A Study On The Response Of Stroke And Coronary To Weather Change And The Prediction Model

In China, cardiovascular and cerebrovascular diseases had become the major killer of human health, which was the NO.1cause of human death. The main lethal diseases include stroke and coronary. Hypertension is the first risk factor of stroke and coronary. To explore which weather condition will cause stroke and coronary of people who have hypertension disease, the data of hypertension, cerebral ischemic stroke (CIS), cerebral hemorrhagic stroke (CHS), and ischemic heart disease (IHD) emergency visits of three hospitals in Beijing during2008to2012, along with the daily meteorological and pollution factors were collected. The study analyzed the relationship between diseases and meteorological factors using data of2008-2011by Generalized Additive Model (GAM) and set prediction models by different method of four seasons, respectively. The prediction methods included Stepwise Regression Model, BP Neural Network Model, Decision Tree Model and GAM. Then using the data of2012tested the goodness of fit and forecast accuracy of the prediction models. The results of this study are as follows:(1) People who have hypertension would be more sensitive to all the meteorological factors, the threshold value was lower and the increase range was much larger. The result showed that the influence of meteorological factors on diseases was significant and the relationship was non-linear, whose pictures always showed L or J-shaped, except wind speed. A threshold point could always be found and relative risk of diseases would increase linearly surpass the threshold when factors changed. When pressure was higher than1030.3hPa, a1hPa increase in it corresponded to an increase of6.36%in ER admissions for CIS in Beijing. When variable temperature was lower than-3.6°C, a1centigrade decrease in it corresponded to an increase of36.4%in ER admissions for CHS in Beijing. When temperature was lower than-5.9°C, a1centigrade decrease in it corresponded to an increase of10.3%in ER admissions for IHD in Beijing. Additionally, in the three diseases, IHD was the only one which was significant affected by wind speed. When wind speed increased by1m/s, the number of coronary emergency visits increased by5.74%.(2) The results of different seasons’study showed that we should focus on the meteorological factors which reveal the change of weather, such as24h variable temperature,24h variable pressure, diurnal temperature.24h variable temperature and24h variable pressure are the major meteorological risk factors of stroke and coronary. In spring and autumn, when24h variable temperature was lower than the threshold, a1centigrade decrease in it corresponded to an increase of18.1%-34.5%in ER admissions for the three diseases, whose influence is2-10times that of other meteorological factors.(3) The result of the comparison between whole year and different seasons showed that when24h variable temperature which was lower than-4.4°C decreases, the relative risk of CIS would increase by13.4%and21.3%(P<0.05), respectively in autumn and winter, which was6-10times of the whole year. The increase of relative risk of CHS at low temperature may due to low temperature of winter and the heating stop period of spring, when temperature of winter decreased, the relative risk of CIS would increase by3.9%(P<0.05). The high temperature of the whole year analysis was not related to IHD, but the high temperature of summer had significant relationship with IHD, when temperature which was higher than27.4increased by1centigrade, the number of IHD emergency visits would increase by7.74%(P<0.05).(4) The sequence of accuracy degree of different models were as follows:GAM>BP Neural Network Model>Stepwise Regression Model>Decision Tree Model. GAM is good at high level prediction, however, the result was higher than actual value. BP Neural Network Model prediction result’s range was closest to the range of actual value. Stepwise Regression Model had better performance at middle level prediction. Decision Tree Model had better performance in random sample prediction rather than time-series diseases prediction. GAM had good performance in high level of disease prediction, whose highest accuracy rate of CIS prediction was76%. GAM is an efficiency model to disease prediction, which basically met the demand for Medical meteorological forecast of CIS.