Over the 290 selected cases, the cause of death was ruptured aneurysm in 160 cases and aortic dissection in 130 cases. In the rAA group, the average age was 70.26 years, while in the AAD group, it was 62.17 years. In the first group the male/female ratio was 1.96: 1.00 (106: 54), in the second it was 1.71: 1.00 (82: 48). In the rAA group the peak of occurrences appears to be in the winter, while in the AD group, it is in the spring. We did not report a statistically significant relationship neither in the seasonal variation nor in the monthly distribution of the events (Table 2.). Table 2(a) and 2(b) shows the seasonal and monthly variation of the events.
(Table 2(a) and 2(b)).
Results of our causal analysis are summarized in Table 3. It can be stated that, the transfer entropy between the timeseries of registered events and lagged climate data reaches its maximum when the lag value
is set to one, which can be interpreted that the weather on the preceding day has the greatest influence on the incidence of rAA and AAD. The only exception is rAA, where the air pressure changes between the 4th and 3rd day before the event counts.
(Table 3.)
Low p-values indicate a weak causal relationship. Between aortic dissection and mean daily temperature TE is not statistically significant, but values obtained in all other combinations are.
Figures 1(a) and 1(b) show the relative occurrence rate for rAA and AAD,calculated from the data. Figure 2 shows the daily mean temperature and air pressure change values recorded during the observation period.
(Figures 1(a) and 1(b), Figure 2)
In Figures 1(a) and 1(b), the rAA and AAD mortality r values are shown as a function of the mean temperature measured on the day before the event and the change in air pressure relative to the day before. The knowledge of the possible values of the weather parameters will help in the interpretation: Figure 2 shows the daily mean temperature and air pressure change values recorded during the observation period.
Figure 1(a) and 1(b) shows the expected percentage change in the daily occurrence of rAA and AAD as a function of the daily average temperature and atmospheric air pressure change in correlation to the previous day. For instance, we can observe that if the daily average temperature were less than 10°C, then 5 hPa increase in the daily air pressure change would result in an at least 10% growth of the incidence of rAA and AAD.
The r values of Figures 1 (a) and 1 (b) at a given point are more significant the denser the point cloud at point 2 is. That is, the more measurements we have around the given point. Based on this, the following conclusions can be drawn:
a) At rAA, in cold weather (mean temperature 0 to 10 ° C), the occurrence of the disease is increased by an increase of air pressure and reduced by a decrease. The figure shows that this could mean up to 20% increase or 10% decrease.
b) At most the incidence of AAD is only slightly dependent on temperature. Here too, cold weather and elevated air pressure are a risk factor, but the effect is weaker than at rAA.
c) In both figures, extreme values must be ignored because there is not enough measurement in these areas (see Figure 2).
Figure 3(a) and 3(b) depicts the real and model cumulative event numbers. The appropriateness of the model is shown by the fact that the values it predicts are well in line with reality, with a difference of less than 10 persons in 20 years.
(Figure 3(a) and 3(b))
Estimated intensities and cumulative trends are presented in Figure 3, where on the left axis, the solid line means the estimated daily intensity rAA and AAD modulated by weather conditions and the thick line stands for the directly unobservable hidden trend which can be traced back to relatively slow changes in the population such as migration, aging, changing habits, etc. The appropriateness of the model is shown by the fact that it can learn complex non-parametric hidden trends from data, moreover by splitting the trend that we can infer any complex nonlinear functional relationship between the response variable and the explicative variables. In Figure 3, we present also the cumulative number of events (right axis), where the solid line denotes the estimated number of events and the dotted line represents the actual number of rAA and AAD up to a given day.
Also shown in Figure 3 is the number of events in a day predicted by the model, indicated by a thin solid line. The global trend that is not directly observable is represented by a thick line. In both diseases, an abnormal peak in the trend is observed over the 2002-2012 period. It is important to note that the statistical model used allows us to decouple the influence of weather from the background trend.