The association between high ambient temperature and risk of hospitalization: a time-series study in eight ecological regions in Vietnam

Viet Nam is among the countries most threatened by and vulnerable to climate change and extreme weather events. However, research on the temperature-morbidity relationship at the national scale has been scarce. This study aimed to assess the impact of high temperatures on the risk of hospital admissions for all causes and heat-sensitive diseases across eight ecological regions in Vietnam. The study utilized a longitudinal dataset that included hospitalization and meteorological data from eight provinces representing eight regions in Vietnam. A time series analysis was applied using the generalized linear and distributed lag models with a quasi-Poisson family to examine the temperature-hospitalization association in each province. A random-effects meta-analysis was used to calculate the pooled estimate of risk for the national scale. The country-level pooled effects (%, [95% CI]) indicated that a 1 °C increase above the threshold temperature (19 °C) increased the hospitalization risk for all causes and infectious diseases by 0.8% [0.4%–1.2%] and 2.4% [1.02%–1.03%], respectively at lag 0–3 d. The effects of heat on respiratory diseases and mental health disorders were not significant. At the regional level, the association varied across eight regions, of which the Northern parts tended to have a higher risk than the Southern. This is among very few national-scale studies assessing hospitalization risk associated with high temperatures across eight ecological regions of Vietnam. These findings would be useful for developing evidence-based heat-health action plans.


Introduction
Climate change, with increasing the frequency and intensity of extreme heat events, has led to detrimental effects on human health [1]. Approximately a third of the world's population has been struggling with extreme heat events for at least 20 d yr −1 , and this figure could rise to 50% by 2100. A large array of studies has shown that heat-related health impacts are numerous and complicated. Exposure to heat can lead to heat-related illness symptoms (such as heat edema, fatigue, and headache to severe fainting, heat exhaustion, and heatstroke), increase the hospitalization risk, exacerbate current chronic conditions, spread outbreaks, and even mortality [2][3][4]. These effects seem to be long-lasting, serious, and costly but difficult to evaluate [5]. However, the World Health Organization declared extreme heat has still not gained adequate consideration despite its serious health impacts [6].
Viet Nam among the most hazard-prone nations in Asia and the Pacific region due to its geographical location, topography, socioeconomic characteristics, and adaptive capacity [7]. Northern Vietnam is characterized by a subtropical climate while the southern regions have a tropical monsoon climate. However, in general, Vietnam has a hot and humid climate with an average maximum temperature is around 28 • C and humidity of roughly 80%. The combination of elevated temperature and humidity obstructs the natural dissipation of heat through evaporation (sweating), disrupts the body's ability to cool down, and amplifies the health hazards associated with heat [8]. According to the high greenhouse gas concentration pathway scenario (RCP8.5 scenario), the pace of warming at the end of the 21st century is projected to increase by 3.3 • C-4.0 • C in the North and 3.0 • C-3.5 • C in the South of Vietnam [9]. A recent health vulnerability and adaptation assessment showed that Vietnam's health sector has been facing a 'high' level of disaster exposure while a 'very low' level of adaptive capacity. These factors make Vietnam highly vulnerable to extreme heat in the current and future. Meanwhile, Vietnam currently lacks evidence-based regulations and guidelines on heat prevention.
In recent years, there have been several studies in Vietnam assessing heat-related health issues both mortality and morbidity [10][11][12][13]. However, most of the existing literature in Vietnam focused on heat effects in individual cities and regions [11,[14][15][16] or particular diseases such as cardiovascular, mental health disorders, and hand-foot-mouth diseases [11,17,18] or among specific age groups such as elders and children [12,16]. Studies about the impacts of temperature on morbidity risk at a national scale have been scarce. Therefore, this study aimed to assess the impact of high temperatures on hospital admission risk across eight ecological regions in Vietnam. This research not only examined the effect of high temperatures on all-cause admissions but also focused on three temperature-sensitive disease groups including infectious diseases, respiratory diseases, and mental health disorders. However, the study did not encompass other disease groups such as cardiovascular, renal diseases, or nutritional and metabolic disorders. These have been purposefully omitted due to their intricate link to temperature change including heat and cold effects that require more advanced statistical analyses [16][17][18]. Also, due to large differences in the incidence rates of cardiovascular diseases and kidney diseases across eight ecological regions in the dataset of this study, the authors suspected some other factors beyond temperature change may predominantly contribute to the prevalence of these diseases. The findings from this study have enabled a better understanding of the evidence base on the relationship between temperature and diseases in Vietnam as well as in low and middle-income countries with subtropical environments.

Research location
The study utilized a longitudinal dataset that included hospitalization and meteorological data from eight provinces in Vietnam. Each province represents one of Vietnam's ecological regions, as defined by terrain, soils, and climate, including the Northwest (Dien Bien), Northeast (Bac Giang), Red River Delta (Ha Noi), North Central Coast (Ha Tinh), South Central Coast (Binh Dinh), Central Highland (Dak Lak), Southeast (Binh Phuoc), and Mekong River Delta (Dong Thap). The location of eight selected provinces is presented in figure 1.

Data collection
Data for daily hospitalization comprised daily admissions due to all causes, infectious diseases, respiratory diseases, and mental health disorders. Supplementary (table S1) presents ICD 10 codes for selected disease groups. The data was extracted from daily admission records of provincial hospitals consisting of admission and discharge dates, residential locations, and discharge diagnoses according to the International Classification of Diseases-10th Revision (ICD 10) code. Hospitalization data excluded patients who were not residents of provinces in our study. Depending on the availability and quality, data for a specific province were available for 5-10 years (from 2005 to 2020).
Daily meteorological data in each province were obtained for the same period from the provincial hydro-meteorological stations. Weather data consisted of daily minimum, maximum and average temperature ( • C) and relative humidity (%). Previous studies use different temperature metrics (daily mean, maximum, minimum temperature, daily mean heat index, or maximum-minimum temperature range) to assess heat effect. However, some large-scale studies suggest that daily mean temperature may be a better indicator of heat-health relationships [19,20]. (Note: This map was created from the shapefile (an empty map) that is publicly available from http://gadm.org)

Statistical analysis
We carried out the analysis in two-part modeling methods consisting of a similar model applied to each province, followed by a meta-analytic approach to produce the country-level pooled estimate [10,13,21,22]. Viet Nam has a primarily tropical monsoon, with high heat and humidity [23]. Therefore, in the current study, heat effects on the risk of hospitalization were analyzed for a whole year to better understand the nature of climate conditions in Viet Nam. Missing temperature records were calculated using the average value of the previous day and the following day. In the first stage, a time series analysis was applied using the generalized linear and the constrained distributed lag models (DLM) with a quasi-Poisson family to examine the short-term relationships between ambient temperatures and hospitalizations. We chose to use DLM after doing sensitivity analyses (described below) and based on some previous studies in Viet Nam [24,25]. In order to control the seasonality and long-term trends, we used a natural spline of time with equally spaced knots and four degrees of freedom (4df) per year [26,27]. The models also integrated the humidity variable (by using a natural spline of time with equally spaced knots and 3df) and days of the week [27]. The cumulative lag effects of temperature on admission were also investigated over a 3 d period (lag 0-1 to lag 0-3). To assess the hospitalization risk of temperature, we calculated the minimum hospitalization temperature (MHT), corresponding to the specific temperature associated with the lowest hospitalization risk in each city based on the cut-off values of 5.0th and 97.5th temperature percentiles, referred to previous studies [14,28]. The model to evaluate the temperature-hospitalization association is presented in the following equation where Y t is the daily count of hospital admission on day t, α is the intercept, Ti is a network obtained by applying the 'cross-basis' DLM functions to the average daily temperature on day i and l is the max lag, H is the daily average humidity, ns(time) is the natural spline function of time (4 knots per year), DOW is the day of the week. In the second stage, a random effect meta-analysis [24] was applied to estimate the country-level pooled temperature-hospitalization effects for all causes and specific diseases. Statistical analysis was conducted using R software (version 4.0.2) [29].
Sensitivity analyses were performed to test the robustness of our results according to quasi-Poisson Akaike information criterion (Q-AIC) [30], in which we compared between using DLM and DLNM and ran separate models by modifying the lag structure (lag 0-3, 0-7, and 0-14 d). The findings showed that DLM and lag 0-3 were more optimal with the smallest QAIC index (supplementary table S3). Table 1 presents the descriptive statistics of the data in eight provinces representing eight ecological regions in Vietnam. The total number of hospital admissions by all causes at eight provincial hospitals was 1614 286 cases. The daily number of hospital admissions fluctuated with mean values ranging between 47.4-140.0 cases across the provinces. Respiratory diseases are the most common among the three selected heat-sensitive diseases. During the study period, the mean temperature ranged from 22.1 • C (Northwest) to 28.3 • C (Southeast). The daily mean relative humidity ranged from 73.9% to 83.3% for the Northwest and the Mekong River Delta, respectively. Time-series plots for daily hospitalization and daily mean temperature ( • C) of eight regions were presented in supplementary figure S1.

The province-level effect of temperature on hospitalization in eight ecological regions
The effects of temperature on all-cause and cause-specific hospital admissions were most notable in the short term, specifically within a cumulative lag of 0-3 d (table 2). The magnitudes of temperature-hospitalization  Regarding infectious disease admissions, except for Southeast and Central Highland regions, consistent negative impacts of temperature on hospitalization were observed from cumulative lag 0-1 to lag 0-3 d. Northern regions posed a higher risk compared to the Southern parts. Specifically, the North Central Coast region showed the highest increased infectious admission risk (4.0%, [2.4%-5.6%]), followed by the Northeast (3.0% [1.3%-4.7%]) and the Northwest (2.9% [0.7%-5.1%]). In terms of admissions for respiratory disease, the effects varied across different regions. In the Northwest, the risk of respiratory hospitalization significantly increased by 1.3% [0.2%-2.5%] for each 1 • C increase above the thresholds at the lag of 0-3 d. Conversely, in the Red River Delta, the risk decreased by −1.0% [−1.7-(−0.3)]. The effects of heat on respiratory diseases were insignificant in the remaining regions. Regarding mental health disorders, non-significant effects of temperature were observed in all regions except North Central Coast. A 1 • C increase in temperature after the threshold was associated with a rise of 4.5% [0.6%-8.6%] in the risk of mental health disorder admissions in the North Central Coast (table 2). Models for other cumulative lags (from lag 0-1 to lag 0-3) were presented in supplementary (table S2).

Country-level pooled effects
Country-level pooled effects of short-term ambient temperature on all-cause hospital admissions and heat-sensitive diseases are shown in figures 2 and 3, respectively. In general, at lag 0-3 d, a 1 • C increase above the threshold temperature (19 • C) has been significantly associated with a 0.8% [0.4%-1.2%] increase in all-cause hospitalization. In addition, the findings indicated an association between a 1 • C increase and the increase in infectious disease admissions by 2.4% [1.8%-3.0%]. The association between average temperature and the risk of hospital admissions due to respiratory diseases and mental health disorders was not significant in this study.

Country-level pooled effects of high temperatures
This was a nationwide study based on a multi-region dataset that examined the short-term effects of high temperatures on the risk of hospitalization in eight ecological regions of Vietnam. The country-level pooled effect indicated a significant relationship between temperature increases above the threshold (19 • C) and the risk of hospitalization for all-cause and infectious diseases within a short-term exposure period of 0-3 d. Nevertheless, the association between high temperatures and respiratory diseases and mental health disorders was not significant.
Previous studies found a similar pattern of heat effect on all-cause admissions while findings for specific diseases varied across studies [10,13]. For example, research in Northern Vietnam revealed that a rise in temperature above 24 • C on the same day had a significant impact on hospitalization for all-cause, respiratory, and infectious admissions [13]. In addition, a study of 16 climatic zones in California found that higher temperatures increased mental health admissions by 4% while it decreased respiratory admissions by 7% [31]. A study in china reported the effects of moderate heat on the hospitalization risk for total respiratory diseases and specific sub-categories including asthma, chronic obstructive pulmonary disease (COPD), and bronchiectasis [32]. However, another study found that increased temperatures only affected total respiratory diseases in the winter, not in the summertime [33].
The effect of high temperature on infectious diseases is in agreement with some previous studies [13,34]. However, it should be noted that the effects of temperatures on specific infectious diseases such as hand-foot-mouth disease, rotavirus, and influenza were different [17,35]. Also, plausible mechanisms behind the temperature-infectious relationship are complex and still unclear. Previous studies found that high temperatures could increase the survival, distribution, and susceptibility of disease vectors which increases the risk of outbreaks [36,37].
The contrasting findings about temperature-associated hospitalization for specific diseases could be attributable to the difference in the socioeconomic and population structure. For instance, chronic diseases are more prevalent among the aged population [38]. In addition, hospital admission may not be a reliable indicator to evaluate the effects of heat on some diseases since patients in severe cases may die before seeking medical attention. Furthermore, it is noted that for the same category of diseases, the effect of heat on sub classifications may be different. For example, a study in California concluded that the risk of hospitalization for acute respiratory proved to be more sensitive to heat than to chronic outcomes [31].

Region-level effects of high temperatures
The risk of hospitalization varied from region to region. For infectious diseases, the effects of temperature tent to higher in the Northern (including Northwest, Northeast, Red River Delta, and North Central Coast) than in the Southern (including Southern Centre Coast, Central Highland, Southeast, and Mekong River Delta). Meanwhile, the risk of hospital admissions for mental health disorders was highest in the North Central Coast region and the risk for respiratory diseases was highest in Northwest. The different findings across regions could be attributed to several factors, such as demographic characteristics, socioeconomic features, healthcare systems, and community adaptability. For example, provinces with high elderly populations or youth children under 5 years old are more prone to respiratory diseases during hot weather. A study in the Vietnam Mekong Delta region identified positive associations between socioeconomic factors (population density, poverty rate, illiteracy rate) and temperature-related hospitalizations, while negative associations were observed with indicators such as access to safe water, standard hygienic toilets, and rural population percentage [10]. Also, our results might be partly explained by the level of heat acclimatization of residents. With a tropical climate all year round, residents in the Southern may better adapt and cope with the hot conditions than residents in North Vietnam where has a humid and subtropical climate. For example, the Central Highland region has a high average daily temperature and the smallest temperature deviations between the hottest and coldest months. This may partly explain why the effect of heat on hospitalization in this region was not significant. This assumption was supported by a national-scale study in Brazil that the impact of heat-related hospitalization risk was most prominent in cold zones [20]. Also, a previous study in the Eastern United States has provided evidence for the health effect of high temperatures as latitude increment increases [39]. Furthermore, it has been proven in some studies that there are spatial and temporal differences in the heat-health association [40,41]. This suggested the importance of both national-and subnational-scale studies to develop an appropriate heat-health action plan and identify vulnerable areas.
It is worth noting that the impact of high temperatures on the risk of hospitalization is intricate and operates through various pathways. The body's thermoregulation system tightly coordinates all major human systems, including the cardiovascular, nervous, gastrointestinal, renal, hematologic, and tegumentary systems [42]. External extreme heat can disrupt the body's temperature regulation mechanisms, disturbing its balance and making individuals more susceptible to both acute and chronic health issues [43]. Additionally, thermoregulation can be influenced by medications for chronic conditions and stimulants [44].
The effect of heat on human health can be indirect through the effect of high temperatures on promoting the survival, reproduction, and transmission of disease-causing pathogens, increasing the risk of infectious diseases [45]. Also, the biological mechanisms that respond to high temperatures are influenced by demographic factors, individual behaviors, socioeconomic and environmental conditions, and the implementation of preventive measures [43]. These factors collectively contribute to the intricate relationship between ambient temperature and the probability of hospitalization.
We acknowledge some limitations in this study. Firstly, since the research used admission data from provincial hospitals, several cases admitted to lower-level hospitals (district hospitals) were not considered. Also, some cases might be admitted to the district hospital for a few days before being transferred to the provincial hospital. Hence, there was a possibility that the date of admission to the provincial hospital did not necessarily represent the date when the health issue first arose. In spite of this, the research still represents the majority of hospital admissions. Secondly, weather data was collected from the central monitoring station of each province, and the study's assumption was that residents of each province were exposed to the same weather conditions. Hence, individual exposure levels may not be accurately assessed due to variations in weather among suburbs. Thirdly, we were unable to access individual factors such as the geographic, socioeconomic, and demographic characteristics in all eight ecological regions because of limited data sharing and accessibility, which is a common issue in developing countries like Vietnam. Lastly, each ecological region only selected one province; therefore, the study's generalizability may be limited since the findings may not represent the entire region.

Conclusion
This is a large-scale study that quantified the effect of high temperatures on hospitalization risk in eight ecological regions in Vietnam, a country with low adaptive capacity and a high vulnerability to climate change. Research findings contributed to our understanding of the morbidity burden attributable to ambient temperature, particularly the association between heat and infectious diseases. Also, the level of effects varied across eight regions. Accordingly, it is essential to identify heat-vulnerable regions and develop local adaptation strategies. For instance, depending on each region's temperature threshold, heat-health warning systems should be applied differently. Moreover, heat vulnerability maps can be developed to identify high-risk populations in order to provide timely support.
For further research, stratified analyses by geography, socio-economic, and latitude characteristics are recommended to better evaluate the effect of high temperature and identify high-risk populations and areas. In addition, instead of classifying disease in the broad cause grouping such as respiratory and mental health disorders, studies about heat effects on the specific sub-category of diseases are needed to have a better understanding of the association. For example, for respiratory disease groups, sub-groups can be asthma, COPD, and pneumonia. Furthermore, digital health and meteorological data should be improved and developed to provide more accurate evidence of the heat-health association.

Data availability statement
The data cannot be made publicly available upon publication because they contain sensitive personal information. The data that support the findings of this study are available upon reasonable request from the authors.