This study was performed in urban areas of La Réunion. The Island is divided into 4 geographic sectors (North/South/East/West), 24 municipalities and 273 neighbourhoods. These neighbourhoods are divided into 960 zones used by the ARS for Aedes albopictus surveillance and control. These homogenous zones are defined according to urban planning and environmental criteria, they extend over 275 km² or 11% of the Island, and mostly cover urban areas. According to the National Institute of Statistics and Economic Studies (or INSEE) in France, an urban area is defined as an agglomeration of more than 2000 residents where no dwelling is separated from the next closest dwelling by >200 metres [12].

Home-owners generally allow LAV officers of the ARS to enter into their private dwellings to carry out routine vector-control activities. We commenced our sampling technique by selecting all zones controlled by the ARS on a minimum of three occasions between 2007 and 2011.We used zones located near the coast (less than 500 metres in altitude) where the presence of mosquitoes from one year to the next is most likely to persist and to ensure a relative homogeneity in environmental factors. Next, we focused on zones that showed relative stability in mosquito density between 2007 and 2011and we classified these as either negative or positive zones using criteria based on the Breteau Index (i.e. the number of positive containers per 100 houses). Using the historical data gathered over this 5-year period, we defined a negative zone as one that had a value for the Breteau Index at LAV routine visits that was lower than 50% of the average Breteau Index value during the same month and year in all zones. A positive zone was defined as having a value for the Breteau Index during LAV routine visits that was higher than 50% of the average Breteau Index value during the same month and year in all zones. A zone was kept for inclusion in the study if it was classified as positive or negative a majority of times during all LAV routine visits (e.g. classified as ‘positive’ twice during three routine visits conducted between 2007 and 2011). This was done to ensure a degree of stability in the classification of zones through time. A total of 184 zones, 68 positive zones and 116 negative zones, were identified using this methodology.

Being that face-to-face interviews in 184 zones were not feasible, a two-stage cluster random sample was drawn from this first selection. In the first stage of this two-stage sampling technique, a random sample of 26 zones (13 positive and 13 negative) was drawn taking into account the geographic distribution of the population on the Island in the four sectors (North/South/East/West). Next, households were randomly selected to achieve a fixed sample of 40 households per zone. The selection of households was undertaken while LAV officers were in the field. All households were randomly chosen and surveys were conducted in locations routinely checked by vector control officers. For each zone, the officers were asked to interview residents living in alternate households. The households were selected in this way while walking through the zone. The LAV officers commenced the survey starting at the four corners of the zone and walked in varying directions that were also chosen at random. The percentage of absentees and refusals varied considerably between zones; the average percentage of absentees was 36% (range of 5% to 50%), and the average percentage of refusals was 12% (range of 4% to 24%). The reason most often quoted for refusing to participate in the survey was a lack of time to respond to the questionnaire. For each selected household, the LAV officers conducted both a face-to-face interview that was addressed to the head of the household as well as an observational survey of the outside of the dwelling itself. The data collected were validated by comparing key characteristics of our sample to information provided in the latest census (e.g. number of household members by age of the head of household, level of education of the head of household, socioeconomic status). No significant differences were found for these key characteristics, an indicator that our sample was representative of the population in La Réunion. Conversely, a significant difference for the Breteau Index measured during the month of interviews (i.e. July 2012) was found between the positive and negative zones in the sample (Mann-Whitney test p-value<0.001). This finding confirms that the survey provides a rather accurate picture of the long-term average density of Aedes albopictus in the selected zones (i.e. our sampling technique resulted in the selection of zones that retained their characteristic classification of vector density through time).

Questions were derived from existing literature on protective behaviours against mosquitoes [13], [14], [15], [16], [17], [18]. Due to the circulation of dengue virus at the time of the study and an epidemic alert level up to 2B issued by the ARS during the month of April 2012, questions on the risk and perception of dengue were also integrated into the questionnaire. A pilot study was launched in June 2012 to test the validity of the questionnaire. The main study was carried out in July 2012, and was conducted according to the rules established by the National Data Protection Authority. Informed consent to answer the 40-minute questionnaire and to allow a LAV officer to conduct an observational survey of the residence was obtained verbally from all participants at the beginning of the interview. Translators were used when necessary.

Analysis was performed using STATA/SE v11.0 (StatCorp, College Station, TX). Descriptive information on behaviour, knowledge, and perceptions related to mosquitoes in our study sample was extracted using univariate and bivariate analyses. However, extrapolating from a simple average expenditure per household, as well as the relationship between expenditure per household and Breteau index, could be confounded by other household- or zone-level cofactors (presence of mosquitoes at home, wealth index, education amongst others), resulting in biased estimates at the Island level. In light of this we decided to analyse the influence of household- and zone–level characteristics on expenditure on protective measures (i.e. direct expenditure at the household level) using multivariate regression analysis. Simple generalised linear regression analysis was first performed using household-level expenditure on protective measures as the outcome variable. All factors significantly related to household-level expenditure were then entered into the final multiple regression model, using a random effects model. Provided that random effects are uncorrelated with the fixed predictors in the model, a random effects model is preferable, as it allows for the consideration of both household- and zone-level characteristics in a single model [24]. As the number of zones is relatively low (26), we didn't perform a mixed multilevel analysis, following Scherbaum et al recommendations (in which case a minimum of 30 level 2 zones is recommended to perform multilevel analysis) [25]. Lastly, univariate sensitivity analysis was conducted to determine how sensitive our estimates of household expenditure were to variation in input parameters, including perceived risk of a new epidemic of chikungunya, age, wealth, and education. Analysis was performed on wealth quintiles distribution across the population (from 100% in poorest category to 100% in richest category), educational levels (from 100% in the primary education category to 100% in college and higher category), age (20 to 99 years old), and perceived risk of a new epidemic of chikungunya (from all people considering that there is a low risk to 100% considering that there is a high risk). We also tested the influence of excluding air-conditioning and fans from the expenditure calculation. Results are displayed graphically using a tornado diagram.

A total of 1024 households in La Réunion were interviewed over the month of July 2012. Figure 2 provides the location of zones interviewed on the Island and Table 2 provides descriptive statistics of the main factors included in the multivariate analysis.

Only about 40% percent of respondents could identify the female as the biting gender in mosquito species. When asked to identify diseases transmitted by mosquitoes, the mean score for respondents in the study sample was of 3.73 correct answers out of 5. Not surprisingly, 92% of interviewees knew that chikungunya is a disease transmitted by mosquitoes, and another 78% answered correctly for malaria. More than a quarter of the respondents answered incorrectly that influenza is a disease transmitted by mosquitoes.

Most respondents declared that mosquitoes were present in their homes (77%). However, when asked about the frequency of mosquito bites over the last 7 days, 93% claimed that they were seldom or never bitten by mosquitoes. In spite of the low level of mosquito bites, 90% of interviewees considered mosquitoes a nuisance and 63% stated that these insects were of no particular use. When asked why mosquitoes were a nuisance, more than 80% replied that this was due to their role in transmitting diseases, 80% stated that mosquito bites and itching were important reasons, and 65% stated that it was due to the noise they created.

When questioned about the risk of epidemics, 47.7% of interviewees perceived that the risk of a new outbreak of chikungunya was reasonable or high. On the other hand, the risk of a dengue epidemic was perceived as being reasonable or high by 52% of respondents. This higher perceived risk of dengue could be explained by the increased attention given to this disease during the time of the study when a small epidemic of dengue was unfolding across the Island. Interestingly, however, when an open question was posed about which diseases could be transmitted by mosquitoes, only 20% of respondents spontaneously quoted “dengue”.

With respect to overall perception of vector control efforts, approximately 70% of respondents were confident that it is possible to reduce the number of mosquitoes. However, 20% insisted that nothing could be done in this regard. More than 75% of interviewees think that science could make further advances in the field of vector control. In terms of the acceptability of different measures to control mosquito numbers, 97% find the elimination of stagnant water acceptable, 82% feel that measures to repel mosquitoes are acceptable, and 74% accept techniques that prevent the reproduction of mosquitoes. Insecticide spraying is deemed acceptable by 65% of the study sample.

When asked whether they eliminate sources of stagnant water in and around their households, 97% of respondents declared that they did this. The frequency of this behaviour varied, however, with 17% claiming that they would eliminate these sources at least once per day, 45% declaring that they did this a few times per week, 11% stating that they did it a few times per month, and about 2% stating once per year. About 23% of the study sample claimed that they had definitively eliminated potential sources of stagnant water by removing empty containers or other potential recipients from their surroundings.

Among the various measures listed in the questionnaire, mosquito coils emerged as the most commonly used protective measure in this study, with 69% quoting that they used this measure at the time of the study. Insecticide/mosquito repellent sprays for the house (53%), non-electric diffusers (38%), and repellent creams and sprays applied to the skin (36%) were less frequently used by the sample during this period. Although use of these measures persists, about 50% of respondents stated that they consider them to be either reasonably or very dangerous to one's health, and another 24% felt that these products were not really effective. According to the health recommendations for travellers [20], mosquito coils are amongst the least effective measures to repel Aedes mosquitoes, intra-domiciliary insecticides and electric diffusers were found to have limited and weak effectiveness, and repellent creams and sprays applied to the skin were judged to be of stronger effectiveness. Therefore, the utilisation pattern of protective measures in this study population appears to be directly inverse to the recommendations of the health authorities [20] in terms of product effectiveness.

The lists provided by respondents of the measures they use to protect themselves against mosquitoes were used to determine the average household-level expenditure on these measures during the study period. This average expenditure per household was estimated to be USD 18.09 during the month of July 2012 and the median expenditure was estimated at USD 15.54. We tested the robustness of these household-level expenditure estimates by verifying whether these correspond with the results of a direct and closed question on household expenditure in the questionnaire. Respondents were asked to specify whether they judged their monthly spend on protective measures against mosquitoes to be less than EUR 10 (USD 12.24), between EUR 10 (USD 12.24) and EUR 20 (USD 24.48), EUR 21 (USD 25.70) to EUR 40 (USD 48.96), or more than EUR 40 (USD 48.96). Just over 60% of respondents declared spending between USD 12.24 and USD 24.48 per month on protective measures against mosquitoes. Taking the middle point of the range of each category, we found an average expenditure of USD 13.60, when using the results of this direct and closed question. A direct declaration of monthly expenditure (i.e. a categorical variable) is judged to be less reliable and accurate compared to the estimation of expenditure based on a list of used products (continuous variable) which is why the latter, i.e. the estimates of household-level expenditure, have been used as the dependent variable in the regression analysis.

Table 3 summarises the findings of the multivariate analysis that shows household-level expenditure on protective measures against mosquitoes as the dependent variable. Wealth quintiles, age (p = 0.039), and an educational level above a college degree (p = 0.087), were found to be positively and significantly associated (at the 10% level) with household-level expenditure on protective measures. Gender of the head of the household was not found to influence expenditure.

Knowledge on diseases transmitted by mosquitoes is shown to positively affect household expenditure on protective measures but the coefficient of this variable just misses the statistically significant threshold of 10% (p = 0.141). Correct knowledge of the biting gender of the mosquito species or the distance travelled by mosquitoes had no significant impact on expenditure on protective measures. In terms of perceived risk of a potential epidemic, the perceived threat of another chikungunya outbreak was found to be significantly associated with expenditure on protective measures. The potential threat of a dengue epidemic, however, did not have the same effect. Perception of the effectiveness of protective measures was found to influence household expenditure on these items only for those who stated that these measures are ‘effective’ (p = 0.021).

The regression analysis shows that while the declaration of having mosquitoes in one's household does not appear to influence expenditure on protective measures, an estimation of the frequency of mosquito bites given by respondents, on the other hand, has a significant and positive relationship with expenditure on products. There is a clear relationship between the perceived biting frequency (a measure of subjective exposure to mosquitoes) and the degree of spending on protective measures (the response to this exposure).

In terms of objective evidence of mosquito density, the number of positive Aedes breeding sites measured in and around the household does not appear to influence expenditure, and it is in fact the vector density for the entire zone (measured using the Breteau index), that has a positive impact on household spending on protective measures (p = 0.004).

Assuming that the positive relationship between the Breteau Index and household spending exists for all seasons of the year, it was possible to extrapolate the results of this study to make household expenditure estimations for the months between July 2011 and July 2012. A prerequisite for this extrapolation was data on the seasonal variation of the Breteau Index over this one-year time-frame, which were obtained using ARS records of monthly-measured Breteau index (publicly available on a monthly basis on the ARS website [26]). When asked about mosquito nuisance 72% of participants claimed to be affected by this mostly during the Austral summer (November to April), and another 42% declared that their use of protective measures would increase during the year, a plausible finding being that this study was conducted during the month of July. The indices recorded by the ARS were compared with the results from our regression analysis (provided in Table 3). We used predictions of the model at fixed values of ARS-measured Breteau index and averaged over the remaining significant covariates to estimate an annual global expenditure for households on the Island. Results for both the average expenditure per household and for all households in Réunion are given in Figure 3 and Table 4. In this study, the definition of a household has been limited to persons living within houses. The total number of households in La Réunion in 2009 was 284,391, as measured by the National Institute of Statistics and Economic Studies in France (INSEE). Among them, INSEE estimates that around 71% reside in houses. This corresponds to a total of 201,917 households (as defined in this study) on the Island of La Réunion.

The analysis shows that the predicted average household expenditure for July 2012 is USD 9.86, which is a better estimate of household expenditure than that derived from the list of used products as explained in the descriptive statistics section. This figure is also more accurate because it takes into account confounders included in the regression analysis. The projected total annual expenditure per household was estimated to be USD 138.92 (when using the survey definition of a household). If we include all households on the Island and assume that residents living in dwellings other than houses have no mosquito-related expenditure (i.e. taking a conservative approach), the total annual expenditure per household is USD 98.63 (the lower bound estimate of the total annual expenditure per household).

Overall, the amount spent by 201,917 households in Réunion from July 2011 to July 2012 on personal protective measures is estimated to be USD 28.05 million (95% CI of USD 25.58 million to USD 30.76 million). By comparison, other authors have found that the chikungunya epidemic resulted in medical expenses of up to EUR 43.9 million (USD 53.74 million) [27]. The annual Gross Domestic Product (GDP) of La Réunion was estimated to be EUR 14.42 billion (USD 17.65 billion) in 2009. Hence, the household-level expenditure related to protective measures in La Réunion amounts to 0.15% of its annual GDP in 2009.

The ARS already dedicates about EUR 10 million (USD 12.24 million) annually to the vector-control service or LAV, which is only a part of the overall budget being spent by municipalities, associations, and other actors to ensure vector control on the Island through various activities. The estimated population of the La Réunion Island in 2009 was about 816,360 inhabitants. ARS expenditure per person on vector control for 2009 was thus about USD 14.99/inhabitant (ARS expenditure does not represent total public expenditure on mosquito prevention and control – this is a value that we cannot capture, due to the role played by a number of different entities/authorities that in some way affect mosquito density on the Island). The annual GDP (Gross Domestic Product) per capita of La Réunion was EUR 17,884/inhabitant (USD 21,890/inhabitant) in 2009.

When compared to the expenditure of the vector-control service of the ARS, the total annual expenditure on protective measures for all households in La Réunion (i.e. USD 28.05 million) appears to be disproportionately high, a finding that has been pointed out by another study [28].

In order to test the sensitivity of our expenditure estimate to variation in the main factors influencing household-level expenditure, a univariate sensitivity analysis was carried out, as stated in the methods section. The factors influencing expenditure included perceived risk of a new epidemic of chikungunya, age, wealth, and education. We also tested the influence of excluding air-conditioning and fans from the expenditure calculation. The results are displayed graphically using a Tornado diagram in Figure 4. As guidance to interpret the findings in the diagram, for instance, age structure of the population has a large impact on estimated expenditure whereas variation in perceived risk of a new epidemic of chikungunya is unlikely to result in savings at the Island scale. The fact that the results are not strongly affected by the exclusion of air-conditioning and fans can be explained by the fact that the prices of these products were discounted in the main analysis.

The general limitations of this study are related to the fact that our expenditure estimates could be underestimated or overestimated. Being that the study was carried out in July which is a less favourable month for mosquitoes, it is probable that our household expenditure is under-estimated (conservative bias). Moreover, we have excluded from the study residents living in high-rise buildings, apartments and flats. This is an additional source of underestimation for our expenditure estimate. However, it is also possible that the expenditure estimates have been overestimated as participants may have reported using more protective measures than they actually do in reality and because the measures they use target mosquito nuisance in general and not only that created by Aedes albopictus, the focus of our study. In La Réunion the name “Tiger mosquito” is not commonly used and it is probable that the majority of people cannot distinguish between different species of mosquitoes. This is a fair assumption considering the low level of general knowledge on mosquitoes we found in this study sample. This assumption should be tested in future studies in La Réunion by specifically asking participants whether they can correctly identify Aedes albopictus, which has been found to be the case in other parts of the world, including the South of France [38]. Our inclusion criteria allowed us to include zones located in specific geographical settings (i.e. located near the coast (less than 500 metres in altitude) where the presence of mosquitoes from one year to the next is most likely to remain stable. This is another source for potential overestimation of household expenditure. Using these inclusion criteria could have resulted in a slight overestimation of the average Breteau index in our sample compared to the Island average. Indeed, in terms of risk indicators, the average Breteau index in our sample is 42.28 [37.92–46.63]. The estimate provided by the ARS in 2012 for the whole Island is 30.82 [27.19–34.44] and 38 [34.37–41.62] in 2011. Nevertheless, this slight overestimation of the Breteau Index should not have much influence on the estimates for household-level expenditure obtained from the regression analyses. This is because we have a variety of situations represented through the positive and negative zones used in this study.

Our household-level expenditure estimations may also be limited by the fact that we have no longitudinal survey to confirm usage patterns of products across time. In addition to this, using model predictions with average values of covariates can lead to over-estimation of expenditure when this is made on the scale of an Island.

In this article, we were particularly interested in expenditure on chemically-based protective measures or repellents (including insecticide-treated nets). We did not focus on ecological interventions (such as eliminating mosquito breeding sites and stagnant water) or other protective behaviours (such as limiting outdoor activities), information which would be valuable when measuring opportunity costs due to time spent on these activities or the impact of mosquito nuisance on quality of life. This could be the subject of another study in the future.

Another limitation is that this cross-sectional study cannot fully explain why people use some chemical measures that are judged to be ineffective by French public authorities. It is probable that habits or tradition continue to play a role in the use of certain measures to repel or kill mosquitoes in La Réunion, for example the use of fire. Identifying the reasons for these continuing behaviours and the use of measures that may actually prove to be ineffective requires a qualitative study approach that would capture more information than is possible through use of standard questionnaires and quantitative methods.

Lastly, a key lesson from this cross-sectional survey is that longer-term research should be undertaken in order to take into account seasonal variations in protective behaviours against mosquito nuisance and disease threats in order to provide more robust conclusions.

Differences in mosquito control practices at the local level involve the interplay of place, scale and politics [39]. This study is one of the first attempts to quantify household-level expenditure on protective measures against mosquitoes, a very important step considering that community involvement is considered to be at the heart of vector-control strategies in La Réunion and elsewhere. More importantly, longer-term studies on this subject, as well as studies on the effectiveness of different products, can be instrumental in determining potential savings at the household-level due to improvements in public messages and the introduction of new policies or interventions that are currently considered as being too expensive. Finally, it is evident that household-level behaviour is differentially affected by subjective and objective measures of exposure to Aedes albopictus. Both variables need to be taken into account when explaining the use of chemically-based protective measures against mosquitoes and any related variations in expenditure.