*Statistical Analysis*

Continuous variables are presented as the mean value (standard error). The univariate associations between NAOI categories and the daily number of EACs for EABP were evaluated by applying ANOVA and Kruskal-Wallis test.

As the numbers of EACs *Yt* are count variable, we suppose that *Yt* followed a Poisson distribution with mean *λt*, depending on predictor variables. We applied multivariate Poisson regression to evaluate the association between daily NAOI variables and daily EACs for EABP, which was specified as:

$$\ln(\lambda\_t) = \beta\_0 + \beta X\_t + \gamma Z\_t \tag{1}$$

where *Xt* is a NAOI variable, *β* is vector of coefficients for *Xt*, *Zt*—vectors of confounding factors—years, seasonality, week days, day length, weather, space weather [15] and air pollution variables, as these

may also affect the daily EACs rate; *γ*—is vector of coefficients for *Zi*. The daily incidence of EACs is defined as E(*Yt*) = *λt*. Poisson regression coefficients were interpreted as the difference between the log of expected counts.

Researchers in the topics of epidemiology and public health have been used the term "the relative risk." It is the ratio of two risks (or, informally, of rates or odds) comparing the risk of disease or death among the exposed to the risk among the unexposed. We could also interpret the Poisson regression coefficients as the log of the rate ratio. This explains the "rate" in incidence rate ratio. In our investigation, it is important for health professionals to evaluate how many times the high or low NAOI value increases the *λ<sup>i</sup>* (the daily mean value of EACs) as compared to reference category (−0.5 ≤ NAOI ≤ 0.5). Let *Xi* = 1, when NAOI with a lag of 2 days > 0.5, *Xi* = 2, when NAOI with a lag of 2 days < −0.5 and *Xi* = 0 otherwise. Then ln(*λi*) = *β*<sup>0</sup> + *β*1(*Xi* = 1) + *β*2(*Xi* = 2) + *γZi*, ln(*λi*|*Xi* = 1) = *β*<sup>0</sup> + *β*1(*Xi* = 1) + *γZi*, ln(*λi*|*Xi* = 0) = *β*<sup>0</sup> + *γZi* and the ratio of daily incidences when *Xi* = 1 and *Xi* = 0 (*λi*|*Xi* = 1)/(*λi*|*Xi* = 0) = exp(*β*1) is defined as Rate Ratio (RR). We presented adjusted rate ratios (RRs) in the multivariate Poisson regression model. The RRs are presented with 95% confidence interval (CI) and p-value. The analysis was performed separately for the number of calls during the whole day, in the morning until the early afternoon, in the afternoon until the evening and at night until the early morning during the colder (November–March) and the warmer (April–October) periods. For a sensitivity analysis, we evaluated the association between EACs for EABP and the NAOI separately for older (>65 years) and younger patients. Statistical analysis was performed using SPSS 19 software.
