adequate behaviors, and risk perception has been shown to be a key component of the adoption of protective practices. A study in Malawi has shown for example that showing freshly captured mosquitoes to participants was more effective than educational leaflets in improving their understanding of the risk and increasing the use of bed nets [29].
The "risk-mapping" exercise allowed us to better understand the perception of risk and barriers preventing the implementation of protective measures. To our knowledge, this is the first time that system effects modelling was employed to explore the factors related to the perceived risk and prevention of mosquito bites. Overall, it highlighted a fatalistic attitude towards disease control combined with a lack of personal responsibility. The majority of participants did not believe that they have control over the risk and perceived the presence of mosquitoes in the environment as inevitable. This corroborates health-belief models, which have shown that perception of self-efficacy is a crucial component of the translation of risk perception to a concrete preventive behavior [28,30]. This can also be related to the findings of a previous study in Martinique, which interpreted the participants' reluctance to apply dengue prevention despite a good risk perception as the result of a "normalization" process and questioned the effectiveness of risk communication for community mobilization in endemic areas [28]. In addition, most participants blamed their neighbors or health authorities for improper management of breeding sites rather than taking responsibility for it. Similar attitudes were described in a study conducted in Hanoi in 2017, where most people would rely exclusively on health authorities for implementing vector control [31].
Our study had several limitations. Firstly, all interviews were conducted during the day when most workers were absent, which can explain the high proportion of retired individuals and limit the representativity of the sample. Prevention practices were self-reported and therefore declaration bias cannot be ruled out, as demonstrated by our contradictory observations on site. Selection of participants for the "mapping exercise" relied mostly on convenience sampling. The exercise would also have benefited from a preparatory training session to give participants more autonomy and limit the influence of the interviewer when constructing the graphs.

Conclusion and Recommendations
Our study demonstrates the existence of barriers to implementing vector control measures in Hanoi, which included improper knowledge of vector biology and disease features, low self-responsibility and low perceived self-efficacy. In order for risk awareness to translate into actual protective behaviors, we recommend the adoption of participatory approaches to engage communities and encourage compliance by demonstrating the feasibility and effectiveness of vector control measures, in coordination with local authorities.