In this study, we conducted a retrospective ecological assessment for the effectiveness of crowdsourced data to detect established tick populations in Quebec, Canada, from 2020 to 2023. Here, an established tick population refers to a locally reproducing population considered persistent over time. The criteria used to infer establishment followed the Institut national de santé publique du Québec’s (INSPQ) risk classification approach [51], itself based on [52] (see section Active Tick Surveillance for details). The unit of analysis was the municipality, defined as CSDs, totaling 1282 across the province [53], representing the finest spatial scale at which all relevant data were available. Approximately 80% of the province’s population of 8 million lives in metropolitan areas, concentrated in the south, with the most populated being Montréal, Québec City, and Sherbrooke [54]. The province has seen ticks and LD emerging gradually over the last two decades. The most endemic regions, with the longest history of emergence, are the southernmost regions of Montérégie and Estrie, near the US border [55,56]. Since initial emergence, endemic regions have progressively expanded across southern Quebec, and human LD cases have been constantly rising. The dominant forest types are deciduous and mixed forests, and the potential exposure to ticks varies depending on land use and human activities, ranging from peridomestic and rural environments to recreational areas, such as natural parks in urban settings.

The eTick platform collects routine passive surveillance data on ticks in Canada. On the basis of the nature of the data collected and the collection process, and acknowledging that no formal research question was specifically formulated to acquire these data, the Bishop’s University Research Ethics Board stated that no ethical review is required for data collection through the eTick platform (Multimedia Appendix 1). Here, we make a secondary use of nonidentifiable data; individuals cannot be identified from any images or materials in the paper or supplementary files, a case defined in the Canadian research ethics guidelines for research involving humans (TCPS 2) as not requiring ethical review [69].

Before submission, all eTick users must consent to the use of their deidentified data for surveillance purposes and for future unspecified research [70]. Personal information (name, username, email address) is collected exclusively for communication purposes and visible only to site administrators. Data extracts used for surveillance or research do not include personal information in their data fields. No compensation was provided to participants.

All analyses were conducted in R version 4.5.1 (R Foundation for Statistical Computing) [71].

This study demonstrates that crowdsourced eTick data can be used to develop robust, evidence-based indicators of blacklegged tick population establishment in Quebec. Although we did not detect spatial clusters in submission rates, we found that the rates were associated with socioecological factors, including climate suitability, population size, landscape configuration, and population socioeconomic characteristics. The results also highlighted key sources of variation that must be accounted for when deriving indicators from crowdsourced surveillance data. Building on these findings, we developed and tested indicators based on eTick submissions that were effective in detecting CSDs with established tick populations while accounting for confounding factors. Overall, the results support integration of crowdsourced surveillance data into the Canadian tick-borne disease surveillance framework as a complementary tool to traditional active surveillance. Previous studies using eTick data have applied different approaches to predict tick distribution or public health risk, providing useful points of comparison for the present findings. Bowser et al [89] assessed whether early eTick data (2018‐2020) could predict LD risk at the CSD level, in an endemic region of Quebec, Canada. They reported weaker associations and poorer model fit when eTick submissions were used as LD risk predictors compared to other indicators, attributing these results to the novelty of the platform at the time and variable levels of public adoption and participation. More recently, Westcott et al [90] used eTick submissions with other citizen science datasets to develop niche models predicting the current and future distribution of I scapularis. Both studies relied on raw submission data and did not explicitly account for socioecological factors influencing participation and reporting intensity, which may have limited risk signal detection in Bowser et al [89] and biased habitat suitability estimates in Westcott et al [90]. While these studies illustrate the potential utility of crowdsourced surveillance data, the lack of consideration of these factors may represent an important limitation. Indeed, results from this study indicate that eTick submission counts are influenced by both human and environmental factors, including population size, socioeconomic characteristics, and habitat suitability affecting tick occurrence and abundance. Submission rates likelihood reflect exposure risk among humans and companion animals as driven by behavior and tick presence and density, as well as factors influencing the decision to submit samples to eTick. Accordingly, the value of these data for contributing to surveillance programs depends on accounting for biases affecting submission rates.

Submissions to eTick were distributed across southwestern and central Quebec, corresponding to the known distribution range of I scapularis in the province [91]. Although no statistically significant spatial clusters were detected, there was geographic heterogeneity in submission rates for both submission types. Submission of ticks found attached to humans were mostly reported in the southernmost regions of the province, which are also the highest risk regions for LD (Outaouais, Montérégie, and Estrie [92]). In contrast, submissions of ticks found attached to animals were distributed across a wider geographic area, including regions located outside the currently defined endemic zones. Several of these regions include municipalities with lower risk levels for LD or are not considered endemic for LD according to the provincial risk classification [91]. Under this classification, municipalities are considered endemic when they meet one or more criteria regarding the number of declared human LD cases, the number of ticks collected through passive surveillance, or the number and stages of Borrelia burgdorferi–infected ticks collected in active surveillance [93]. Taken together, these patterns highlight the broader geographic scope of detections obtained through animal-origin submissions and suggest that combining submissions from both origins may provide complementary information for surveillance of tick presence beyond established endemic areas. As eTick data continue to accumulate over time, there may be sufficient statistical power to detect higher than expected submission rates, if present, which could help identify and test for currently unknown local factors (human and pet population, or tick) affecting exposure risk.

In zero-inflated negative binomial models analyzing the factors associated with tick submissions, the total human population of CSDs was the strongest predictor of both submission likelihood and count, indicating that more populated areas are more likely to report tick presence. This result aligns with previous studies, which have shown that submission counts are generally proportionate to population size, where the odds of submission increase with the number of potential contributors [18-20,40,41]. These findings reinforce the importance of accounting for population size when analyzing passive surveillance data [18,19]. Additional socioecological factors also influenced submission occurrence and counts.

First, municipalities that accumulated more annual DD >0 °C were more likely to have tick detections. Ticks require ambient temperatures above 0 °C to quest for hosts, with warmer conditions accelerating interstadial development [94]. Globally warmer conditions throughout the year are thought to accelerate the completion of their life cycle and facilitate their establishment. Indeed, previous modeling and field studies have shown that a threshold of approximately 2800 DD >0 °C is needed for ticks to establish viable populations [67,95]. While DD >0 °C was a determinant of the presence of tick populations, it did not significantly affect submission counts once populations were established. This suggests that once tick populations are well established, other factors such as host densities, human-tick encounters, or participation behavior may drive the numbers of submissions to eTick.

The proportion of WUI in CSDs was positively associated with tick submission counts. This variable measures the degree to which human settlements (ie, neighborhoods where people live, infrastructures where people work, or roads providing access to the territory) are surrounded by natural areas, including woodlands, ecotones, and prairies, which may be suitable habitats for ticks and their hosts [65]. Results indicate that CSDs with higher WUI represent higher tick exposure risk areas [45]. Conversely, the WUI indicator was negatively associated with the occurrence of tick submissions (as denoted by a significant positive relationship with the likelihood of structural zeros), which is a surprising result that warrants further analysis. Here, we studied a region that may not be entirely suitable for tick establishment, with very different biomes across the territory [5]. It is unknown, for example, if boreal forest provides a habitat where I scapularis populations can become established [96]. Consequently, this result may reflect another underlying relationship not captured by our current indicators or may relate to the coarse ecological nature of the indicator, which is aggregated at the municipal level and includes all types of natural habitats. Moreover, ticks require specific ecological conditions regarding habitat type, host availability, and landscape connectivity at fine spatial scales, which likely could not be detected at the municipal scale used in this work [45,97,98].

Finally, CSDs with higher median household income and age were associated with more submissions. This aligns with other findings that citizen science initiatives tend to predominantly attract middle-aged and older groups and those with greater access to educational and financial resources [35,37]. This is a common point of critique [37,99] and an area where improvement is needed through initiatives aimed at reaching more diverse groups. Doing so would help ensure that the benefits of participation are more equitably distributed across the population [31] and that the indicators derived from these data are more representative of all areas and population profiles. Our proxy of general population awareness of ticks and LD, constructed using Google Trends data, showed a weak but positive relationship with submission counts in CSDs. This aligns with expectations that public awareness of the problematics under focus is generally associated with higher participation in citizen science initiatives [31,37,100]. However, several limitations inherent to Google Trends may have contributed to the weak association observed. Keyword-based indicators rely on specific search terms and may not capture the broader range of queries related to tick information seeking [101]. Additionally, Google search activity can be strongly influenced by short-term, media-driven attention, which may not reflect sustained awareness or a durable willingness to engage with surveillance initiatives such as eTick [101]. More broadly, public health intelligence derived from open-access, user-generated sources, such as Google Trends, is subject to participation bias, with contributions disproportionately originating from populations with greater internet access and digital literacy. As a result, these data may not accurately represent the population at risk to the public health threat [102,103]. Furthermore, Google Trends data can be unstable or imprecise when analyzed at small spatial scales or in areas with low population sizes, such as many CSDs, in the study area [104]. Collectively, these limitations may reduce the validity of this indicator as a proxy for awareness, as noted in previous studies [101]. Integrating complementary data from diverse sources, ranging from traditional public health surveillance systems to nontraditional data, such as open-access social media, may help strengthen signal detection and mitigate participation biases [102].

To evaluate the ability to detect established I scapularis populations, we built and compared models using 1 of 3 crowdsourced surveillance indicators based on tick origin: ticks found attached to humans, animals, and from all origins combined. For each indicator, we compared simple models with more complex ones that accounted for socioecological factors associated with submissions in objective 1. In all cases, the complex models significantly outperformed simpler alternatives, including those accounting for population size or only using submission counts. While simple models can be useful for ease of interpretation and indicator development for public health use [18], the loss in performance here would be too high to justify this approach. This result highlights the relevance of analytically considering socioecological covariates when using passive or crowdsourced data, consistent with findings and approaches from other studies [20,41,44].

The 3 top-performing models yielded similar results: model A1, based on ticks found attached to humans and socioecological factors; model B1, based on ticks found on animals and socioecological factors; and model C1, based on all tick origins combined with socioecological factors. Given these similar performances and the complementarity observed in the geographic distributions of both submission types, we pursued subsequent analyses using the submissions from all origins to identify the threshold optimizing the detection of established tick populations. By doing so, we also ensured that all data collected by users were leveraged.

These analyses showcase a methodology for indicator development that can be adapted according to the objectives of the surveillance or research program and the epidemiological context. Previous Canadian passive surveillance studies have shown that ticks recovered from animals, particularly dogs, were effective at providing early signals and wider geographical extent of emerging tick populations [18-20,105]. This relationship has been attributed to differences in exposure, as dogs are generally more directly in contact with vegetation and leaf litter, where they are more likely to be bitten by questing ticks, even with low tick densities [13,14]. However, ticks recovered from animals may be less specific indicators than those from humans, as very low tick densities detected through animals may reflect adventitious ticks transported by migratory birds rather than established tick populations [18,19]. On the other hand, our analyses suggest that ticks from animals can be efficient at detecting emerging tick populations where active surveillance may not be possible or may be insufficient. Although active surveillance remains the most standardized method available for tick surveys, its efficacy depends on multiple factors, including vegetation type and weather conditions before and during sampling [106-108]. Drag sampling may be less effective for detecting emerging tick populations in northern Quebec, where coniferous forests and dense understory vegetation dominate, compared to deciduous forests in southern Quebec [106-109]. This highlights that both active and crowdsourced surveillance indicators may have strengths and limitations and that we would benefit from both by developing integrated indicators [15,100,110].

To identify submission thresholds that optimize tick population detection, we searched for cutpoints that maximized sensitivity, specificity, or both, using models based on submission types. In the best-performing model, a threshold of 3 submissions (rounded from 3.32) maximized both sensitivity and specificity for detecting tick populations. At this threshold, both sensitivity and specificity reached 76%. These levels are comparable to those of a previous study conducted in Quebec, which used passive surveillance to predict the presence of established I scapularis populations, as measured by active surveillance carried out from 2007 to 2008 [20]. That study identified an optimal threshold of 0.073 (Predicted Linear Index), which predicted the presence of ticks in the environment with 74.19% sensitivity and 73.68% specificity, using a similar methodology that maximized both indicators. Their model included passive surveillance data, an index of environmental suitability based on cumulative annual DD >0 °C, and an index estimating the number of ticks arriving on migratory birds. Overall, our results of crowdsourced surveillance shared similar characteristics and limitations with comparable traditional passive surveillance datasets from other Canadian studies: more adult female ticks were submitted than immature ticks, more ticks were recovered from animals than from humans, and similar factors affected the submission data [14,18-20]. The predictive performance of this dataset in identifying established tick populations, as confirmed by active surveillance, was consistent with that of traditional passive surveillance. Therefore, our analysis supports the validity of crowdsourced data as a complement to traditional active and passive surveillance. We believe these approaches should remain complementary, as each provides distinct information. Active surveillance provides standardized measures of tick density and confirmation of locally reproducing populations, while crowdsourced data offer broader geographic coverage and rapid detection of emerging areas [111], and traditional passive surveillance also more readily enables pathogen testing, which is generally not feasible with crowdsourced submissions. Future work should include an assessment of how these data should be used to build indicators that are adaptive to evolving epidemiological context, such as the one prevailing in Canada [11,105].

The optimal thresholds of submissions maximizing tick population detection accuracy were highly variable across the study area. This variability reflects the province’s pronounced heterogeneity regarding population size, land cover, climate, and socioeconomic characteristics. To facilitate interpretation, we exemplified what these thresholds would be across different groups of CSDs, regrouped by population size and climate suitability. Climatically suitable regions for ticks are concentrated in the southern parts of Quebec. These areas are well-known endemic zones for I scapularis and LD, including the highest incidence regions of Estrie and Montérégie [56]. Most CSDs in these regions have small population sizes and are suburban or rural. These areas also have a high proportion of WUI, meaning there is significant overlap between tick-suitable natural environments and human habitation. In such contexts, only a very small number of tick submissions was sufficient to indicate the likely presence of an established tick population. This suggests that even limited crowdsourced surveillance signals in these areas should raise awareness about possible reproducing tick populations and emerging risk of tick-borne disease [112]. In contrast, large population centers in climatically suitable areas required a higher threshold of submissions to balance sensitivity and specificity. In climatically unsuitable regions, a similar threshold was identified for small population centers, while the highest one was identified for medium population centers. These findings suggest that in emerging areas, a high number of ticks must be submitted before suspecting an established population, which may indicate that adventitious ticks are detected well before actual establishment occurs. Alternatively, it could reflect a mismatch between active and passive surveillance data in these regions. For example, active surveillance may fail to detect ticks in more northern environments due to limitations in sampling sensitivity, especially in the dense understory vegetation of coniferous forests [106,108], or because of lower sampling effort deployed in these areas. In contrast, passive surveillance may be less affected by these limitations, as it relies on ticks encountered by the public or professionals across a wider range of habitats and conditions.

This study has several limitations that should be considered when interpreting its findings. First, the spatial accuracy of eTick submissions implies an inherent level of uncertainty, as reported locations reflect the user’s presumed site of tick acquisition and cannot be assumed to be exact. These locations remain subject to recall and information bias, as users may have incomplete recollection of where exposure occurred or may infer locations based on routine activities. Spatial accuracy is expected to be higher when ticks are detected promptly after exposure or when individuals perform regular tick checks and are aware of their recent movements [28]. Excluding submissions from users reporting recent travel history mitigated the impact of this source of uncertainty; however, it was not possible to fully control for this limitation with the available data. Second, uncertainty in indicator precision may arise from spatial aggregation at the CSD level and from the use of active surveillance as an imperfect reference standard. Ecological processes relevant to tick establishment operate at finer spatial scales than those captured here, and active surveillance may fail to detect established populations under certain environmental or sampling conditions. Together, these limitations reflect the challenge of relying on relatively coarse spatial indicators. Likewise, although we attempted to account for participation biases to which crowdsourced initiatives are subject by including sociodemographic covariates in our modeling framework, residual bias likely remains. Finally, the generalizability of the proposed indicators beyond Quebec and over time will require further evaluation. Differences in surveillance systems, population engagement, ecological contexts, and public awareness across provinces or countries may affect indicator performance. Over time, the usefulness of a crowdsourced approach such as eTick depends on sustained participation from both the general public and professionals [113]. Since future engagement is unknown and may not remain stable over time, the continued usefulness of this surveillance method cannot be guaranteed. Regular reassessment will therefore be necessary to ensure data quality and long-term relevance.

In this study, we analyzed the patterns in blacklegged tick submissions to the platform eTick in Quebec, Canada. This work introduces a rigorous analytical framework for interpreting crowdsourced surveillance data by comparing submission rates with long-term active surveillance. A novel contribution is using a socioecological approach to adjust for potential participation bias in eTick submission rates. Unlike prior Canadian tick passive surveillance studies, which rely primarily on traditional surveillance methods or that use unadjusted crowdsourced data, our approach explicitly incorporates socioecological factors associated with reporting into the modeling framework. This work demonstrates that crowdsourced surveillance data, when appropriately modeled, can yield reliable inferences about disease vector distributions. It provides a methodological approach applicable to other medically important vector species and geographic contexts, and contributes to improving the robustness of ecological inference of vector distributions. Specifically, our findings show that eTick can effectively complement traditional passive surveillance methods for detecting established blacklegged tick populations. Although this analysis focused on I scapularis, eTick allows the collection of all tick species and could therefore support monitoring of other medically important species in Canada, such as Ixodes cookei, Dermacentor variabilis, Rhipicephalus sanguineus, or Amblyomma americanum. These contributions have practical implications for public health practice. More confident identification of tick distributions supports targeted public health actions and improved surveillance strategies. The broad geographic coverage of eTick makes it well suited for early detection of emerging tick-borne diseases in undersurveyed regions. Integrating eTick data into formal surveillance systems could significantly enhance Canada’s capacity to detect those emerging risks in a cost-effective way. Future work should focus on expanding the analytical framework to other tick species and pathogens, conducting field validation of predicted tick populations using active surveillance, and improving the representativeness of reporting populations across diverse and historically excluded communities.