This paper is in the following e-collection/theme issue:

Environmental Health (81) Innovative Methods in Public Health and Surveillance (336)

Published on in Vol 9 (2023)

Preprints (earlier versions) of this paper are available at https://preprints.jmir.org/preprint/46792, first published .
The Combined Effects of Short-Term Exposure to Multiple Meteorological Factors on Unintentional Drowning Mortality: Large Case-Crossover Study

The Combined Effects of Short-Term Exposure to Multiple Meteorological Factors on Unintentional Drowning Mortality: Large Case-Crossover Study

The Combined Effects of Short-Term Exposure to Multiple Meteorological Factors on Unintentional Drowning Mortality: Large Case-Crossover Study

Authors of this article:

Yingyin Liu1 Author Orcid Image ;   Xiaomei Dong1 Author Orcid Image ;   Zhixing Li1, 2 Author Orcid Image ;   Sui Zhu1 Author Orcid Image ;   Ziqiang Lin1 Author Orcid Image ;   Guanhao He1 Author Orcid Image ;   Weiwei Gong3 Author Orcid Image ;   Jianxiong Hu4 Author Orcid Image ;   Zhulin Hou5 Author Orcid Image ;   Ruilin Meng6 Author Orcid Image ;   Chunliang Zhou7 Author Orcid Image ;   Min Yu3 Author Orcid Image ;   Biao Huang5 Author Orcid Image ;   Lifeng Lin6 Author Orcid Image ;   Jianpeng Xiao4 Author Orcid Image ;   Jieming Zhong3 Author Orcid Image ;   Donghui Jin7 Author Orcid Image ;   Yiqing Xu7 Author Orcid Image ;   Lingshuang Lv7 Author Orcid Image ;   Cunrui Huang8 Author Orcid Image ;   Tao Liu9 Author Orcid Image ;   Wenjun Ma1 Author Orcid Image
Article Authors Cited by (1) Tweetations (2) Metrics

    Original Paper

    1Department of Public Health and Preventive Medicine, School of Medicine, Jinan University, Guangzhou, China

    2Department of Nosocomial Infection Management, Affiliated Nanfang Hospital of Southern Medical University, Guangzhou, China

    3Zhejiang Provincial Center for Disease Control and Prevention, Hangzhou, China

    4Guangdong Provincial Institute of Public Health, Guangdong Provincial Center for Disease Control and Prevention, Guangzhou, China

    5Yunnan Provincial Center for Disease Control and Prevention, Kunming, China

    6Jilin Provincial Center for Disease Control and Prevention, Changchun, China

    7Guangdong Provincial Center for Disease Control and Prevention, Guangzhou, China

    8Hunan Provincial Center for Disease Control and Prevention, Changsha, China

    9Vanke School of Public Health, Tsinghua University, Beijing, China

    10Disease Control and Prevention Institute of Jinan University, School of Medicine, Jinan University, Guangzhou, China

    *these authors contributed equally

    Corresponding Author:

    Wenjun Ma, PhD

    Department of Public Health and Preventive Medicine

    School of Medicine

    Jinan University

    601 West Huangpu Avenue

    Tianhe District

    Guangzhou, 510632

    China

    Phone: 86 13622280436

    Email: mawj@gdiph.org.cn


    Background: Drowning is a serious public health problem worldwide. Previous epidemiological studies on the association between meteorological factors and drowning mainly focused on individual weather factors, and the combined effect of mixed exposure to multiple meteorological factors on drowning is unclear.

    Objective: We aimed to investigate the combined effects of multiple meteorological factors on unintentional drowning mortality in China and to identify the important meteorological factors contributing to drowning mortality.

    Methods: Unintentional drowning death data (based on International Classification of Diseases, 10th Edition, codes W65-74) from January 1, 2013, to December 31, 2018, were collected from the Disease Surveillance Points System for Guangdong, Hunan, Zhejiang, Yunnan, and Jilin Provinces, China. Daily meteorological data, including daily mean temperature, relative humidity, sunlight duration, and rainfall in the same period were obtained from the Chinese Academy of Meteorological Science Data Center. We constructed a time-stratified case-crossover design and applied a generalized additive model to examine the effect of individual weather factors on drowning mortality, and then used quantile g-computation to estimate the joint effect of the mixed exposure to meteorological factors.

    Results: A total of 46,179 drowning deaths were reported in the 5 provinces in China from 2013 to 2018. In an effect analysis of individual exposure, we observed a positive effect for sunlight duration, a negative effect for relative humidity, and U-shaped associations for temperature and rainfall with drowning mortality. In a joint effect analysis of the above 4 meteorological factors, a 2.99% (95% CI 0.26%-5.80%) increase in drowning mortality was observed per quartile rise in exposure mixture. For the total population, sunlight duration was the most important weather factor for drowning mortality, with a 93.1% positive contribution to the overall effects, while rainfall was mainly a negative factor for drowning deaths (90.5%) and temperature and relative humidity contributed 6.9% and –9.5% to the overall effects, respectively.

    Conclusions: This study found that mixed exposure to temperature, relative humidity, sunlight duration, and rainfall was positively associated with drowning mortality and that sunlight duration, rather than temperature, may be the most important meteorological factor for drowning mortality. These findings imply that it is necessary to incorporate sunshine hours and temperature into early warning systems for drowning prevention in the future.

    JMIR Public Health Surveill 2023;9:e46792

    doi:10.2196/46792

    Keywords

    drowning; exposure mixture; quantile g-computation; environmental epidemiology; meteorological factor 


    Drowning is the process of experiencing respiratory impairment from submersion or immersion in liquid [1]; it is a serious public health problem worldwide. The World Health Organization (WHO) reported that an estimated 236,000 people died from drowning in 2019. Globally, drowning is the third leading cause of death by unintentional injury, accounting for 7% of all injury-related deaths [2]. Despite the downward trend in drowning mortality rates [3], there are still considerable disparities across countries. Developed countries tend to have lower drowning mortality; for example, in the United States, the drowning mortality rate was 1.23 per 100,000 from 2010 to 2019 [4]. However, developing regions, like Southeast Asia, are the regions with the highest drowning mortality rates in the world. In the last decades, China has made great progress in drowning prevention and control, and the drowning mortality rate has decreased from 15.09 per 100,000 in 1990 to 3.97 per 100,000 in 2019 [5]; however, this is still a high drowning burden compared with most other countries.

    Prior studies on drowning mainly focused on epidemiological characteristics or risk factor analysis [6-8], clinical first aid [9,10], and intervention policy assessment [11,12]. Few studies have investigated the association of meteorological factors with drowning. Among these few studies, several reports indicate that there is a clear seasonal distribution of drowning deaths, with the highest drowning mortality in the summer [13]. The summer heat itself may not increase the risk of drowning, but it may lead to changes in behavior that increase susceptibility to drowning, such as increased water-related activities, increased alcohol consumption, or decreased use of personal flotation devices [14]. In terms of humidity, several studies report that high relative humidity (RH) at high temperatures significantly increases body temperature and increases the chance that people engage in water-related activities such as swimming [15], which in turn leads to an increased risk of drowning. Flooding due to extremely heavy rainfall is also an important cause of drowning mortality. For example, in July 2021, a flood occurred in Henan Province, China, due to persistent extremely heavy rainfall, and this incident resulted in hundreds of drowning deaths [16]. Even though there is no research on the association between duration of sunlight (DS) and drowning, there is research illustrating the importance of daylight hours for those who plan to enjoy swimming in a lake or outdoor pool, thereby potentially increasing the chance of drowning [17]. These studies focused on the effects of single weather factors on drowning. However, in the real world, people are exposed to multiple weather factors simultaneously, not only one at a time. Regrettably, no prior studies have examined the combined effects of exposure to multiple meteorological factors on drowning.

    Indeed, it is a challenge to quantify the combined effects of multiple meteorological exposure factors on drowning using traditional statistical models. In order to reduce collinearity in the model, previous multivariate regression models, such as linear regression, generalized linear regression, and generalized additive models (GAMs) have usually explored the association between a single or limited environmental factors and health outcomes. However, this approach could not accurately reveal the true impact of multiple environmental exposures [18]. In recent years, several advanced statistical methods, such as quantile g-computation, have been developed to examine the joint health effects of exposure mixture [19], which provides an opportunity for combined-effect analysis of exposure mixture in environmental epidemiology. Quantile g-computation not only allows for nonlinear correlation analysis, but also calculates the positive and negative weights contributed by each exposure.

    This study aimed to investigate the combined effects of multiple meteorological factors, including mean temperature (MT), RH, DS, and rainfall, on drowning mortality based on a data set from 5 provinces in China using a quantile g-computation model [19] and to further quantify the relative importance of each meteorological factor. Our findings may provide scientific information for developing an early warning system for drowning based on multiple meteorological factors to prevent drowning in the context of global warming.


    Data Collection

    In this study, the study sites in China were Guangdong Province in south China, Hunan Province in the middle part of the Yangtze River, Zhejiang Province on the southeast coast of China, Yunnan Province in southwest China, and Jilin Province in northeast China. Guangdong and Hunan are both regions with rich sunlight, heat, and water resources. Zhejiang Province is rich in marine resources. Both Jilin and Yunnan provinces have many rivers and lakes.

    Drowning mortality data from January 1, 2013, to December 31, 2018, in these 5 Chinese provinces were collected from the Provincial Disease Surveillance Points System. The information on each death included the date of death, cause of death, address, gender, age, education level, and cause of death code. In the International Classification of Diseases, 10th Edition (ICD-10), the codes W65 to 74 are unintentional drowning deaths, excluding drownings due to cataclysms or transportation accidents. The spatial distribution of drowning deaths in the 5 provinces is shown in Figure 1.

    ‎
    Figure 1. Spatial distribution of drowning deaths at the district and county level in 5 provinces. (A) Yunnan Province; (B) Jilin Province; (C) Hunan Province; (D) Zhejiang Province; (E) Guangdong Province.

    Daily meteorological data, including daily MT, RH, DS, and rainfall from 2013 to 2018 were collected from the China Meteorological Science Data Center, which collects data from 698 meteorological surveillance points across the country [20]. In order to obtain accurate daily MT and RH at a county level, we used Australian National University Splines (ANUSPLIN) thin plate smoothing [21] to interpolate station meteorological data to form raster data for daily MT and RH with a resolution of 0.01° × 0.01°. In the spatial interpolation model, we introduced latitude, longitude, and altitude as independent variables. The interpolation was validated using the 10-fold crossover method with good prediction accuracy, as previously described [22]. Daily mean DS and rainfall amount were calculated by averaging the values of all stations in each city because these 2 weather factors returned inaccurate data from interpolation in our preliminary analysis.

    Air pollution might be a confounding factor, so we also collected data on fine particulate matter (PM2.5) as a representative of air pollutants to control for possible confounding effects. Daily air pollution data (ie, PM2.5) were derived from the National Urban Air Quality Real-Time Release Platform [23]. Because the air pollution surveillance points did not completely cover our study provinces, the land use regression method, based on the random forest model, was used to predict ambient air pollutants; the model showed good prediction capability [24].

    Ethics Approval

    Ethical approval was obtained from the Medical Research Ethics Committee of Guangdong Provincial Center for Disease Control and Prevention (2019025), and all procedures followed the ethical standards specified by the institution.

    Statistical Analysis

    This study used a case-crossover design based on temporal stratification, which was further adapted from the case-crossover design of Levy et al [25]. The basic principle is to stratify time so that each case with an outcome is treated as a case group, while exposures in the same year, month, and day of the week are treated as controls, meaning that each case has 3 to 4 controls, thereby effectively controlling for confounding factors, including long-term trends, gender, and age, and effectively avoiding control selection bias in the original case-crossover design and enabling unbiased estimates to be obtained. This research design has been widely used to investigate the effects of short-term environmental exposure on population health, especially in the field of health-effect assessment of meteorological factors [26] and air pollution [27].

    Descriptive analyses were represented by numbers, means, maximum values, minimum values, and medians (quartiles), respectively. Spearman correlation was used to examine the correlations between meteorological factors and PM2.5.

    Few previous studies have reported the effects of wind speed and air pressure on drowning. Moreover, after preliminary analysis, we observed that wind speed and air pressure had little effect on drowning mortality (Multimedia Appendix 1). Therefore, these 2 weather factors were not included in the joint-effect analysis. In our study, we first conducted individual meteorological factor analysis using a GAM [28] after adjusting for confounding factors (ie, PM2.5) to assess the individual effect of each meteorological factor (MT, RH, DS, and rainfall) on drowning mortality. The GAM equation is as follows:

    Logit(case)=s(Variable,df)+s(PM2.5,df)+α,

    where Logit() is the connection function, case indicates the case type with 1 for cases and 0 for controls; Variable represents the meteorological factors to be analyzed, that is, MT, RH, DS and rainfall; PM2.5 is the concentration of PM2.5; s() is the nonparametric smoothing function, which is used to fit the nonlinear relationship between meteorological factors (ie, PM2.5 and drowning, respectively), with 3 degrees of freedom selected by generalized cross-validation [29]; and α denotes the intercept term.

    Second, in order to further estimate the joint effect of multiple meteorological factor exposure on drowning mortality, we used a quantile g-computation model [30]. Quantile g-computation is a new approach to estimate the joint effects of exposure mixture [19]. It has been used to evaluate the change in health outcome for each quantile of mixed exposure to multiple environmental factors [31-33], and it does not require the same effect directions between exposure variables and outcome [18]. In this study, the model was implemented through categorizing MT, RH, DS, and rainfall in quartiles, , coded as 0, 1, 2, and 3. The mixture is first fitted nonlinearly to drowning and then linearized according to the results. The quantile g-computation model equation is as follows (confounders Z could also be included):

    where (the change in drowning for a 1-unit change in all exposures) and each meteorological factor is given a negative or positive weight. The quantile g-computation estimator of the exposure response, ψ, is the sum of the regression coefficients across the included exposures. If all βj are in the same direction, then the weight for each component (indexed by k) is defined as , which is the proportion of the effect due to that component and sums to 1.0. If factors have different directions of effect, then the weights are interpreted as the proportion of the positive or negative partial effect, and the positive and negative weights together sum to 2.0 [32]. After obtaining regression coefficients for 71 cities, values from different cities were combined using meta-analysis to obtain the exposure-response relationship between exposure mixture and drowning risk. The results of the nonlinear fit were expressed as exposure response curves.

    The linearized results were expressed in terms of excess risk (ER) with the 95% CI. The formula for ER is as follows:

    ER = [exp(β) – 1] × 100%

    where β is the effect of exposure mixture in meta-analysis.

    Moreover, we used the quantile g-computation model to conduct subgroup analysis by sex (male, female), age group (<25 years, 25-64 years, ≥65 years) and season (the cold season, from October to March, and the warm season, from April to September). The following formula was used for the between-group variability test:

    where b1 and b2 are the effect estimates of the 2 groups, respectively, and SE1 and SE2 are the SEs of the estimated effects of the 2 groups, respectively [34].

    Sensitivity analysis was performed by changing the parameters of the model as well as by adjusting the confounding factors in order to verify the stability and reliability of the model. In the GAM, we conducted a sensitivity analysis by changing the degrees of freedom (df=2,4,5) for MT, RH, DS, and rainfall. In the quantile g-computation model, model 1 removed the confounding factor PM2.5 for sensitivity analysis. Model 2, model 3, and model 4 used 2-day, 3-day, and 4-day moving averages for MT, RH, and PM2.5 for sensitivity analysis.

    All tests were conducted 2-sided, and effects with P<.05 were considered statistically significant. All analyses were conducted with R (version 4.1.0; R Foundation for Statistical Computing) and the mgcv, qgcomp, mvmeta, and metafor R packages.


    Descriptive Analysis

    There were 46,179 drowning deaths in the 5 Chinese provinces from 2013 to 2018, with an average of 21 and a maximum of 68 drowning deaths per day. Each city had an average of 1 drowning death per day, with a maximum of 10 deaths (Table 1). A total of 17,408 people died from drowning during the summer seasons. The number of drowning deaths was the highest in Hunan Province and the lowest in Jilin Province. The number of drowning deaths was much higher for males (n=29,805, 64.5%) than females (n=16,374, 35.5%). The daily MT, RH, rainfall and PM2.5 during the study period were 20.4 °C, 76.7%, 4.5 mm, and 37.6 µg/m3, respectively (Table 2).

    Table 1. Descriptive statistics of meteorological factors and daily drowning deaths by season, province, sex, and age group from 2013 to 2018.
    CharacteristicsDrowning deaths overall (n=46,179), n (%)Drowning deaths per day, n


    		Mean (total=21.1)Minimum (total=3)P25a (total=14)P50b (total=19)P75c (total=26)Maximum (total=68)
    In each cityN/Ad1.3111210
    Season

    		Warm29,145 (63.1)26.5620263268

    		Cold17,034 (36.9)15.6312151940
    Sex

    		Male29,805 (64.5)13.618121849

    		Female16,374 (35.5)7.51571025
    Age (years)

    		≤2413,827 (30)6.4135926

    		25-6416,918 (36.6)7.81571027

    		≥6515,434 (33.4)7.1157923

    aP25: 25th percentile.

    bP50: 50th percentile.

    cP75: 75th percentile.

    dN/A: not applicable.

    Table 2. Descriptive statistics of meteorological factors from 2013 to 2018.
    Meteorological factorsMeanMinimumP25aP50bP75cMaximum
    Average temperature (°C)20.60.115.421.827.133.0
    Relative humidity (%)76.950.670.677.984.295.2
    Sunlight duration (hours)5.30.01.35.68.813.9
    Rainfall (mm)2.90.00.00.22.735.0
    PM2.5 (µg/m3)38.13.620.731.547.6302.9

    aP25: 25th percentile.

    bP50: 50th percentile.

    cP75: 75th percentile.

    Individual Effect of Meteorological Factors on Drowning Mortality

    Among the 4 meteorological factors, the strongest correlation was observed between RH and DS (r=–0.61; P<.05; Multimedia Appendix 2). As shown in Figure 2, in an individual-effect analysis, there were U-shaped relationships between MT, rainfall and drowning mortality, with the lowest risk at 15.4 °C for MT and 16.6 mm for rainfall, respectively. RH was negatively associated with drowning mortality, while DS was positively associated with drowning mortality.

    In the sensitivity analysis of the GAM, the relationship curves for MT, RH, DS, and rainfall with drowning deaths changed little after changing the degrees of freedom in the spline function. (Multimedia Appendix 3)

    ‎
    Figure 2. The associations of daily mean temperature, relative humidity, sunlight duration, and rainfall with drowning mortality in generalized additive models. Solid lines represent the exposure-response curve; dotted lines represent the 95% CI.

    Combined Effect of Multiple Meteorological Factors on Drowning Mortality

    In the exposure mixture analysis using quantile g-computation, mixed exposure to the 4 meteorological factors was approximately linearly associated with the risk of drowning. The risk tended to increase with each percentile (Figure 3). Meanwhile, we conducted an effect modification analysis of city-level characteristics and found that the risk of drowning caused by meteorological exposure mixture was higher in cities with higher population density, while urban gross domestic product, latitude, longitude, and altitude were found to have only weak modification effects (Multimedia Appendix 4).

    After linearization, we observed a 2.99% (95% CI 0.26%-5.80%) drowning mortality increase per quartile rise in the 4 meteorological factors (Figure 4). Subgroup analyses showed that the effect for males was higher than for females with a statistically significant difference (Z=2.21; P=.03). The risk of drowning was heterogeneous across age groups in combined effects. Although those aged 25 to 64 years had the highest risk (ER 4.56%, 95% CI 0.07%-9.25% per quartile rise in the 4 meteorological factors), the difference among all age groups was not statistically significant. The drowning risk attributed to mixed exposure to the 4 weather factors in the warm season seemed higher than in the cold season, but the difference was also not statistically significant (Z=1.61; P=.11).

    Figure 5 further shows the relative contribution of each meteorological factor to the combined effect. Sunlight duration had a positive weight with a much higher proportion among all subgroups: 93.1%, 94.1%, 100%, 88.9%, 83.1%, and 83.9% for the total population, males, females, and people aged 0 to 24 years, 25 to 64 years, and ≥65 years, respectively. RH negatively contributed to the combined effect in the total population (9.5%), females (26%), males (9.9%), and young people aged 0 to 24 years (43.2%) and positively contributed in people aged 25 to 64 years (12.4%) and ≥65 years (16.1%). Temperature contributed positively in the total population (6.9%), males (5.9%), the population aged 0 to 24 years (11.1%), and the population aged 25 to 64 years (4.5%), while it made a negative contribution in females (17.8%) and people aged ≥65 years (37.2%). Rainfall always contributed negatively to drowning mortality among all populations (90.5%, 90.2%, 56.2%, 56.8%, 100%, and 62.8% for the total population, males, females, people aged 0 to 24 years, 25 to 64 years, and ≥65 years, respectively; Figure 5).

    In a quantile g-computation model sensitivity analysis, the results were basically stable whether adjusting or not adjusting for PM2.5 in the model. The excess risk of drowning increased when moving averages of each weather factor were used in the model (Multimedia Appendix 5).

    ‎
    Figure 3. Exposure-response relationships between meteorological exposure mixture and drowning risk in 71 cities from a quantile g-computation analysis. The line is the exposure response curve, and the blue shaded area is the 95% CI at each percentile.
    ‎
    Figure 4. Excess risk (as a percentage) of drowning mortality per quartile increase in mixed daily mean temperature, relative humidity, sunlight duration, and rainfall in 71 cities.
    ‎
    Figure 5. Weights representing the proportion of the positive or negative partial effect of each meteorological factor in quantile g-computation model by sex and age group.

    Principal Findings

    In this study, we examined the combined effect of daily MT, RH, DS, and rainfall on drowning mortality in China. We found that for every quartile increase in meteorological factors, the risk of drowning death in the population increased by 2.99%. We also observed that the 4 meteorological factors had a positively combined effect on drowning mortality in all subgroups except those aged 0 to 24 years; adults aged 25 to 64 years may be more susceptible to mixed exposure to the 4 meteorological factors. Moreover, we showed that DS, rather than MT, was the most important meteorological factor contributing to drowning mortality. These findings have important public health implications for developing an early warning system to reduce drowning mortality in the context of global warming.

    Results Interpretation and Implications

    An increasing number of epidemiological studies have examined the relationship between exposure to single meteorological factors and the risk of drowning. Consistent with our findings, a study conducted by Chauvin et al [35] reported an increase in the risk of drowning death as ambient temperature rose. When the human body is exposed to high temperatures, the blood vessels in the skin dilate rapidly and sweating increases to accelerate the discharge of excess heat; when entering the water, heavy sweating could lead to electrolyte deficiencies, and swimming in water at this time may lead to dangerous reactions, such as convulsions [36]. Therefore, high temperatures increase the risk of drowning. Additionally, alcohol consumption at high temperatures is an important risk factor for drowning [37,38], as this could affect physiological reactions and behavioral patterns, thereby increasing the risk of drowning. A prior study reported that alcohol consumption was positively correlated with MT [39]. Individuals that consume alcohol have an increased risk of drowning due to greater skin blood flow and sweating. Also, alcohol can cross the human cerebrovascular barrier directly and affect mental and cognitive performance, and the heat sensation associated with alcohol consumption may increase exposure to water [40]. For instance, an Australian study reported that there was a higher tendency to see people consuming alcohol alongside rivers on hot days [41].

    To date, we have not found any study on the association of RH with drowning mortality. Among prior studies, RH is often treated as a confounding variable in the model fitting [42,43]. In our study, we observed that RH was negatively associated with the risk of drowning in an individual analysis of meteorological factors and negatively contributed to drowning mortality in most subgroups in a combined-effect analysis. The underlying mechanism of this finding is unclear and requires further in-depth studies in the future.

    In terms of DS, existing studies mainly focus on sunlight exposure and skin cancer or suicide [44-46]. The combined effect analysis in our study shows that DS was the most important weather factor in increasing drowning mortality in all subgroups. Possible explanations for this are that (1) longer daylight hours may be associated with higher temperatures, thereby increasing opportunities for water activities, and (2) light has the neurobiological effect of interacting with the brain serotonin system and may influence serotonin-related behavior [46]. Finally, when daylight hours are short, people are less likely to be active outside and are likely to reduce their contact with bodies of water, which in turn may lead to a decreased incidence of drowning.

    We further observed that rainfall was a negative factor for unintentional drowning risk in the combined-effect analysis. Previous studies have shown a negative correlation between rainfall and physical activity and a positive association with sedentary time during weekdays [47]. Therefore, rainfall might make it more difficult for people to work outdoors, which in turn may reduce the chance of contact with bodies of water. In contrast, Murray and Carter [48] showed that high rainfall (180.55 mm to 202.25 mm) was associated with increased drowning deaths in Fiji. The rainfall amount in this study was less than 35 mm, which is more likely to reduce people’s willingness to go outside; thus, we observed a reduced risk of drowning. In addition, the deaths in this study did not include flood-related drownings or disaster drownings.

    In the stratified analysis, we further observed that males exhibited higher risk than females when exposed to meteorological mixed factors, which is consistent with previous studies. For example, the risk of drowning death is 1.67 times higher in men than women on hot days [14]. Compared to females, males are more likely to work in hot outdoor environments, resulting in potentially greater access to bodies of water and higher risk of drowning. We also observed that for people aged 0 to 24 years, there was no statistically significant association between mixed exposure to the 4 meteorological factors and drowning. This is contrary to previous studies. For instance, a study in Shanghai reported that drowning deaths among children aged 0 to 14 years mainly occurred in the hot season, while other age groups showed less seasonality in drowning [49]. Possible reasons for this are that the young population is more likely to be influenced by individual personality traits (eg, an active and strong spirit for exploring outside) and behavioral patterns and less influenced by external weather factors and the environment; on the other hand, children also end up in bodies of water due to unintentional falls and fights. Meteorological exposure mixture had the greatest effect on people aged 25 to 64 years; one possible explanation for this finding was that adults were more likely to consume alcohol in hot weather, and drinking alcohol increases exposure to water [50]. Moreover, this age group is the main workforce, with a high chance of occupational exposure to bodies of water, which may increase drowning deaths [51].

    Several previous studies have estimated the risk of drowning associated with exposure to a single meteorological factor. However, people are exposed to a variety of meteorological factors simultaneously, so it is important to use an exposure mixture approach to measure the combined effects of weather factors on human health. In this study, we estimated the combined effects of 4 weather factors using a quantile g-computation model. This model helps design potential public health interventions for specific sources of exposure [19].

    Multiple sectors in the government, such as education departments, health departments, and meteorological departments, should collaborate to develop an early warning system taking into account DS as a potentially critical weather factor to reduce the risk of drowning in the high-risk season. In addition, in areas with a high risk of drowning, governments should provide adequate and safe swimming places to reduce the possibility of people going into dangerous waters for recreation.

    Limitations

    This study has several limitations that must be noted. First, due to the lack of individual exposure data, we used meteorological data obtained by spatial interpolation as a proxy in each city, and there may be some exposure misclassification bias. Second, although this study is based on 5 provinces, which are to some extent nationally representative, extrapolation of the results needs to be done with caution. Third, drowning is influenced by family factors and socioeconomic factors, which were not included in this study. Finally, we only included unintentional drowning (ICD-10 codes W65-74) in this study, which excludes drowning due to cataclysm (ICD-10 codes X34-X39) and transportation accidents (ICD-10 codes V01-V99). This may not adequately reflect the true number of drowning deaths [52].

    Conclusion

    This study shows that mixed exposure to the 4 weather factors was positively associated with drowning mortality, which was primarily driven by DS and MT. This finding implies that it is necessary to incorporate sunshine hours and temperature into early warning systems for drowning in the future.

    Acknowledgments

    This work was supported by the National Key Research and Development Program of China (2018YFA0606200), the National Natural Science Foundation of China (42275187), and the Foreign Expert Program of the Ministry of Science and Technology (G2022199006L).

    Data Availability

    The data sets generated and/or analyzed during this study are available from the corresponding author on reasonable request.

    Authors' Contributions

    YL and XD are joint first authors. WM had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. WM and TL conceived the original idea and supervised the study. YL and Z Li conducted the data analysis and drafted the manuscript. YL, XD, and Z Li participated in manuscript writing. WG, ZH, RM, CZ, MY, BH, L Lin, JX, JZ, DJ, YX, and L Lv conducted the investigation and contributed to resource organization and data curation. CH was responsible for project administration, resource organization, and data curation. JH, SZ, and Z Lin participated in results review. WM, TL, and GH participated in revision. All authors reviewed the final manuscript. WM and TL are joint corresponding authors and have contributed equally to this paper.

    Conflicts of Interest

    None declared.

    Multimedia Appendix 1

    Preliminary analysis of air pressure and wind speed in a generalized additive model.

    PNG File , 252 KB
    Multimedia Appendix 2

    Spearman correlations between meteorological factors and PM2.5.

    PNG File , 101 KB
    Multimedia Appendix 3

    Sensitivity analysis, changing the degrees of freedom in a spline function with mean temperature, relative humidity, sunlight duration, and rainfall in a generalized additive model.

    PNG File , 461 KB
    Multimedia Appendix 4

    The modification effects of city-level characteristics.

    PNG File , 346 KB
    Multimedia Appendix 5

    Sensitivity analysis with a quantile g-computation model.

    DOCX File , 20 KB
    1. van Beeck EF, Branche C, Szpilman D, Modell J, Bierens JJLM. A new definition of drowning: towards documentation and prevention of a global public health problem. Bull World Health Organ. Nov 2005;83(11):853-856. [FREE Full text] [Medline]
    2. Drowning. World Health Organization. URL: https://www.who.int/news-room/fact-sheets/detail/drowning [accessed 2022-09-15]
    3. Franklin RC, Peden AE, Hamilton EB, Bisignano C, Castle CD, Dingels ZV, et al. The burden of unintentional drowning: global, regional and national estimates of mortality from the Global Burden of Disease 2017 Study. Inj Prev. Oct 28, 2020;26(Supp 1):i83-i95. [FREE Full text] [CrossRef] [Medline]
    4. Drowning facts. US Centers for Disease Control and Prevention. URL: https://www.cdc.gov/drowning/facts/index.html [accessed 2022-09-15]
    5. Global Burden of Disease Study 2019 (GBD 2019) data resources. Institute for Health Metrics and Evaluation. URL: https://ghdx.healthdata.org/gbd-2019 [accessed 2022-09-15]
    6. Willcox-Pidgeon S, Franklin RC, Leggat PA, Devine S. Epidemiology of unintentional fatal drowning among migrants in Australia. Aust N Z J Public Health. Jun 2021;45(3):255-262. [FREE Full text] [CrossRef] [Medline]
    7. Yang L, Nong Q, Li C, Feng Q, Lo SK. Risk factors for childhood drowning in rural regions of a developing country: a case-control study. Inj Prev. Jun 2007;13(3):178-182. [FREE Full text] [CrossRef] [Medline]
    8. Zhu Y, Jiang X, Li H, Li F, Chen J. Mortality among drowning rescuers in China, 2013: a review of 225 rescue incidents from the press. BMC Public Health. Jul 10, 2015;15(1):631. [FREE Full text] [CrossRef] [Medline]
    9. Culp W, Brewster B, Wernicki P. Drowning: lifeguard rescue and resuscitation. Anesthesiology. Jan 2010;112(1):245; author reply 246. [FREE Full text] [CrossRef] [Medline]
    10. Szpilman D, de Barros Oliveira R, Mocellin O, Webber J. Is drowning a mere matter of resuscitation? Resuscitation. Aug 2018;129:103-106. [CrossRef] [Medline]
    11. Brenner R. Prevention of drowning in infants, children, and adolescents. Pediatrics. Aug 2003;112(2):440-445. [CrossRef] [Medline]
    12. Denny S, Quan L, Gilchrist J, McCallin T, Shenoi R, Yusuf S, et al. Council on Injury‚ Violence‚ and Poison Prevention. Prevention of drowning. Pediatrics. Aug 2021;148(2):e20190850. [CrossRef] [Medline]
    13. Peden AE, Franklin RC, Willcox-Pidgeon S. Media reporting of summer drowning: A partial picture, useful for advocacy. Health Promot J Austr. Sep 28, 2020;31(3):491-496. [CrossRef] [Medline]
    14. Fralick M, Denny CJ, Redelmeier DA. Drowning and the influence of hot weather. PLoS One. 2013;8(8):e71689. [FREE Full text] [CrossRef] [Medline]
    15. Mora C, Dousset B, Caldwell IR, Powell FE, Geronimo RC, Bielecki C, et al. Global risk of deadly heat. Nature Clim Change. Jun 19, 2017;7(7):501-506. [CrossRef]
    16. Lancet. Flooding, drowning, and health equity: The Lancet Regional Health - Western Pacific August editorial. Lancet Reg Health West Pac. Aug 2021;13:100264. [FREE Full text] [CrossRef] [Medline]
    17. Falk M. Impact of weather conditions on tourism demand in the peak summer season over the last 50years. Tour Manag Perspect. Jan 2014;9:24-35. [CrossRef]
    18. Rosato I, Zare Jeddi M, Ledda C, Gallo E, Fletcher T, Pitter G, et al. How to investigate human health effects related to exposure to mixtures of per- and polyfluoroalkyl substances: A systematic review of statistical methods. Environ Res. Apr 01, 2022;205:112565. [CrossRef] [Medline]
    19. Keil AP, Buckley JP, O'Brien KM, Ferguson KK, Zhao S, White AJ. A quantile-based g-computation approach to addressing the effects of exposure mixtures. Environ Health Perspect. Apr 2020;128(4):47004. [FREE Full text] [CrossRef] [Medline]
    20. China Meteorological Science Data Center. URL: http://data.cma.cn/ [accessed 2022-09-15]
    21. ANUSPLIN version 4.4 user guide. Australian National University. URL: https://fennerschool.anu.edu.au/research/products/anusplin [accessed 2023-06-30]
    22. Liu T, Zhou C, Zhang H, Huang B, Xu Y, Lin L, et al. Ambient temperature and years of life lost: a national study in China. Innovation (Camb). Feb 28, 2021;2(1):100072. [FREE Full text] [CrossRef] [Medline]
    23. National urban air quality real-time release platform. URL: https://air.cnemc.cn:18007/ [accessed 2022-09-15]
    24. Liu T, Meng H, Yu M, Xiao Y, Huang B, Lin L, et al. Urban-rural disparity of the short-term association of PM with mortality and its attributable burden. Innovation (Camb). Nov 28, 2021;2(4):100171. [FREE Full text] [CrossRef] [Medline]
    25. Levy D, Lumley T, Sheppard L, Kaufman J, Checkoway H. Referent selection in case-crossover analyses of acute health effects of air pollution. Epidemiology. Mar 2001;12(2):186-192. [CrossRef] [Medline]
    26. Lee H, Myung W, Kim H, Lee E, Kim H. Association between ambient temperature and injury by intentions and mechanisms: A case-crossover design with a distributed lag nonlinear model. Sci Total Environ. Dec 01, 2020;746:141261. [CrossRef] [Medline]
    27. Kim Y, Ng CFS, Chung Y, Kim H, Honda Y, Guo YL, et al. Air pollution and suicide in 10 cities in northeast Asia: A time-stratified case-crossover analysis. Environ Health Perspect. Mar 06, 2018;126(3):037002. [FREE Full text] [CrossRef] [Medline]
    28. Ravindra K, Rattan P, Mor S, Aggarwal AN. Generalized additive models: Building evidence of air pollution, climate change and human health. Environ Int. Nov 2019;132:104987. [FREE Full text] [CrossRef] [Medline]
    29. Wood S. Generalized Additive Models: An Introduction with R. 2nd ed. Boca Raton, FL. Chapman and Hall/CRC; 2017.
    30. Keil A. qgcomp: Quantile G-Computation. Cran R Project. URL: https://CRAN.R-project.org/package=qgcomp [accessed 2022-09-15]
    31. Li G, Liu J, Lu H, Hu W, Hu M, He J, et al. Multiple environmental exposures and obesity in eastern China: An individual exposure evaluation model. Chemosphere. Jul 2022;298:134316. [CrossRef] [Medline]
    32. Niehoff NM, Keil AP, O'Brien KM, Jackson BP, Karagas MR, Weinberg CR, et al. Metals and trace elements in relation to body mass index in a prospective study of US women. Environ Res. May 2020;184:109396. [FREE Full text] [CrossRef] [Medline]
    33. Sun Y, Li X, Benmarhnia T, Chen J, Avila C, Sacks DA, et al. Exposure to air pollutant mixture and gestational diabetes mellitus in Southern California: Results from electronic health record data of a large pregnancy cohort. Environ Int. Jan 2022;158:106888. [FREE Full text] [CrossRef] [Medline]
    34. Schenker N, Gentleman JF. On judging the significance of differences by examining the overlap between confidence intervals. Am Stat. Aug 2001;55(3):182-186. [CrossRef]
    35. Chauvin M, Kosatsky T, Bilodeau-Bertrand M, Gamache P, Smargiassi A, Auger N. Hot weather and risk of drowning in children: Opportunity for prevention. Prev Med. Jan 2020;130:105885. [CrossRef] [Medline]
    36. Maughan RJ, Shirreffs SM. Muscle cramping during exercise: Causes, solutions, and questions remaining. Sports Med. Dec 2019;49(Suppl 2):115-124. [FREE Full text] [CrossRef] [Medline]
    37. Driscoll TR, Harrison JA, Steenkamp M. Review of the role of alcohol in drowning associated with recreational aquatic activity. Inj Prev. Apr 2004;10(2):107-113. [FREE Full text] [CrossRef] [Medline]
    38. Pajunen T, Vuori E, Vincenzi FF, Lillsunde P, Smith G, Lunetta P. Unintentional drowning: Role of medicinal drugs and alcohol. BMC Public Health. May 19, 2017;17(1):388. [FREE Full text] [CrossRef] [Medline]
    39. Hagström H, Widman L, von Seth E. Association between temperature, sunlight hours and alcohol consumption. PLoS One. 2019;14(9):e0223312. [FREE Full text] [CrossRef] [Medline]
    40. Wasielewski JA, Holloway FA. Alcohol's interactions with circadian rhythms. A focus on body temperature. Alcohol Res Health. 2001;25(2):94-100. [FREE Full text] [Medline]
    41. Peden AE, Franklin RC, Leggat PA. Breathalysing and surveying river users in Australia to understand alcohol consumption and attitudes toward drowning risk. BMC Public Health. Dec 19, 2018;18(1):1393. [FREE Full text] [CrossRef] [Medline]
    42. Barnett A, Tong S, Clements A. What measure of temperature is the best predictor of mortality? Environ Res. Aug 2010;110(6):604-611. [FREE Full text] [CrossRef] [Medline]
    43. Zhang K, Li Y, Schwartz JD, O'Neill MS. What weather variables are important in predicting heat-related mortality? A new application of statistical learning methods. Environ Res. Jul 2014;132:350-359. [FREE Full text] [CrossRef] [Medline]
    44. Lindqvist PG. The winding path towards an inverse relationship between sun exposure and all-cause mortality. Anticancer Res. Feb 2018;38(2):1173-1178. [CrossRef] [Medline]
    45. Lindqvist P. On the relationship between sun exposure and all-cause mortality. Adv Exp Med Biol. 2020;1268:115-122. [CrossRef] [Medline]
    46. Vyssoki B, Kapusta ND, Praschak-Rieder N, Dorffner G, Willeit M. Direct effect of sunshine on suicide. JAMA Psychiatry. Nov 2014;71(11):1231-1237. [CrossRef] [Medline]
    47. Harrison F, Jones AP, Bentham G, van Sluijs EM, Cassidy A, Griffin SJ. The impact of rainfall and school break time policies on physical activity in 9-10 year old British children: a repeated measures study. Int J Behav Nutr Phys Act. May 24, 2011;8(1):47. [FREE Full text] [CrossRef] [Medline]
    48. Murray K, Carter P. Fatal drownings in Fiji. Asia Pac J Public Health. Jan 2017;29(1):28-34. [CrossRef] [Medline]
    49. Yi-chen C, Xiao-pan L, Chen Y, Li-peng HA. Analysis on the epidemiological characteristics and trend of traffic fatality data among residents in Pudong New Area of Shanghai, 2002-2015. Chin J Dis Control Prev. 2017;21(6):602-606. [CrossRef]
    50. Clemens T, Tamim H, Rotondi M, Macpherson AK. A population based study of drowning in Canada. BMC Public Health. Jul 13, 2016;16:559. [FREE Full text] [CrossRef] [Medline]
    51. Yang L, Lin Q, Wang Y, Tong Z. Mortality and potential life lost of drowning in Zhoushan residents. J Prev Med. Published online. 2019:1116-1119. [CrossRef]
    52. Peden AE, Franklin RC, Mahony AJ, Scarr J, Barnsley PD. Using a retrospective cross-sectional study to analyse unintentional fatal drowning in Australia: ICD-10 coding-based methodologies verses actual deaths. BMJ Open. Dec 21, 2017;7(12):e019407. [FREE Full text] [CrossRef] [Medline]

    ‎
    ANUSPLIN: Australian National University Splines
    DS: duration of sunlight
    ER: excess risk
    GAM: generalized additive model
    ICD-10: International Classification of Diseases, 10th Edition
    MT: mean temperature
    OR: odds ratio
    PM: particulate matter
    RH: relative humidity
    WHO: World Health Organization

    Edited by A Mavragani, T Sanchez; submitted 25.02.23; peer-reviewed by F Huang, A Peden, Z Zhang; comments to author 18.04.23; revised version received 05.05.23; accepted 15.06.23; published 20.07.23.

    Copyright

    ©Yingyin Liu, Xiaomei Dong, Zhixing Li, Sui Zhu, Ziqiang Lin, Guanhao He, Weiwei Gong, Jianxiong Hu, Zhulin Hou, Ruilin Meng, Chunliang Zhou, Min Yu, Biao Huang, Lifeng Lin, Jianpeng Xiao, Jieming Zhong, Donghui Jin, Yiqing Xu, Lingshuang Lv, Cunrui Huang, Tao Liu, Wenjun Ma. Originally published in JMIR Public Health and Surveillance (https://publichealth.jmir.org), 20.07.2023.

    This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR Public Health and Surveillance, is properly cited. The complete bibliographic information, a link to the original publication on https://publichealth.jmir.org, as well as this copyright and license information must be included.