| objectives:(1)Analyze the temporal and spatial distribution characteristics of bacterial dysentery(BD)in Chongqing from 2009 to 2016,and grasp its prevalence.(2)Analyze the correlation between meteorological factors in Chongqing and the incidence of BD,explore the key meteorological factors affecting the incidence of BD,and find its vulnerable population.(3)Based on the potential impact of meteorological factors on the incidence of BD,establish a high-efficiency BD incidence prediction model,and explore the importance of meteorological factors to improve the accuracy of model prediction.Through the above three aspects of research to provide a scientific basis for the prevention and control of bacterial dysentery in Chongqing.Methods:1 In this study,data on bacterial dysentery cases and meteorological factors were collected from January 1,2009 to December 31,2016 in Chongqing.Epidemiological characteristics of BD analyzed by descriptive epidemiological techniques and Spatial-temporal scan statistics.2 The correlation between various meteorological factors and BD was analyzed using Distributed lag non-linear models(DLNM),so as to estimate the possible intervention effect of meteorological factors on BD in Chongqing.3 Boruta algorithm was combined with particle swarm optimization algorithm(PSO),support vector machine regression model(SVR)and LSTM neural network model to establish BD incidence prediction models.RMSE,MAPE and MAE were used to evaluate the prediction performance of models.Meanwhile,the importance of meteorological factors in improving the prediction performance of BD incidence prediction models were discussed.Results:1 Spatial and temporal distribution characteristics of BDA total of 68 855 cases of bacillary dysentery were included,and the incidence declined from 36.312/100 000 to 23.613/100 000,with an obvious seasonal peak from May to October.Males were more predisposed to the infection than females(the ratio was 1.118:1).Children < 5 years old had the highest incidence(295.892/100 000)among all age categories,and scattered children had the highest proportion(34 658 cases,50.335%)among all occupation categories.The Spatial-temporal scan statistics indicated that the most likely clusters for disease were found mainly in the core urban areas,and the main gathering time was from June to October.2 Effects of meteorological factors on BD2.1 The overall correlation between meteorological factors and BDThe daily average air temperature,daily average air pressure and daily average water vapor pressure in Chongqing have a non-linear relationship with the number of BD patients,and there is a strong immediate effect.When the daily average temperature is 36.5?(0 day lag),RR is the highest,which is 1.11(95%CI:1.07~1.14);when the daily average pressure is 961.5 hPa(0 day lag),RR is the highest,which is 1.07(95%CI:1.03~1.10);when the daily average water vapor pressure is 35 hPa(0 day lag),RR is the highest,which is 1.13(95%CI:1.08~1.17).In the 30 days lag time,when the daily average temperature reaches the highest value of 36.5?,the CRR value is the highest,which is 2.65(95%CI:2.19~3.21);when the daily average pressure is the lowest value of 961.5 hPa,the CRR value is the highest,which is 3.21(95%CI:2.50~4.11);when the daily average water vapor pressure is the highest value of 35 hPa,the CRR value is the highest,which is 1.98(95%CI:1.59~2.47).2.2 The extreme effect and cumulative extreme effect of meteorological factors on BDThe RR value of high temperature effect is 1.10(95%CI:1.07~1.13)when the extremely high daily mean temperature(32?)lags 0 day,and the lag effect lasts for about 10 days.The RR value of low-pressure effect is 1.05(95%CI:1.03~1.06)when the extremely low daily mean pressure(970 hPa)lags 0 day,and the lag effect lasts for about 30 days.When the extremely high daily average vapor pressure(28 hPa)lags 0 days,the RR value of the high vapor pressure effect is the largest,which is 1.09(95%CI:1.06~1.12),lasting for about 10 days.There was no statistical significance in the effects of daily mean air temperature extreme low(7?),daily mean air pressure extreme high(997 hPa)and daily mean water vapor pressure extreme low(8 hPa)on population BD.The CRR values of the cumulative extreme effects of daily mean air temperature extreme high,daily mean air pressure extreme low and daily mean water vapor pressure extreme high on BD were 1.79(95%CI:1.58~2.02),2.20(95%CI:1.91~2.53)and 1.54(95%CI:1.37~1.74)within the maximum lag days,respectively.The cumulative effects of daily mean air temperature extreme low,daily mean air pressure extreme low and daily average water vapor pressure extreme low were not statistically significance.2.3 Gender and age differences of daily average temperature on the incidence of BDThe high temperature effect on women is greater than that on men,and the high temperature effect on people aged 15-64 is greater than that on other age groups.The lag time of high temperature effect is also different among different genders and age groups.Men are about 13 days,women are about 10 days,people aged 0-14 are about 20 days,people aged 15-64 are about 10 days,and people aged 65 and over are about 3 days.There was no statistically significance in the effect of low temperature among different populations.The cumulative effect of temperature was the largest in the 0-14 age group(2.19(95%CI:1.83~2.60)),followed by the female population(1.98(95%CI:1.69~2.31)).From 2009 to 2015,the attribution score(AF)of daily mean temperature to the incidence of BD was 24.99%(95%CI: 22.12%~27.54%).Among them,the AF of high temperature(daily average temperature > 11?)was 25.63%(95% CI: 22.83~28.41),and that of low temperature(daily average temperature < 11?)was-0.65%(95%CI:-1.33% ~0.05%).The AF value of the influence of high temperature on BD of 0-14 years old people is the largest,which is 37.13%(95%CI:33.84%~39.92%).The AF value of low temperature has no statistical significance.3 Establishment and evaluation of four prediction models of BD incidence3.1 Modeling establishment and effect evaluation of BD monthly incidence predictionIn this study,RMSE,MAPE and MAE of PSO_SVR_MONTH model are 0.299,0.104 and 0.237,respectively,compared with SVR_MONTH model,RMSE,MAPE and MAE of PSO_SVR_MONTH model decreasing by 40%,44% and 43% respectively.Moreover,RMSE,MAPE and MAE of LSTM_MONTH model was 0.177,0.067 and 0.144,respectively,compared with LSTM_MONTH1 model,RMSE,MAPE and MAE of LSTM_MONTH model decreasing by 22%,14% and 22%,respectively.3.2 Modeling establishment and effect evaluation of BD weekly incidence predictionIn this study,the RMSE,MAPE and MAE of PSO_SVR_WEEK(time point =5,independent variables include meteorological factors)are 0.069,0.101 and 0.055,respectively.Compared with SVR_WEEK(time point =5,independent variables do not include meteorological factors),the RMSE,MAPE and MAE of PSO_SVR_WEEK decrease by 29%,30% and 29% respectively.The RMSE,MAPE and MAE of LSTM_WEEK(time point =5)constructed in this study are 0.072,0.108 and 0.057,respectively.Compared with LSTM_WEEK1(time point =5),the RMSE,MAPE and MAE of LSTM_WEEK(time point =5)decrease by 33%,41% and 37%,respectively.Conclusions:1 The situation of BD epidemic in Chongqing is still severe from 2009 to 2016.The relevant health departments should regard the main urban areas and northeastern areas in Chongqing as the key prevention and control areas of BD.Targeted responses to key populations such as children <5 years old,scattered children,farmers,and women should be carried out to maximize the control of BD transmission and prevalence according to both correlation between meteorological factors and the incidence of BD and its seasonal high incidence characteristics.2 Monthly and weekly incidence prediction models of BD established based on meteorological factors have showed satisfactory predictive performances.And the meteorological factors such as air temperature,air pressure and vapor pressure are important related feature sets of BD incidence prediction models,which can significantly improve the prediction accuracies of the models.The results of this study should serve as a model for the BD study in Chongqing.It provides a strong theoretical and technical support for the prevention and control of BD epidemic situation in Chongqing and the timeliness and effectiveness of rational allocation of health resources.At the same time,it can also provide a more complete research ideas and frameworks for other infectious diseases such as handfoot-and-mouth disease and influenza,etc. |
PDF Full Text Request