Statistical And Dynamic Model Analysis Of The Impact Of Air Pollutants On The Number Of Tuberculosis In Xinjiang

Objective:To constructed a prediction model in line with the epidemic characteristics of tuberculosis in Xinjiang,and to explore the impact of air pollutants on the number of tuberculosis by using multivariate models and dynamic models.so as to provide scientific basis for relevant departments to take preventive and control measures against tuberculosis in advance.Methods:（1）The SARIMA model,Holt-Winters additive model,Holt-Winters multiplicative model,GM（1,1）model and linear combination prediction model were created based on the number of pulmonary tuberculosis from January 2004 to June 2018.The predictive value of each model was evaluated using absolute percentage error（APE）,average absolute percentage error（MAPE）and root mean square error（RMSE）,and the best mode was selected based on minimum APE,MAPE and RMSE.（2）we separately constructed ARIMA,ARIMAX and RNN models to determine whether there exists an impact of the air pollutants on the number of pulmonary tuberculosis cases from January2014 to December 2018 and the air pollutant data of the same period.In addition,by using a new comprehensive evaluation index DISO to select the optimal model and predict the number of pulmonary tuberculosis cases in 2019.（3）Firstly,the correlation between air pollutants and tuberculosis cases was analyzed by Spearman rank correlation method,and the air pollutants unrelated to tuberculosis cases were excluded.Secondly,the air pollutant dynamics model was established,and the least square method and Bootstrap are used to fit the air pollutant concentration data.Thirdly,use the MCMC algorithm to fit the tuberculosis data,find out the optimal parameters of the tuberculosis dynamic model,and calculate the basic regeneration number of the tuberculosis model.Finally,sensitivity analysis of air pollutants and model parameters.Results:（1）SARIMA model is the optimal model（APE=10.94%,MAPE=7.96%,RMSE=419）.The linear combination prediction model is a suboptimal model,while the Holt-winters additive model,Holt-winters multiplication model and GM（1,1）model showed a low predictive value.（2）ARIMAX（1,1,2）×（0,1,1）₁₂+PM_2.5（lag=12）model was the optimal one,which was applied to predict the number of tuberculosis cases in 2019.The predicting results were in good agreement with the actual pulmonary tuberculosis cases and shown that tuberculosis cases obviously declined.（3）There is a negative correlation between nitrogen dioxide and the number of tuberculosis cases（r=-0.44）.The basic regeneration number（R₀）of the tuberculosis dynamics model is2.6975,95%CI:（2.1843,3.3952）.Increase the concentration of nitrogen dioxide and improve the recovery rate（r₁、r₂）of patients infected with sensitive and drug-resistant strains of tuberculosis can reduce the number of tuberculosis cases.Conclusion:（1）SARIMA model is better than linear combination prediction model,Holt-Winters addition model,Holt-Winters multiplication model and GM（1,1）model,and is more suitable for the prediction of the number of tuberculosis cases in Xinjiang.（2）Compared with the univariate model,the prediction accuracy of the multivariate model is improved.（3）The basic regeneration number R₀>1 of tuberculosis dynamic model,means that tuberculosis in Xinjiang will continue to spread among the population,and improving the recovery rate of individuals infected with tuberculosis is the key to reducing the number of tuberculosis cases.