Roles of Different Transport Modes in the Spatial Spread of the 2009 Influenza A(H1N1) Pandemic in Mainland China

Abstract

There is increasing concern about another influenza pandemic in China. However, the understanding of the roles of transport modes in the 2009 influenza A(H1N1) pandemic spread across mainland China is limited. Herein, we collected 127,797 laboratory-confirmed cases of influenza A(H1N1)pdm09 in mainland China from May 2009 to April 2010. Arrival days and peak days were calculated for all 340 prefectures to characterize the dissemination patterns of the pandemic. We first evaluated the effects of airports and railway stations on arrival days and peak days, and then we applied quantile regressions to quantify the relationships between arrival days and air, rail, and road travel. Our results showed that early arrival of the virus was not associated with an early incidence peak. Airports and railway stations in prefectures significantly advanced arrival days but had no significant impact on peak days. The pandemic spread across mainland China from the southeast to the northwest in two phases that were split at approximately 1 August 2009. Both air and road travel played a significant role in accelerating the spread during phases I and II, but rail travel was only significant during phase II. In conclusion, in addition to air and road travel, rail travel also played a significant role in accelerating influenza A(H1N1)pdm09 spread between prefectures. Establishing a multiscale mobility network that considers the competitive advantage of rail travel for mid to long distances is essential for understanding the influenza pandemic transmission in China.

Keywords: 2009 influenza A(H1N1) pandemic; China; quantile regression; rail travel; spatial spread; transport modes.

Figure 1

The spatial and temporal distribution of all 127,797 laboratory-confirmed influenza A(H1N1)pdm09 cases reported to the China Information System for Disease Control and Prevention in mainland China from 10 May 2009 to 30 April 2010: (a) spatial distribution of geocoded residential addresses; (b) the daily epidemic curve from 10 May 2009 to 30 April 2010; (c) the enlarged daily epidemic curve from 10 May 2009 to 31 August 2009.

Figure 2

Illustration of arrival days and peak days using the daily epidemic curve of confirmed cases in Beijing. The gray, green, and red vertical dashed lines represent, respectively, the date of the first case in mainland China (10 May 2009), the date of the first case in Beijing (16 May 2009), and the date with the highest incidence in Beijing (23 October 2009).

Figure 3

The presence/absence (1/0) of (a) airports and (b) railway stations in prefectures, and the distributions of (c) arrival days and (d) peak days for prefectures. The green line in (c) indicates the Heihe-Tengchong (Hu) line, and the one in (d) indicates the Hukun Railway that connects Shanghai and Kunming. The arrows illustrate that arrival days (or peak days) between prefectures with and without airports (or railway stations) are compared using the Mann-Whitney U test, and their spatial heterogeneity stratified by airports (or railway stations) are detected using q-statistic test.

Figure 4

Violin plots of arrival days and peak days between prefectures with (in red) and without (in cyan) airports (or railway stations). Box plots are embedded into violin plots to add summary statistics. The dark red points represent the mean arrival days (or peak days). * p < 0.05; *** p < 0.001; NS., not significant for comparing arrival days (or peak days) between prefectures with and without transport hubs using a Mann-Whitney U test.

Figure 5

Quantile process regression of arrival days in 115 prefectures. In (a) the density plot of arrival days, the red, green, and blue vertical lines indicate, respectively, the $\tau$ = 0.25, 0.50, and 0.75 quantiles of arrival days, whereas the dark green vertical line indicates the mean arrival days. In quantile process plots for log-transformed air (b), rail (c), and road (d) passenger volumes, the blue curves and shaded areas represent the quantile regression coefficients and 95% confidence intervals.