Background
Lyme disease (LD), or Lyme borreliosis, is the most frequent vector-borne disease in temperate countries [
1]. In most cases, the disease causes non-specific flu-like symptoms and a typical skin lesion known as
erythema migrans. More severe systemic infections can occur in some cases, and may lead to arthritis, cardiac and neurological problems [
2]. With a recent adjusted estimation of about 300,000 human cases annually in the United States [
3] and about 85,000 cases in Europe [
1], Lyme disease is a growing concern in many countries, including Canada where it is currently emerging. In the province of Québec (Canada), locally acquired cases were first identified in 2008 [
4]. Populations of black-legged ticks (
Ixodes scapularis), the only known vector of Lyme disease on the eastern-American coast, are now recognized as established in the southern part of the province, in the Montérégie region. In this region, 8-13% of the black-legged ticks have been found to be infected with
Borrelia burgdorferi, the bacteria causing LD [
5]. In Switzerland, LD cases have been reported for more than 30 years [
6]. The disease has not been notifiable since 2003, but current estimates place this country third highest for LD incidence in Europe with 83 cases per 100,000 inhabitants reported in 2010 [
7]. In this country,
Ixodes ricinus is the vector responsible for the transmission of LD, and prevalence of
Borrelia burgdorferi infection in ticks is as high as 40% in some regions [
8]. Moreover, in several regions of the country, these ticks are known to carry tick-borne encephalitis virus (TBEV), the agent of tick-borne encephalitis (TBE), another severe and notifiable disease in Switzerland [
9].
Although LD ecology differs in Europe and North America with regards to the importance of different reservoir species of the bacteria and the primary tick vector involved in transmission [
10], the main preventive strategy is the same in both regions and relies primarily on individual-level preventive behaviors [
11]. Preventive behaviors such as checking for ticks after visiting affected wooded regions, wearing long trousers or repellent containing DEET, have been shown to be efficient in the prevention of LD [
12‐
18]. However, beyond their demonstrated efficacy, studies have also shown that people do not apply these measures with the same consistency, even in highly prevalent regions [
19‐
29].
Predictors of individual-level preventive behaviors have been studied for many health conditions. The Health Belief Model is one widely used theoretical model developed to study health behaviors. In this model, one main determinant of a health behavior is the perception of risk, defined as the subjective assessment of the probability and the consequences of a specified type of hazard [
30]. Risk perception is composed of the perceived severity of and the perceived susceptibility to the disease in question [
31]. This model has been validated for many diseases and health conditions including LD, for which a higher level of risk perception was associated with an increased adoption of preventive behaviors [
19,
28,
29,
32‐
34]. As a result, risk perception has become a major point of interest for decision-makers involved in the design and the implementation of preventive communication programs. An extensive literature exists on risk perception, Paul Slovic being a pioneer of the psychometric approach, which recognizes that risk perception is a construct reflecting individual and social level influences [
35‐
37]. Studies have shown tendencies which seem to persist among different fields of research: the perceived risk in the general public differs from the risk as perceived (or evaluated) by experts [
38]; determinants of risk perception are numerous and multidimensional, and they include characteristics of the hazard in question such as the novelty of the hazard and its potential catastrophic impacts, as well as individual and sociological factors, such as gender, age, education, income, personality, culture and values [
35,
39‐
41].
Past studies have described LD risk perception in particular regions or countries [
19,
21,
22,
25,
27‐
29,
32], but none have explored the differences between the determinants of risk perception in different epidemiological contexts, such as in a population experiencing the emergence of LD versus a population that has been living in a region endemic for LD during a long period of time. Are determinants of LD risk perception universal, or do they vary according to the context, such as the epidemiologic situation? The identification of context-specific determinants of LD risk perception would provide additional insights for decision-makers in the planning of LD risk communication that could be better adapted to emerging or endemic situations. Moreover, it could help decision-makers in emerging contexts to anticipate the changes in their population’s risk perception that may occur once LD becomes endemic.
With this perspective in mind, the main objective of this study was to compare risk perception of LD and to describe its determinants within and between two different populations: residents of the Neuchâtel canton, in Switzerland where LD has been endemic for more than 30 years, and residents of the Montérégie region, in Québec, Canada, where LD is emerging and where the indigenous cases were first reported in 2008. A second objective was to compare perceptions of the general population with perceptions of regional LD experts, and between experts from both regions. Estimated LD incidence in the Neuchâtel canton ranged from 49 to 95 cases per 100,000 inhabitants by 1996–2001 [
42,
43], which was above the national mean incidence for Switzerland. Montérégie had an estimated incidence of 0.5 cases per 100,000 inhabitants in 2012, making it the most affected region in the province of Québec (Canada) [
44].
Discussion
Risk perception of LD has been studied in the past, mostly as a predictor of individual level preventive behavior along with knowledge and other factors. The vast majority of these studies were undertaken in the United States [
19,
21,
23,
27‐
29,
32,
33,
45,
59], while more recent studies were done in the Netherlands [
24,
25] and in the United Kingdom [
22]. To our knowledge, this is the first study to measure risk perception regarding LD and its determinants in Canada and in Switzerland, and to address risk perception of LD in different epidemiological contexts with an international comparative perspective.
The populations of these two regions were different on several aspects. First of all, nearly half of the surveyed population in Montérégie had never heard about LD (46%), with only a minority of the participants (15%) demonstrating a good level of knowledge, whereas in Neuchâtel, more than 8 out of 10 people knew of the disease and close to 60% had a good level of knowledge of the disease. These differences may be due to several contextual factors including a longer experience with LD in Switzerland, where the disease is highly endemic, through public health messages, media coverage, social networks, personal history of infection and schools. Neuchâtel residents declared themselves as more often exposed through outdoor activities than in Montérégie, which also reflects the fact that most people in this region live near (if not ‘in’) the tick inhabited regions. Accordingly, the mean global risk perception score was higher in Neuchâtel. This is consistent with previous findings comparing risk perception in low and high incidence states in the United States and showing that risk perception was positively correlated to incidence of LD [
28,
29].
Looking at the same results, we can also highlight that a lack of knowledge about the risks of LD still persists in Neuchâtel despite the high regional incidence: 22% of the respondents declared they had never heard about the disease, 35% did not know that the disease was transmitted by a tick and three out of four did not know the difference between LD and TBE. A previous national study in the United States also reported that 7% of people had never heard about LD in high-incidence States, and that 22% declared that they did not know how LD is contracted [
29]. These results suggest that living in an endemic area established for a long time does not guarantee that the entire population will be aware of the risks and have sufficient knowledge of how to protect themselves. This underscores the need to adjust, strengthen and maintain communication efforts about LD risks even as the epidemiological situation evolves over time.
Some surprising findings arose. First, the mode of the
perceived regional susceptibility was found to be equal between regions even though the incidence was nearly 200 times higher in Neuchâtel compared to Montérégie for this period (95
vs 0.5/100,000). In Montérégie, the population
perceived regional susceptibility was greater than the expert’s
perceived regional susceptibility, who most often consider the risk to be low in this region. One possible explanation for this observation could rest on the novelty of the hazard for the Montérégie population. New threats frequently lead to higher perceived risk in the general population, as has been previously demonstrated in studies comparing risk perception between different kinds of hazards [
35].
Second, Montérégie respondents rated the risk for themselves (mode = 2, mean = 2.7) and the risk for the residents of their region (mode = 4, mean = 3.2) differently. The underestimation of the personal risk as opposed to the general population risk has been described before for other hazards and is known as ‘unrealistic optimism’ [
39]. Explanations for this optimism have been extensively studied before and are reviewed in Shepperd and colleagues [
60]. In Neuchâtel, this phenomenon is not observed. One possible explanation is that past history with LD among respondents or their relatives is more prevalent in this region. Personal experiences with a hazard has been shown to decrease unrealistic optimism [
61].
This study showed that the perceived risk of LD differed between the population and their regional experts. In Montérégie, experts rated the measured components of risk as smaller and more ‘controllable’ than the population. Many studies have demonstrated differences between public and expert risk perception for other hazards [
35,
38,
62] and this trend can be problematic when decisions have to be made about risk management options. Given that risk perception can affect the adoption of preventive behaviors, as well as the social acceptability of public health actions, our results suggest not only that risk perception of a hazard has to be taken into account when making such decisions, but also that risk perception should be measured directly in the target population, and cannot be extrapolated from studies carried out in different contexts, nor by regional experts. Because of the limited number of experts who participated in this study, statistical analysis could not be performed to compare perceptions between the population and experts and between both groups of experts.
One interesting aspect of this study lies in the use of EFA to build a global risk perception score for LD. Past studies of risk perception and LD have used individual perception variables such as the perceived susceptibility and the perceived severity of the disease as the dependant variable or as independent variables to predict preventive behaviors [
25,
28,
29,
32]. However, we hypothesised that risk perception is a complex construct that can only be imperfectly captured by individual survey questions. Most individual perception variables are correlated, and factor analysis can be used to verify the internal consistency of a set of questions designed to measure a construct (internal consistency) and to reduce the measurement bias related to individual questions by identifying which composition of items best represents a single factor (composite reliability) [
63]. Although identifying determinants of a global perception score could be more interesting for public health decision-making than focusing on individual perception variables, the use of EFA has been criticized, mainly because of the absence of objective criteria to guide decisions necessary to complete the analysis, particularly in the choice of the type of rotation of factors [
64]. In this study, no rotations were performed in the final model given that only one factor was retained. We used EFA to explore which perception variables to include in a global risk perception score and we interpreted the results in light of previous findings. The Health Belief Model recognizes two main dimensions of risk perception: the severity of the hazard and the susceptibility of individuals to this hazard [
31]. Empirical studies have underlined that individual susceptibility can be perceived differently than the susceptibility for the general population [
39]. These three dimensions
(perceived severity, perceived individual susceptibility, perceived regional susceptibility) were identified in this study, along with
feeling of worry, as the main contributors to a factor with the EFA realized in both populations and this strengthens the choice of these four variables in the construction of a global score.
Another important result of our study was the identification of different determinants of risk perception regarding LD in both populations, suggesting that the determinants may not be universal but rather context-dependant. The only common predictor was gender, a well-known determinant of risk perception. Possible explanations for gender differences in risk perception have been explored in the risk perception literature and include differences in social roles and activities [
65]. In Montérégie, the effect of age was also highlighted, where being less than 35 years old decreased risk perception. This effect has been demonstrated before for other hazards, particularly regarding risk perception of road accidents [
66]. One interesting finding is that in Neuchâtel, where the disease has been endemic for a long time, the level of knowledge was not significantly associated with risk perception, in contrast to the Montérégie region. It is both the exposure (living in a high risk region) and past history with LD that constituted the strongest predictors of risk perception. Only a handful of other studies have previously identified determinants of LD risk perception, being that the main focus of these other studies has generally been to identify predictors of the adoption of preventive behavior. Knowledge of LD [
24], knowing someone who has had LD [
24], the presence of tick populations [
59], and cultural identity [
21] have been identified before as factors that may affect risk perception of LD.
Globally, these results suggest that in populations facing an emerging threat such as LD in Montérégie, risk perception is mostly determined by globally available information. In the Montérégie context of LD emergence, risk perception seemed less affected by an individual’s specific circumstance, i.e. their exposure and past history with LD, than it was in the LD endemic region of Neuchâtel. This further suggests that the availability of reliable information becomes particularly important in a context of emergence. This comparison can provide useful insights for both Canadian and Swiss decision-makers, as well as for other countries facing a challenge of LD emergence. On the one hand, this study provides important information for local populations and on the other hand, international comparisons may allow us to understand what might occur in future epidemiological contexts.
Nevertheless, this study presents some limitations. First, by recruiting participants through two web panels, the population samples were not probabilistic and were restricted to internet-users. Generalisation of the results should be interpreted in consequence. The mean response rate was considerably low. Previous Canadian studies using the same Canadian panel had response rates around 20 to 25% (Léger & Marketing, personal communication). A wide variety of factors are known to affect the response rate of web surveys, such as the methods of delivery [
67]. For this study, the firm which administered the survey closed the survey access when 400 participants had completed the survey in a region, which took three days in Montérégie (response rate of 8.3%) compared to 12 days in Neuchâtel (response rate of 36%). A longer response period, especially in Montérégie, might have led to a better response rate. Though, these response rates depend in fact on the number of people who were initially contacted (5,222 in Montérégie vs 1,233 in Neuchâtel), and the mean response rate should be considered with regards to the recruitment process.
Secondly, all participants read a descriptive text before answering questions pertaining to perception. This was a strategic decision implemented in the study design with the objective of increasing the number of eligible respondents, particularly in Montérégie, where we expected that the majority of residents would not know enough about LD to complete the survey. But the content of this text may have altered participants’ perception and consequently, may have biased their ‘true’ perception (i.e. the perception they would have had without reading the text, influenced by the information they already had about the disease).
Another limitation is the cross-sectional design of this study. Measures of risk perception, such as psychometric variables, can change rapidly over time [
63]. Future work should include additional administrations of the risk perception questionnaire in the same regions in order to provide insights on the temporal evolution of risk perception and their determinants in both populations, and to allow confirmation of the risk perception factor structures.
Finally, regression models revealed interesting determinants in both regions, but explained only 12% and 11% of the variance in Montérégie and Neuchâtel, respectively. Even if these percentages are low, they are consistent with other psychometric studies of risk perception [
39]. When interpreting the multivariate analysis, we must keep in mind that several other possible unmeasured factors may have an impact on risk perception.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
CA conceived and designed the study and drafted the manuscript. AR and PM helped in the design and statistical analyses. LG helped in the conception of the study and in the design of the survey. FM, JPW and DB helped in the conception of the study. All authors read, revised and approved the final manuscript.