Deep learning-based mortality surveillance: implications for healthcare policy and practice

Mortality modeling is critical for healthcare policy and resource allocation. Multilayer parameterization and static features are not needed for deep learning (DL) models. To enhance prediction accuracy, DL models like LSTM, Bi-LSTM, and GRU have shown promise. However, research on using DL models to mortality modeling remains restricted. Hence, this paper presents a unique technique that blends DL models with a Hotelling \(T^2\) $T^2$ control chart. The DL models anticipate the number of fatalities, while the Hotelling \(T^2\) $T^2$ control chart examines the abnormalities. Furthermore, MTY decomposition is employed for diagnostic analysis of age groups in the population. Using the Great Britain mortality dataset, a comparison study is developed between the Hotelling \(T^2\) $T^2$ control chart based on traditional mortality and the DL models. The data demonstrated that the suggested strategy outperformed the current methods. Moreover, this study illustrates the methodology’s potential for identifying mortality variations owing to emerging diseases.