Climate signal or surveillance artefact? Disentangling environmental and observation processes in Lyme borreliosis trends

Climate change is increasingly invoked to explain long-term shifts in vector-borne disease, yet observed incidence reflects both ecological processes and the systems used to detect disease. Distinguishing environmental signals from changes in surveillance is therefore essential for attributing disease trends to global change. We used Lyme borreliosis in Poland as a national case study, analysing 259,003 notifications across 16 administrative regions from 2009 to 2024 (1,024 region-quarters). Multivariable negative binomial models estimated associations with lagged climatic conditions while accounting for temporal trend, a post-2016 structural step, COVID-19 healthcare disruption, seasonality and regional heterogeneity. Hospitalisation proportion provided an independent population-level indicator of changing observed case mix. Reported incidence increased 2.93-fold, whereas hospitalisation proportion declined from 26.4% to 3.5%. Temperature at a three-quarter lag (incidence rate ratio [IRR] 1.069 per 1'C, 95% CI 1.045 - 1.093) and cloud cover at a one-quarter lag (IRR 1.127 per okta, 1.063 - 1.195) were independently associated with reported incidence. However, changes in the observation process were of greater magnitude: the post-2016 structural step was associated with a 57% increase in notifications (IRR 1.571, 1.465 - 1.684), whereas the COVID-19 period was associated with a 48% reduction (IRR 0.524, 0.482 - 0.568). The divergence between increasing notifications and declining hospitalisation indicates substantial expansion in ascertainment of milder disease, such that notification trends cannot be interpreted as proportional changes in transmission. Our findings show that climatic signals can coexist with, and be obscured or amplified by, changes in disease observation. Robust attribution of vector-borne disease responses to global environmental change therefore requires explicit modelling of surveillance processes and integration of epidemiological, healthcare and ecological data.