The cut-off point estimation for piecewise function-based growth model

Growth models are commonly used to analyze the spread of diseases and make predictions during outbreaks. While single growth models can be effective in the early stages of an outbreak, they often fail to capture real-world fluctuations, leading to inaccurate predictions. To address this limitation, a novel piecewise model, the Modified Piecewise Growth Curve Model (MPGCM), was developed and applied to COVID-19 mortality data in Türkiye. The study analyzed 358 days of COVID-19 mortality data and found that a single growth model did not adequately align with the observed values. To improve accuracy, outputs from the generalized growth model were used to determine the cut-off points for the piecewise functions, identifying five change points. Within each segment, the best-fitting growth model was selected based on AIC and BIC values. Results indicated that the single logistic model consistently overestimated mortality, with percentage errors between +9.5% and +13.4%. In contrast, the MPGCM kept the error within ±0.5%, offering predictions much closer to actual values. These findings suggest that piecewise functions significantly enhance prediction accuracy for non-linear epidemiological trends. The proposed model offers a promising alternative for analyzing pandemic progression and could be adapted for future outbreaks.