Exploring linkages between weather factors and the risk of cerebral infarction through the application of Bayesian networks

Hiroshi Morimoto

A number of studies, mainly applying regression models, have investigated the correlation between meteorological factors and the incidence of cerebral infarction. One assumed subjectively the roles of the variables of the onset of cerebral infarction and the variables of the change in weather, and tried to find correlation. However, correlation can not induce causality. An exploratory method such as the Bayesian network approach, which is based on exhaustive searches, is thus required to determine causality. This approach considers all the possible causal associations and determines the most probable causality by estimating network scores. This study was based on the daily data on the number of patients transported by ambulance in the city of Nagoya in Japan. The patients were first transported to hospitals by ambulance prior to being diagnosed with cerebral infarction.  The use of heuristic search techniques enabled us to determine a causal relation between the weather and the incidence of cerebral infarction, including the influence of weather states such as high pressure or low pressure types considered as one of the nodes of the network. At the same time, we observed the influence of delayed effects of cold exposure on the onset of cerebral infarction. As an application of these findings, we showed the existence of the threshold for the mean temperature that divided risky temperature and safe temperature. These results provided clear evidence of the links between human physiological responses and changes in weather, and suggest the possibility of predicting cerebral infarctions.