The COVID-19 pandemic underscored the urgent need for forecasting models that integrate epidemiological surveillance with human mobility data. Although conceptual frameworks highlighting the role of mobility in epidemic spread exist, the operational use of such models in routine public health surveillance has been limited. Traditional approaches often rely solely on historical case trends, which, while informative, fail to account for inter-regional transmission dynamics driven by human movement. This gap has important implications for early-warning systems, as overlooking mobility data can result in delayed hotspot detection and suboptimal allocation of scarce resources. In this study, we developed and applied a Bayesian spatial–temporal Generalized Linear Mixed Model (GLMM) to COVID-19 case data obtained from the New York Times COVID-19 GitHub repository. The model integrated epidemiological and mobility information through a negative binomial likelihood with region-level random effects. Predictor variables included one-day and seven-day temporal lags, rolling averages, spatial spillovers, and normalized mobility inflows derived from gravity-model constructs. Weakly informative priors were assigned, and inference was performed using approximate Bayesian methods. The model’s performance was evaluated using posterior predictive checks and hotspot forecasting procedures. Results indicated that all key predictors were significantly associated with daily case counts. The one-day lag (β=0.35, 95% CrI: 0.25–0.45) and seven-day lag (β=0.28, 95% CrI: 0.14–0.41) captured epidemic inertia, while spatial lag effects (β=0.42, 95% CrI: 0.31–0.54) and mobility inflows (β=0.18, 95% CrI: 0.10–0.26) highlighted the importance of inter-regional connectivity. Posterior predictive checks showed strong calibration, with more than 90% of observed counts falling within the 95% credible intervals, and hotspot forecasts achieved precision above 80%, correctly identifying high-incidence regions. These findings demonstrate the value of Bayesian GLMMs in combining surveillance and mobility data to improve epidemic hotspot forecasting. By providing probabilistic, uncertainty-aware forecasts, this approach enhances the capacity of public health authorities to design timely, targeted, and cost-effective interventions, especially in resource-constrained settings.