Risk-prediction model for acute myocardial infarction using atmospheric pressure data <Original Articles>

Prediction of high-risk days for onset of acute myocardial infarction (AMI) would aid prevention by taking specific actions on such days. To construct a predictive model, we investigated meteorological conditions related to high risk, with particular attention to atmospheric pressure. The data used were records of conveyance by ambulances in the city of Hiroshima from January 1993 to December 2002, and corresponding meteorological data in the area. We used a Poisson regression model and several variables representing different critical conditions of atmospheric pressure decline. Finally, we selected the best model according to Akaike's Information Criterion (AIC). A prediction model using a continuous variable of daily mean atmospheric temperature was established as the baseline model. Among models using different variables, one using weather pattern variables achieved the lowest AIC, showing it to be the best choice. In this model, strong winter patterns on the previous day were correlated with high risk of AMI. The following meteorological factors were particularly related to high AMI risk: 1) A weather chart showing a strong winter pattern on the previous day, and 2) a decline in atmospheric pressure ≥ 16 hPa in addition to low atmospheric temperature. Such a strong winter pattern is easy to use and will improve performance of the Hiroshima prefectural AMI alert system