The effects of “Fangcang, Huoshenshan, and Leishenshan” hospitals and environmental factors on the mortality of COVID-19

Data collection and mortality calculation

From January 21 to February 21, 2020, daily total number of confirmed cases by nucleic acid testing, daily total number of severe cases (i.e., confirmed cases who met one of the following conditions: 1. Respiratory rate ≥ 30 times per minute; 2. Resting state oxygen saturation ≤ 93%; 3. Partial arterial pressure of oxygen (PaO2)/concentration of oxygen (FiO2) ≤ 300 mmHg) (Xinyi & Yuanyuan, 2020), and daily total number of deaths in Wuhan city, Hubei Province and non-Hubei regions (as a contrast so as to reduce bias) were collected by two authors independently. All the above data were available on the official website of Chinese National Health Commission (http://www.nhc.gov.cn/). Growth rate of confirmed cases was calculated using the following formula: (1)${\displaystyle G{R}_n=\frac{N{C}_n}{T{C}_{\left(n-1\right)}}}$

where

GR_n = the growth rate on day n

NC_n = the new cases on day n

TC_(n−1) = the total cases on day (n − 1)

And the daily mortality rate was calculated using the following formulas: (2)${\displaystyle MC{C}_n=\frac{2\ast N{D}_n}{TC{C}_n+TC{C}_{\left(n-1\right)}}}$ (3)${\displaystyle MS{C}_n=\frac{2\ast N{D}_n}{TS{C}_n+TS{C}_{\left(n-1\right)}}}$

where

MCC_n = the mortality of confirmed cases on day n,

ND_n = the new deaths on day n,

TCC_n = the total confirmed cases on day n,

TCC_(n−1) = the total confirmed cases on day (n − 1),

MSC_n = the mortality of confirmed cases on day n,

TSC_n = the total severe cases on day n,

TSC_(n−1) = the total severe cases on day (n − 1).

The daily average data of three environmental factors, AT, RH, and AQI, were collected from Chinese Weather Net (http://www.weather.com.cn/), and the AT of Hubei Province was represented by the average AT of its seventeen cities (i.e., Wuhan, Huangshi, Shiyan, Yichang, Xiangyang, Ezhou, Jingmen, Xiaogan, Jingzhou, Huanggang, Xianning, Suizhou, Enshi, Xiantao, Qianjiang, Tianmen, and Shennongjia).

Statistical analysis

First, outliers of the datasets were detected and then deleted using SPSS software. Second, the data was transformed using z-score normalization, a method to standardize observations obtained at different times and from different cohorts, thus allowing comparisons between these observations (Guilloux et al., 2011). It was assumed that T was the original time series and Z was the Z-normalized time series: (4)${\displaystyle T=\left\{t_1,t_2,t_3,\dots \dots ,t_n\right\}}$ (5)${\displaystyle Z=\left\{z_1,z_2,z_3,\dots \dots ,z_n\right\}}$

Then (6)${\displaystyle z_i=\frac{{\mathrm{t}}_i-{\mathrm{\mu }}_T}{{\sigma }_T}}$

where  µ_T and  σ_T were the arithmetic mean value and standard variance of sequence T.

The data of each region was then divided into group A (from January 21 to February 5, before the use of MSHs) and group B (from February 6 to February 21, after the use of MSHs). Since the sample size was small (less than 50), the normality of the data was determined using Shapiro–Wilk test, and P value > 0.05 was considered as normally distributed (Mishra et al., 2019). If the data of the two groups were both normally distributed, Student’s t-test would be performed to compare their difference, and if the data of at least one group had a skewed distribution, Mann–Whitney U test would be performed instead (Parab & Bhalerao, 2010). We compared the data of four days, eight days, twelve days, and sixteen days after the use of MSHs, respectively, with the data of sixteen days before the use of MSHs. As for the correlation analysis, if the data of the environmental factors and the data of the growth rate/mortality were both normally distributed, Pearson correlation analysis would be performed to investigate the correlation between them, otherwise, Spearman’s correlation analysis would be performed instead (Schober, Boer & Schwarte, 2018). SPSS 26.0 statistical software (IBM, New York, USA) was used for statistical data processing, and GraphPad Prism 8.3 (GraphPad Software Inc., New York, USA) was used to plot graphs. All tests were two-sided, and P value < 0.05 was considered statistically significant.

Mortality difference before and after the use of MSHs

Daily number of confirmed cases, severe cases, new deaths, and daily AT, RH, and AQI in different regions were summarized in Table 1. The results of normality tests and the selection of statistical methods for comparative analyses are shown in Table 2. As shown in Fig. 1 and Table 3, no matter on day 4, day 8, day 12, or day 16 after the use of MSHs, the growth rates of confirmed cases were all significantly decreased both in Wuhan and Hubei; but in non-Hubei regions, changes were also significant.

As shown in Fig. 2 and Table 3, eight days after the use of MSHs, the mortality of confirmed cases was significantly decreased both in Wuhan (t = 4.545, P < 0.001) and Hubei (U = 0, P < 0.001), (t and U are the test statistic used to test the significance of the difference), while in non-Hubei regions, in contrast, the mortality of confirmed cases remained unchanged (U = 76, P = 0.106). While on day 12 and day 16 after the use of MSHs, the reduce in mortality was still significant both in Wuhan and Hubei; but in non-Hubei regions, the reduce also became significant this time (U = 123, P = 0.036; U = 171, P = 0.015, respectively).

As shown in Fig. 3 and Table 3, four days after the use of MSHs, the mortality of severe cases was significantly decreased in Hubei (U = 0, P = 0.002); and in non-Hubei regions, in contrast, changes were not significant (U = 48, P = 0.080). Similarly, on day 8, day 12, and day 16 after the use of MSHs, the reduce in mortality was still significant both in Wuhan and Hubei; but in non-Hubei regions, the reduce also became significant (U = 82, P = 0.039; U = 129, P = 0.015; and U = 177, P = 0.007, respectively).

In brief, the mortality of confirmed and severe cases was found to be significantly decreased after the use of MSHs both in Wuhan and Hubei; while in non-Hubei regions, the reduction in mortality was not significant on day 4/day 8, but became significant over time.

Correlation between environmental factors and outcomes

The results of normality tests and the selection of statistical methods for correlation analyses are shown in Table 4. As shown in Fig. 4. The negative correlation between the growth rate of confirmed cases and AT was not significant in Wuhan (P = 0.580), but significant in Hubei region (r =  − 0.644, P < 0.001). There was a significant negative correlation between AT and the mortality of confirmed cases both in Wuhan (r =  − 0.460, P = 0.014) and Hubei (r =  − 0.535, P = 0.004). And the mortality of severe patients was also found to be negatively correlated with AT in Hubei (r =  − 0.522, P = 0.005). This means that, if the AT rises 1 Celsius, the mortality of confirmed cases would drop by about 0.5% and the mortality of severe cases would drop by 0.522% on average.

As shown in Fig. 5, no significant correlation between the growth rate of confirmed cases and RH was found no matter in Wuhan (P = 0.946) or Hubei (P = 0.144). The correlation between the mortality of confirmed cases and RH was also insignificant both in Wuhan (P = 0.943) and Hubei (P = 0.107). As for the mortality of severe cases, its correlation with RH in Hubei was also found to be insignificant (P = 0.128).

As shown in Fig. 6, the growth rate of confirmed cases was found to be significantly correlated with AQI both in Wuhan (r = 0.373, P = 0.042) and Hubei (r = 0.426, P = 0.021). And the correlation between the mortality of confirmed cases and AQI was also significant in Wuhan (r = 0.620, P < 0.001) and Hubei (r = 0.634, P < 0.001). As for the mortality of severe cases, its correlation with AQI in Hubei was also significant (r = 0.622, P < 0.001). This means that, if the AQI drops 1 unit, the mortality of confirmed cases might drop by about 0.63% and the mortality of severe cases might drop by about 0.622%.

In brief, the mortality of confirmed/severe cases was negatively correlated with AT no matter in Wuhan or in Hubei, while the negative correlation between the growth rate of confirmed cases and AT was significant in Hubei, but not significant in Wuhan. In addition, both the growth rate and the mortality of COVID-19 cases were significantly correlated with AQI, but not with RH.