Assessing the Effectiveness of Crowdsourced Data to Detect Established Tick Populations in Quebec, Canada:...

The range expansion of ticks transmitting zoonotic diseases in Canada poses challenges for public health surveillance. Active monitoring of where tick populations are emerging helps inform public health responses, but is resource intensive and logistically complex. Crowdsourced data may provide a cost-effective approach to augment surveillance. In this study, we assessed the value of crowdsourced data from eTick, a platform whereby the public submits tick photographs for expert identification.

We aimed to (1) characterize spatial patterns of eTick submissions and identify socioecological factors associated with their occurrence and frequency, and (2) assess the ability of indicators derived from eTick to predict census subdivisions (CSDs) with established tick populations.

We conducted a retrospective ecological assessment (2020-2023) for the effectiveness of crowdsourced data to detect established blacklegged tick (Ixodes scapularis) populations in Quebec, Canada. First, we applied a spatial scan approach to assess significantly high submission rates in human- and animal-origin data separately. Then, using zero-inflated regression models, we assessed socioecological factors associated with the occurrence and counts of tick submissions. Next, we used logistic regression models to assess the value of using eTick surveillance data to predict the presence of established tick populations, as documented by active surveillance. Finally, we calculated thresholds for eTick submissions to maximize sensitivity, specificity, or a balance of both metrics for detecting established tick populations, and developed scenarios to interpret linear predictors of submission counts across different socioecological contexts.

No clusters in submissions were detected for human and animal-origin data. However, high concentrations of ticks found on humans corresponded with areas reporting a high risk of Lyme disease. Tick submissions of animal origin extended beyond endemic areas, highlighting their complementary surveillance value. Zero-inflated models identified socioeconomic, climatic, and land cover factors significantly associated with submission rates. Logistic regression models, including significant socioecological factors, performed best to identify CSDs with established tick populations. Predictive performance was similar for ticks from humans (area under the curve [AUC] 0.86), animals (AUC 0.83), and all submissions combined (AUC 0.85). Using the best model adjusted for population size, median income, median age, and cumulative degree days above 0 °C, an optimal threshold of 3.32 crowdsourced tick submissions allowed the detection of CSDs with established tick populations while maximizing sensitivity and specificity (both 0.76).

This work introduces a framework for analyzing crowdsourced surveillance data by correcting for participation bias and benchmarking against active surveillance. Unlike approaches relying on traditional or unadjusted data, it incorporates socioecological factors associated with reporting. We show that crowdsourced data can yield reliable inferences about vector distributions and provide a methodological approach that could be applied across species and regions. Applied to eTick, it supports integration into surveillance systems to improve early detection, targeting interventions, and cost-effective monitoring of emerging risks by public health.