This study examines recent advancements in hybrid epidemiological modelling aimed at enhancing disease outbreak forecasting. Traditional compartmental models, such as the SEIR framework, have served as foundational tools in understanding infectious disease dynamics; however, their ability to capture the complexity of real-world outbreaks is limited. Recent studies have explored the integration of real-time environmental factors—such as temperature, rainfall, and humidity—and dynamic policy interventions, including lockdowns, social distancing, and vaccination campaigns, to address these limitations. Furthermore, the emergence of machine learning techniques, particularly transformer-based artificial neural networks, has opened new avenues for modelling complex, stochastic patterns inherent in disease transmission. This study highlights the benefits of ensemble approaches that combine mechanistic insights with adaptive learning capabilities. Emphasis is placed on the integration of environmental factors and policy interventions in the development of epidemiological models to improve prediction accuracy. Experimental design results, utilising 2020–2023 Covid-19 data for Zambia, reveal that ensemble-based models incorporating environmental and policy interventions data exhibit superior accuracy of RMSE = 361.1213, MAE = 229 when compared to models oblivious of this data that had an accuracy of RMSE = 843.8066, MAE = 785.2666 and R2 =  − 1.8930 at 200 epochs. Based on these results we recommend similar studies focusing on multi-disease models. The study also recommends the development of tools built on such models that can provide useful insights into the impact of policy interventions taking into consideration variable environmental factors to reduce the impact of disease epidemics.