Abstract: Objective To assess the predictive models commonly used in statistics, and determine the optimal model (s) for predicting the trend of tuberculosis. Methods Commonly used statistical predictive models were described concerning the principles, goodness of fit test and model selection, and applied to predicting the trend of tuberculosis in Xiamen city for choice of the best one(s) in following prevention and control of the epidemics. Results The fitting results into the model and data showed that 11 models had statistical significance (P<0. 05). In the 11 models, the largest R² model was Cubic, followed by the Quadratic and Logarithmic. Quadratic model was used, and revealed 191 (95% CI: 124-259), 192 (95% CI: 124-260), 193 (95% CI: 125-261), 194 (95% CI: 126-262) and 195 (95% CI: 127-263) and 196 cases of tuberculosis (95% CI: 128-264), respectively, reported in Xiamen area from July to December of 2019. Conclusion The commonly used statistical model can be used to predict the incidence trend of tuberculosis in Xiamen area, and the number of reported cases in Xiamen tends to slightly increase in the short term.

Key words: Mathematical model, Tuberculosis, Prediction