Tropical cyclone and daily respiratory mortality across East Asia: a time series study

Extract

The threat of tropical cyclones in East Asia is a critical public health issue and a key reason for constructing adaptation strategies for the challenges of climate change. Some studies have indicated general evidence of the relationship between tropical cyclone exposure and the increased incidences of various kinds of diseases [1–3]. Specifically, there are several pathways for the impact of tropical cyclone exposure on respiratory diseases, such as increased dispersion of bacteria and allergens [4], and a sharp drop in both temperature and air pressure [5]. However, current evidence is limited to residents in the USA, and no epidemiological studies had explored the impact of tropical cyclones on respiratory mortality from an international perspective, particularly in East Asia, which is one of the regions with the most frequent tropical cyclone events in the world.

Tweetable abstract

This multi-country study offers crucial evidence on the associations between tropical cyclone exposure and heightened daily respiratory mortality in East Asia. Significant impacts were observed for pneumonia and COPD deaths, rather than asthma. https://bit.ly/42rCIqM

To the Editor:

The threat of tropical cyclones in East Asia is a critical public health issue and a key reason for constructing adaptation strategies for the challenges of climate change. Some studies have indicated general evidence of the relationship between tropical cyclone exposure and the increased incidences of various kinds of diseases [1–3]. Specifically, there are several pathways for the impact of tropical cyclone exposure on respiratory diseases, such as increased dispersion of bacteria and allergens [4], and a sharp drop in both temperature and air pressure [5]. However, current evidence is limited to residents in the USA, and no epidemiological studies had explored the impact of tropical cyclones on respiratory mortality from an international perspective, particularly in East Asia, which is one of the regions with the most frequent tropical cyclone events in the world.

To fill these gaps, we first matched daily mortality data with a tropical cyclone exposure database. We collected historical data on the spatial dynamics of tropical cyclones associated with different wind structures from previous studies [6, 7]. This database includes all attributes of tropical cyclones that had occurred across the Northwestern Pacific Ocean with a 6-h temporal resolution and has a good agreement on validation analysis against other reanalysis models [7]. By adopting a series of spatial identification tools, we identified a tropical cyclone exposure day to be when a city centre was within the affected buffer area during that day; we further categorised these days as either cyclone days (gale to violent storm winds, ≥34 knots) or hurricane days (winds ≥64 knots) [3]. Our mortality dataset comprises death records from 32 East Asian cities in mainland China (including Hong Kong), South Korea, Japan and Taiwan, spanning from 1972 to 2010. Detailed information on the mortality dataset was listed in our previous studies [8, 9]. According to the International Classification of Diseases (ICD-9 and ICD-10), we gathered mortality data for all natural causes (ICD-9: 001–799, ICD-10: A00–R99), cardiovascular diseases (ICD-9: 390–459; ICD-10: I00–I99), total respiratory diseases (ICD-9: 460–519; ICD-10: J00–J99), and three specific respiratory diseases: asthma (ICD-9: 493; ICD-10: J45–J46), pneumonia (ICD-9: 480–486; ICD-10: J12–J18) and COPD (ICD-9: 491–492, ICD-10: J41–J44). Although the dataset includes cases coded by both ICD systems, this does not impact our evaluation, as noted in our prior studies [9, 10].

The analysis period was limited to May–October, since 98.1% of tropical cyclone days were within these months in our dataset. Consistent with a previous study, a two-stage analytical model was used [11]. In the first stage, we calculated the city-level relative risk (RR) of respiratory mortality by comparing the tropical cyclone days to non-tropical cyclone days for each city. The RRs were estimated by a generalised additive model with a quasi-Poisson link function to account for the overdispersion in the distribution of daily death counts. The main model was adjusted by a natural cubic B-spline of time (four degrees of freedom per year) and the day of the week [12]. The cumulative effects of tropical cyclone exposure over 14 lag days were captured using a distributed lag model with a natural cubic B-spline (four degrees of freedom) and two internal knots placed at equally spaced values in the log scale [13]. The length of lag period was selected based on previous studies [1, 2]. We also controlled for the daily mean temperature using a cross-basis function with the same lag period as the tropical cyclone constructed by the distributed lag nonlinear model [2, 14]. In the second stage, univariate meta-regression models were used to pool the city-specific risk estimates [11]. We included location-specific average temperature, average humidity, population size and per capita gross domestic product as the predictors in meta-regression models. To ensure the robustness of our findings, we performed several sensitivity analyses by changing the degree of freedom for lag response and time variable, and also the maximum lag [15].

Of the 32 cities studied, 21 experienced at least one tropical cyclone during the study period, with six from Chinese mainland, three from Taiwan, six from South Korea and six from Japan (figure 1a and e). This study identified 1457 tropical cyclone days and 402 hurricane days. Results indicate a significant increase in respiratory mortality associated with tropical cyclone exposure, with the risk peaking on the concurrent day (RR 1.04, 95% CI 1.01–1.08) and decreasing in the following days (figure 1b). The RR for overall respiratory mortality was 1.20 (95% CI 1.06–1.37), and 1.25 (95% CI 1.03–1.47) for exposures to tropical cyclones (≥34 knots) and hurricanes (≥64 knots), respectively (figure 1c).

FIGURE 1.

The relationship between tropical cyclone exposure and daily mortality in 21 East Asian cities. a) The spatial locations of the 21 cities experiencing tropical cyclones during the study period in East Asia. b) The overall lag structure in the relative risk of daily respiratory mortality associated with tropical cyclone exposure in 21 East Asian cities. The solid line represents the mean relative risk estimates, and the shaded areas represent the 95% confidence intervals. c) Cumulative relative risk of daily respiratory mortality in 21 East Asian cities, classified by cyclone intensity, age, country/region and specific respiratory disease. The blue dots represent the relative risk estimates, and the solid lines represent the 95% confidence intervals. d) The cumulative relative risk (with 95% confidence intervals) of respiratory mortality associated with tropical cyclone exposure using different methods of adjustment for the sensitivity test. e) The number of the tropical cyclone and hurricane exposure days per year across 21 East Asian cities throughout the study years. Hong Kong was categorised as part of mainland China.

The cyclone-associated RRs for respiratory mortality in each city and country/region are presented in figure 1c. The RRs were 1.20 (95% CI 1.06–1.35) for the Chinese mainland, 1.19 (95% CI 1.02–1.37) for Taiwan, 1.23 (95% CI 1.06–1.40) for Japan, and 1.24 (95% CI 0.88–1.60) for South Korea. Significant associations were observed between tropical cyclone exposure and respiratory mortality for those aged 65 years and older, with RRs of 1.20 (95% CI 1.01–1.42) for residents aged between 65 and 75 years, and 1.24 (95% CI 1.04–1.51) for those older than 75 years. However, no significant association was found for populations under 65 years (RR 1.00, 95% CI 0.61–1.63). Pneumonia and COPD were found to be significantly impacted by tropical cyclone exposure, with RRs of 1.17 (95% CI 1.03–1.32) and 1.24 (95% CI 1.01–1.47), respectively. The point estimate for asthma was similar to those for pneumonia and COPD, but was not statistically significant (RR 1.21, 95% CI 0.78–1.84), which may be attributed to the fewer death cases. In terms of all natural causes of death and cardiovascular deaths, Using the same models with respiratory mortality, the results indicate a significant impact on all natural causes of death (RR 1.04, 95% CI 1.01–1.08), but an insignificant impact on cardiovascular deaths (RR 0.95, 95% CI 0.83–1.08). The results of the sensitivity analysis suggested that our findings were robust to using alternative model parameters (figure 1d).

To our knowledge, this is the first study evaluating the effects of tropical cyclones on daily mortality across multiple countries. This study provides first-hand epidemiological evidence in East Asia demonstrating significant associations between tropical cyclone exposure and increased mortality from respiratory diseases. This association was plausible due to a range of environmental changes induced by tropical cyclones, including high wind speeds, sudden drops in atmospheric pressure, the dissemination of bacteria and allergens, and infrastructure damage such as power outages. Taken together, these factors represent a potent trigger for respiratory disease exacerbations. Furthermore, our study demonstrates that elderly individuals are particularly vulnerable to tropical cyclones. The findings highlight the critical need for more effective policies for managing climate disasters and preparedness strategies that mitigate risks from tropical cyclones, particularly for elderly populations with pre-existing respiratory conditions. It should be noted that our results could be limited by the ecological study design and by the inevitable coding errors based on national death registries, but the direction of their impacts on our results remains unclear.

Shareable PDF

This one-page PDF can be shared freely online.

Shareable PDF ERJ-00546-2023.Shareable ^{(331.2KB, pdf)}