| Objective:To investigate seasonal and monthly variation of upper gastrointestinal bleeding (UGIB) and associations with meteorological factors in Liaocheng, China.Methods:All the records of patients with UGIB in Liaocheng city between1January2007and31December2012were reviewed. The meteorological data were obtained from Liaocheng meteorological bureau. Patient’s sex, age, diagnosis, endoscopic findings, and time of onset were recorded. The number of monthly, quarterly, and annual cases was counted. The seasonal and monthly variation of UGIB incidence was analyzed by sided x2test. The Spearman rank correlation coefficient was used to determine the correlations of seasonality, monthly and daily UGIB incidence and meteorological factors. The data were analyzed with SPSS17.0software. A value of p<0.05was considered significant.Results:A total of3,782cases were enrolled,2,247(59.41%) cases were male, and1,535(40.59%) cases were female. Mean age was (56.3±21.1) years old. There were two peaks of UGIB incidence, one was30-39years old, and the other was60-69years old. The etiologies of UGIB were as follows:peptic ulcer,1,558(41.20%); esophageal varices,1,062(28.08%); acute gastric mucosal lesions,252(6.66%); Mallory-Weiss syndrome,161(4.26%); esophageal or gastric cancer,213(5.63%); and the others,536(14.17%). Due to the patients of peptic ulcer and esophageal variceal bleeding were the majority,2,620(69.28%), our study also analyzed separately the circadian patterns of peptic ulcer and esophageal varices and their relationship with meteorological factors.The study showed clearly seasonal variation of the incidence of UGIB. With regard to3,782bleeding episodes, the highest percentage (1,034,27.34%) was seen during the winter, followed by the spring (972,25.70%), fall (900,23.80%), and summer (876,23.16%) in decreasing order (x2=19.272, p<0.001). Different etiologies of UGIB also existed seasonal variations, peptic ulcer had a spring peak and summer trough (26.70% vs.22.40%, x2=15.752,p<0.001); esophageal varices had a winter peak and summer trough (31.92%vs.20.62%, x2=18.312,p<0.001); acute gastric mucosal lesions had a summer peak and spring trough (30.16%vs.21.03%, x2=16.935, p<0.001); Mallory-Weiss syndrome had a summer peak and spring trough (26.71%vs.23.60%,x2=8.837, p=0.257); esophageal or gastric cancer increased in spring and decreased in fall (25.82%vs.23.94%, x2=5.852,p=0.124).Based on Spearman rank correlation analysis, there was a parallel relation between the seasonal number of UGIB and the mean atmospheric pressure, and inverse relations to the mean temperature, rainfall, and≥0℃accumulated temperature, there were no correlated with humidity, wind speed and hours of sunshine. There were parallel relations between the seasonal number of peptic ulcer and the mean atmospheric pressure and wind speed, and inverse relations to the mean temperature, rainfall, and≥0℃accumulated temperature, there were no correlated with humidity and hours of sunshine. There were parallel relations between the seasonal number of esophageal varices and the mean atmospheric pressure and wind speed, and inverse relations to the mean temperature, rainfall, hours of sunshine, and≥0℃accumulated temperature, there was no correlated with humidity.The number of UGIB incidence showed significant monthly fluctuations. For each month, the number was356in January,352in February,340in March,318in April,314in May,296in June,289in July,291in August,308in September,288in October,304in November, and326in December. It peaked in January (356,9.41%) and the trough was October (288,7.62%)(x2=39.964,p<0.001). Different etiologies of UGIB also had monthly fluctuations. Peptic ulcer was highest in March (151,9.69%) and fewest in August (111,7.12%)(x2=38.095,p<0.001);esophageal varices was highest in January (117,11.02%) and fewest in July (69,6.50%)(x2=51.601, p<0.001); acute gastric mucosal lesions was highest in August (27,10.71%) and fewest in May (15,5.95%)(x2=62.379, p<0.001); Mallory-Weiss syndrome was highest in July (17,10.56%) and fewest in May and June (11,6.83%)(x2=21.186,p=0.385); esophageal or gastric cancer was highest in May and November (21,9.86%) and fewest in September (13,6.10%)(x2=15.371, p=0.296).Based on Spearman rank correlation analysis, there were parallel relations between the monthly number of UGIB and the mean atmospheric pressure and wind speed, and inverse relations to the mean temperature, rainfall, and≥0℃accumulated temperature, there were no correlated with humidity and hours of sunshine. There was a parallel relation between the monthly number of peptic ulcer and the mean atmospheric pressure, and inverse relations to the mean temperature, rainfall, and≥0℃accumulated temperature, there were no correlated with humidity, wind speed, and hours of sunshine. There were parallel relations between the monthly number of esophageal varices and the mean atmospheric pressure and wind speed, and inverse relations to the mean temperature, rainfall, hours of sunshine, and>0℃accumulated temperature, there was no correlated with humidity.Based on Spearman rank correlation analysis, there was a parallel relation between the incidence of UGIB, peptic ulcer or esophageal varices and the daily atmospheric pressure (average, maximum, minimum), and an inverse relation to the daily temperature (average, maximum, temperature), there were no correlated with daily humidity, wind speed, rainfall, and hours of sunshine.Conclusions:There was an obvious seasonality variation and monthly fluctuations of UGIB. Meteorological factors may play an important role in the circadian patterns of UGIB. Objective:To predict the incidence of upper gastrointestinal bleeding (UGIB) by establishing back propagation (BP) neural network model when meteorological factors change.Methods:The hospitalization data of UGIB in Liaocheng city from2007to2012were collected. The meteorological data were obtained from Liaocheng meteorological bureau. MATLAB software was used to establish BP neural network model. The data were divided into two groups:training group and simulation group. The hospitalization data of UGIB and meteorological data from January2007to December2011was the training group, and the data from January2012to December2012was the simulation group. The meteorological data from January2007to December2011was as input layer, and the hospitalization data of UGIB from January2007to December2011was as the output layer. When the network structure was determined, the meteorological data from January2012to December2012was used to verify accuracy of the BP neural network model. The criteria of evaluate of the neural network were:(1) mean square error (MSE, the average squared difference between actual values and predictive values. Lower values are better), and Regression R Values (R, measure the correlation between actual values and predictive values. An R value of1means a close relationship,0a random relationship);(2) mean error rate between actual values and predictive values;(3) the paired-samples t test was used to determine the statistical significance of the difference between actual values and predictive values.Results:BP neural network model was established based on the hospitalization data of UGIB and meteorological data from January2007to December2011.The training algorithm was Levenberg-Marquardt, The best BP neural network model was constructed when the hidden neurons was11, the network structure was7-11-1.The training results showed MSE=4.928, R=0.948for training data, and MSE=6.793, R=0.875for all data. The model was evaluated by meteorological data from January 2012to December2012to predict the incidence of UGIB, and the results showed MSE=16.905,R=0.737, the average error rate was12.77%, there was no significant difference between actual values and predictive values (t=0.248,p=0.809).Conclusion:BP neural network model shows good predictive capability. It is valuable to predict UGIB when meteorological factors change. |
PDF Full Text Request