Hefei is the capital of Anhui province and located in the middle of Anhui. It occupies a total area of 11,445 square kilometers, with a permanent population of 5,118,200 in the urban area ( http://www.hefei.gov.cn/mlhf/index.html ). In this study, we obtained all AAD admissions in the First Affiliated Hospital of University of Science and Technology of China (Anhui provincial hospital) from January 1, 2015 to December 31, 2020. The collected hospitalization information of AAD patients includes age, gender, home address, hospitalization date, and hypertension status. In addition, AAD is defined as dissection onset no more than 14 days (Yu et al., 2021) and the gold standard is CT angiography of the aorta. The exclusion criteria were as follows. (1) Iatrogenic aortic disease secondary to previous cardiac surgery or interventional repair. (2) Unidentified atherosclerosis without any evidence of multi-row computed tomography or echocardiography. (3) Traumatic AAD. (4) Exclude chronic dissection; more than 14 days is chronic aortic dissection. (5) Permanent residence outside Hefei. (6) Patients with congenital arterial malformations, Marfan syndrome, connective tissue disease, vasculitis and Loeys-Dietz syndrome.
Our meteorological data includes daily mean temperature, daily maximum temperature, daily minimum temperature, air pressure and relative humidity. Daily air quality data includes aerodynamic diameter < 10µm (PM10), sulfur dioxide (SO2), nitrogen dioxide (NO2). The above data comes from China Meteorological Data Network (http://data.cma.cn/) and China National Environmental Monitoring Center (https://www.aqistudy.cn/historydata/).
Time Series Analyses
Meteorological factors (daily mean temperature, daily maximum temperature, daily minimum temperature, air pressure, relative humidity) and air pollutants (PM10, SO2, NO2). Median, standard deviation, minimum, maximum, 25th percentile (P25) and 75th percentile (P75) parameters were used to describe data in descriptive analysis. Spearman correlation analysis was used to explore meteorological and air pollutants variables. If variables have a strong correlation, and we selectively eliminated during model fitting to avoid possible collinearity problems. As far as we know, the association between temperature and disease generally shows a non-linear trend (Chen et al., 2018). The DLNM statistical model was used to evaluate the two-dimensional relationship for the data with "cross-basis" function. DLNM not only can express the relationship of exposure-response between temperature and AAD, but also imply the non-linear hysteresis effect for both (Gasparrini et al., 2010). As in previous studies, we use a natural cubic spline to control the influence of air pollutants, including PM10, NO2, SO2, air pressure and relative humidity (Zhang et al., 2020). Furthermore, based on the conclusions of published articles that the cold effect can last for several weeks, we used a 28-day lag period for AAD (Li et al., 2016). According to the quasi-Poisson Akaike information criterion (Q-AIC) for temperature and lag with 3 degrees of freedom. We control long-term trends with 7 degrees of freedom per year (Pan et al., 2021). The basic model is shown as below:
Yt ~ Poisson(µt)
Log (µt) = α + β∗tempt, l +ns (PM10, 3) + ns (SO2, 3) + ns (NO2, 3) + Air pressure + ns (relative humidity, 3) + (time, 7/year) + DOWt
Where t represents the day of observation; Yt is the expected daily counts of AAD on day t (1,2,3,…2192); α is the model intercept and β is the coefficient in DLNM; tempt, l represents the "cross-basis" function and lag effect from lag 0 to lag l days, the maximum of l was set to 28 days (Guo et al., 2014); ns ( ) means a natural cubic spline, we control long-term trends with 7 degrees of freedom per year and 3 degrees of freedom for relative humidity, PM10, SO2 and NO2; a dichotomous variable representing the day of the week (DOW) in the model.
The DLNM model was used to fit the lag effect of ambient temperature (daily maximum temperature, daily mean temperature, daily minimum temperature) and AAD. Based on multiple analyses of the data, we choose the percentile of the daily average temperature to calculate the effect of AAD (cold effect: 10th ; hot effect: 90th ) and the median was used as the reference temperature (17.56°C) (Anderson and Bell, 2009).Through the comparison of the results of three temperature variables, the risk effect between daily mean temperature and AAD was significant. Thus, here we mainly report lag effect of daily average temperature on AAD (Yang et al., 2012).
In order to discover people who are susceptible to ambient temperature, we conducted a stratified analyses of gender, age, and presence or absence of hypertension. Among them, gender was divided into two groups: male and female; age was divided into group of < 55 years old and ≥ 55 years old due to the characteristics of AAD onset; existing evidence indicated that a positive correlation between hypertension and the onset of AAD, so, hypertension is a key risk factor that we should to be considered (Dong et al., 2019). Analyses were conducted by R software, packages "dlnm" and "splines".
Since conclusions usually vary with model specifications, sensitivity analysis needs to be performed to verify that risk estimates were robust. The degree of freedom was adjusted to 5 to control the confounding effect of the time trend, and the degree of freedom was adjusted to 4 and 5 to control the factors of weather and pollutants (Lin et al., 2013). The results have not changed substantially.