https://orcid.org/0000-0002-8494-2585
Figures
Abstract
Climate change has led to unpredictable, intense temperature fluctuations during warming winters, particularly in subtropical regions. Among that, abrupt shifts from warm to cold temperatures over short periods pose health challenges yet remain poorly understood. Historical data from Hong Kong revealed a rising trend in the frequency of temperature flip events, characterized by larger temperature drops and faster cooling rates. The temperature flip events were associated with increased risks of total non-external (RR: 1.08, 95% CI: 1.05-1.11) and cause-specific mortality, with a higher risk observed at or below the cold weather warning threshold (RR: 1.13, 95% CI: 1.09–1.17). Greater magnitude of temperature drops, shorter durations, and more rapid cooling during flip events intensified the mortality risk, while the starting temperature showed an inverse relationship. Events that began with warming temperature followed by rapid cooling were particularly associated with increased pneumonia mortality risk (RR: 3.17, 95% CI: 2.14–4.68), while cold-start flips with rapid (RR: 1.21, 95% CI: 1.14–1.29) or slow (RR: 1.08, 95% CI: 1.03–1.13) temperature declines were linked to increased total mortality. The health impacts varied across genders and causes of death, with the elderly being especially vulnerable under both cold-slow (RR: 1.11, 95% CI: 1.04–1.18) and cold-rapid (RR: 1.20, 95% CI: 1.11–1.30) flip scenarios. These findings highlight that the increased mortality risk among vulnerable populations caused by sudden temperature flips may be overlooked. The potential neglect underscores the urgent need to enhance cold weather warning systems and health interventions to better protect vulnerable populations from the adverse effects of temperature shifts in aging society of subtropical climate regions.
Citation: Wang Y, Liu S, Ren C, Woo J, Chong KC, Ng E (2026) Increased mortality risks of winter temperature flips: A growing concern in aging society of subtropical climate regions. PLOS Clim 5(6): e0000859. https://doi.org/10.1371/journal.pclm.0000859
Editor: Noureddine Benkeblia, University of the West Indies, JAMAICA
Received: November 22, 2025; Accepted: March 17, 2026; Published: June 10, 2026
Copyright: © 2026 Wang et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: Daily meteorological data can be found on the Hong Kong Observatory website (https://www.hko.gov.hk/en/index.html), while daily air pollution data are publicly available from the Environmental Protection Department of Hong Kong (https://www.epd.gov.hk/epd/english/top.html). Mortality data was collected from the Hong Kong Census & Statistics Department, and the data is available upon request (https://www.censtatd.gov.hk/en/page_1338.html, Email: population@censtatd.gov.hk). The tutorial R code for the DLNM model is available on the package author’s personal page (http://www.ag-myresearch.com). Code for events identification and their health risk assessment are provided in the Text A and Text B in S1 Appendix. The dates of the identified events are available upon request to the corresponding author (renchao@hku.hk) and is also available by running the code with the historical meteorological data.
Funding: This work was funded by the Research Impact Fund 2022/23 of the Hong Kong Research Grant Council (Project Ref. No. R4040-22). Prof. Edward Ng of the Chinese University of Hong Kong is the Project Coordinator. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
1. Introduction
Global climate change and population aging are two prominent and interlinked concerns of the 21st century, with the impact of temperature-related extreme weathers on health emerging as a significant global public health challenge [1,2]. According to the Global Burden of Disease (GBD) 2019, low temperatures accounted for a higher percentage of global deaths due to noncommunicable disease than high temperatures, patrtically among the elderly population. In addition to extreme low or high temperatures, increasing concerns have been raised about temperature fluctuations between warm and cold conditions and the related health risks. Previous studies have established strong positive associations between mortality, hospitalization, and winter temperature fluctuations, including consecutive day-to-day temperature change (TCN) [3], diurnal temperature range (DTR) [4], and overall temperature variability (TV) [5]. These links are particularly pronounced for cardiovascular and respiratory diseases, especially among the elderly. In general, older adults experience a cumulative physiological response to cold weather exposure [6–8]. The human body’s physiological reactions to a temperature drop, which can begin even before the cold weather warning threshold temperature is reached, are influenced by both the magnitude and rate of the temperature decrease. Research indicates that the body’s primary regulatory responses to a drop in ambient temperature include peripheral vasoconstriction [9], an increase in heart rate and blood pressure [8], and elevated inflammatory blood markers [9–11]. These reactions place a greater strain on the cardiovascular and respiratory systems of older adults, as their body’s blood vessel walls become less elastic with age [12] and pre-existing chronic immune inflammation further amplifies the physiological stress during sudden temperature drops [13]. Under the context of accelerating global population aging, there is a critical need to investigate the health impacts of cold weather on the elderly.
Previous research examining the impacts of cold weather exposure on health outcomes of older adults has predominantly used temperature thresholds (such as fixed values [14] or historical percentiles [15]) and single-dimensional metrics (e.g., frequency, duration, or intensity) to define cold events. However, subtropical regions have experienced increasingly frequent and severe temperature fluctuations [16,17] in a warming winter. Such phenomenon is linked to Arctic amplification [18], which disrupts the jet stream and can destabilize the stratospheric polar vortex, thereby enhance the advection of warm and cold air masses and leading to more extreme temperature variability [19,20]. Furthermore, populations in subtropical cities exhibit a higher vulnerability to such fluctuations, primarily due to their long-term physiological adaptation to a warmer climate that leaves thermoregulatory mechanisms less prepared for cold stress [21,22], and behavioral adaptation where the rarity of extreme cold leads to inadequate protective responses [23]. Evidence from studies conducted in subtropical cities such as São Paulo [24], Guangzhou [25], and Taipei [26] underscores the critical need for greater attention to cold-related health risk in these areas.
Regarding cold weather warnings, most warning systems worldwide rely on absolute low-temperature thresholds as their primary trigger. While some systems may incorporate metrics such as a 24- or 48-hour temperature drop, warnings are typically activated only upon reaching a critical low temperature. This approach limits their effectiveness in preventing and mitigating health risks caused by temperature fluctuations, and their corresponding health impact still remains poorly understood. Indices of temperature fluctuations in existing research primarily focus more on a relatively normal range, thereby inadequately capturing the health impacts of sudden temperature drop events characterized by large fluctuations during winter seasons. This is particularly challenging because temperature change is a nonlinear, non-stationary process with multi-scale oscillations. Thus, it would be timely needed to explore the relationship between winter sudden temperature drop events and their corresponding public health risks in subtropical climate regions.
In this study, we go beyond the traditional methods that rely on absolute or percentile-based temperature thresholds to investigate the public health impacts of overlooked temperature fluctuations from warm to cold. Hong Kong, one of the world’s most rapidly aging cities, where more than 30% of the population is projected to be aged over 65 years by 2039 and 40.6% by 2050, is used as a case study [27]. Furthermore, residential and public buildings in subtropical regions typically prioritize ventilation and cooling over winter thermal insulation [28–30]. Coupled with the limited installation of heating systems, indoor environments in these areas would be susceptible to sudden outdoor temperature drops [31]. The purpose of this study is to identify changing patterns and characteristics of warm-to-cold temperature flip events by using Hong Kong’s long-term meteorological records, and to investigate their associated health risks. Our findings will provide a scientific basis for improving the existing cold weather warning system and developing early prevention services to mitigate cold-related health risks for the old adults, and also serve as an useful reference for other cities in subtropical regions.
2. Materials and methods
2.1. Study setting
This study was conducted in Hong Kong, a coastal subtropical city located in southern China, with a high population density of 6777 person/km2 and a total area of 2755 km2 (Fig 1). The city’s climate features hot and humid summers, along with cold and dry winters. Hong Kong’s population is ageing rapidly. By 2050, Hong Kong is forecasted by the World Health Organization to rank fifth in the world for cities with the largest percentage of older adults, i.e., 40% of the population in Hong Kong will be aged 65 years or above. Census data from the last three decades indicates a rising trend in the elderly population proportion in each of Hong Kong’s 18 districts (Fig A in S1 Appendix).
2.2. Data collection
Daily mortality data from November to March for the years 1995–2021 was obtained from the Hong Kong Census & Statistics Department. The International Classification of Diseases (ICD-9, ICD-10 and ICD10-updated) was used for recording the causes of deaths. Cause-specific mortalities include total non-external deaths, non-cancer non-external deaths, cardiovascular deaths, influenza deaths, pneumonia deaths, and deaths attributed to symptoms, signs and abnormal clinical and laboratory findings that not elsewhere classified. Additionally, information regarding the age and gender of each death was gathered to enable subgroup analyses.
Daily meteorological data including daily minimum temperature (°C), mean relative humidity (%), windspeed (km/h), and rainfall were represented by the weather station located in the city center and obtained from the Hong Kong Observatory. Daily mean concentrations of inhalable particulate matter (PM10, ug/m3) from 13 general monitoring stations across the city were collected, and the average from these stations was used to represent overall exposure.
2.3. Warm to cold temperature flip identification
A warm-to-cold temperature flip event is defined as a short period in which the daily minimum temperature (Tmin) suddenly shifts from a warm state to a cold one. Different from indices such as DTR, TCN, and TV, which primarily quantify intraday or interday temperature fluctuations within a relatively normal range, a warm-to-cold flip event represents a transition from an anomalous warm state to a subsequent cold one, capturing the health impacts of sudden multi-day transitions rather than daily fluctuations. To identify these events, a warm/cold day is first detected when the detrended Tmin exceeds/falls below the climatological average by 1 standard deviation (s.d.). A warm-to-cold temperature flip event was then confirmed if a warm day was followed by a cold day within five days. To ensure the robustness of our findings, we also defined the flip period using seven-day intervals, which yielded similar patterns of increasing annual frequency, decreasing duration, and a rising cooling rate (Fig B in S1 Appendix).
To exclude the effects of long-term warming on temperature flip events, we first detrended the daily Tmin series since 1884. This process utilized a 31-day rolling window calendar-day detrending method to separate the seasonal variability, which is the season-dependent climatological average temperature. Based on the detrended temperature data and in alignment with WMO climate normals, we used 1991–2020 to calculate the multi-year daily average and the corresponding standard deviation
for each calendar day. This 30-year span accounts for the recent warming winter climate, thereby ensuring a more accurate and unbiased baseline for detecting warm and cold day anomalies. These standard deviations are derived from detrended data to ensure they reflect natural climate variability rather than being biased by the pronounced warming trend observed over the past 30 years. We then used the average plus or minus one standard deviation as a threshold to identify and mark warm and cold days across the long time series. The daily detrended Tmin average and standard deviation series are plotted in Fig C in S1 Appendix.
A day’s status is marked as follows:
• is the status of the day.
• is the detrended temperature on day i.
• is the climatological mean for calendar day
.
• is the corresponding standard deviation.
• Building on this, a warm-to-cold flip event was detected when a warm day was followed by a cold day within five days (Fig 2). An event was then marked as a continuous period starting from the initial warm day and including the first and all subsequent consecutive cold days.
2.4. Statistical analysis
2.4.1. Warm to cold temperature flip event groups.
We systematically analyzed the characteristics of warm-to-cold temperature flip events in Hong Kong, focusing on their frequency, drop magnitude, cooling rate, duration, and starting temperature. Event frequency was defined by the 5-year growth rate of winter events since 1884. For each warm-to-cold flip event, we also analyzed its specific characteristics over the past 140 years, the past 60 years, and the past 30 years: duration was the time interval from the first day of the warm period to the coldest day within the event; drop magnitude was the absolute difference between the Tmin of the starting warm day and the coldest day; the cooling rate was the ratio of the drop magnitude to the duration; and the starting temperature, defined as the Tmin of the first warm day of the event, was used to describe the initial temperature conditions that triggered such events within the Hong Kong’s warming winters context.
To investigate the health impacts of warm-to-cold flip events, we clustered the characteristics of “warm-to-cold flip” events, using the 75th percentile of winter event metrics since 1884 as a threshold. This threshold was selected to capture high-intensity exposure characteristics while ensuring sufficient sample sizes for statistically robust health impact analysis, a strategy consistent with established environmental health methodologies [32]. Accordingly, all winter events were categorized into four groups: high/low starting temperature, high/low drop magnitude, rapid/slow cooling rate, and long/short duration. Following this initial categorization, we further combined the ‘starting temperature’ and ‘cooling rate’ characteristics to define four distinct event types: Cold-slow, Cold-rapid, Warm-slow, and Warm-rapid events, to better understand their specific health risks. This grouping provides a dynamic perspective on the evolving characteristics of these events from their onset to sustained phases. We further analyzed how “warm-to-cold” events were associated with Hong Kong’s cold weather warning system by examining if these events triggered the cold weather warning threshold, and if yes, how the above- and below-threshold components affect human health. This work will support optimizing current static-threshold warning models to effectively incorporate dynamic “warm-to-cold” events, crucial for subtropical cities facing more frequent and more severe rapid temperature drops.
2.4.2. Health effect of the temperature flip events.
A quasi-Poisson generalized linear model combined with a distributed lag nonlinear model was applied to evaluate the impact of sudden temperature flip exposures on human mortality risk [33]. The model is expressed as:
where μt is the expected number of daily deaths on day t. β0 represents the intercept. The cb(•) denotes a cross-basis function designed to capture the delay in the effects of the temperature flip events and to account for short-term harvesting. Flip is a categorical variable represents the presence of temperature flips (yes/no) or flips with different characteristics. The maximum lag period was set as 28 days as such lag was tested to be able to capture the full effects of sudden flip events. The natural cubic spline function ns(•) denotes a smoothed relationship between log(μt) and the other factors of day t. Meteorological and air pollution confounders including windspeed, relative humiaity, and PM10 were adjusted as potential confounders. Both the year and the day of a year (i.e., doy) were included to control the long-term trends and seasonality, while day of week (i.e., dow) effect was also considered. The logged annual population log(Nt) was included as an offset, while holidays indicates whether day t is a public holiday. The degrees of freedom was selected based on quasi Akaike information criterion (QAIC) [34].
2.4.3. Subgroup analysis.
Subgroup analysis included evaluating the impact of warm-to-cold temperature flip events on various cause-specific mortalities. The effects of temperature flip exposure were also stratified according to gender and age groups (i.e., 0–17 years, 18–44 years, 45–64 years, 65–74 years, and 75 years and above).
2.4.4. Sensitivity analysis.
Sensitivity analysis was conducted by changing the lag period to 21 days (a commonly tested lag period) [35,36]. To ensure the robustness of our findings, we also tested seven days intervals to define the flip period, which yielded similar results with a trend of increasing annual growth, decreasing duration, and a rising cooling rate. Temperature flip events were identified with Python, while health impacts were assessed with R software (version 4.3.2) with “dlnm” and “splines” packages. P < 0.05 was considered as statistically significant.
3. Results
3.1. Increasing frequency and intensity of warm-to-cold temperature flip under warming winters
As subtropical regions experience warming winters and people become accustomed to the decrease in extreme cold days, our research reveals an increased frequency in warm-to-cold temperature flips with an annual rate of approximately 0.76% (Fig 3). These events are reshaping Hong Kong’s winter climate pattern. While the traditionally defined winter months are from December to February in next year, a new pattern has evolved over the last three decades and we found that winter warm-to-cold temperature flip events are most prominent in January, with an increasing frequency also observed in March. This indicates that sharp drops in temperature are on the rise, extending beyond the traditionally defined winter season and into the winter’s shoulder months and early spring. From a dynamic perspective of the events’ evolution, their starting temperature has been on an upward trend over the past century, rising from about 19°C to nearly 22°C (Fig 4). This trend has been particularly pronounced in the last 30 years, with an annual increase of about 0.1°C, which aligns with the context of global warming. Further analysis reveals that for about 60% of all such flip events (Fig 3), their daily minimum temperature could reach and drop below 12° C, the 10th percentile of daily minimum temperatures from Hong Kong’s 1991–2020 Climatological Normals and the trigger point of the local cold weather warning.
Note: Non-triggered events refer to warm-to-cold drops where temperatures remain above the Hong Kong’s cold weather warning (CWW, i.e., 12°C) level. Conversely, CWW-triggered events involve a temperature drop that could reach the CWW threshold. Three patterns including 5-year event growth rate (top-left), Monthly distribution of events (top-right), and percentage of non-triggered/triggered events (bottom) are shown.
The cooling process during “warm-to-cold temperature flip” events in Hong Kong’s warming winter has intensified as evidenced by two key trends: a slight upward trend in drop magnitude with an increas rate of approximately 0.6°C per decade and an accelerated cooling rate with a decadal increase of about 0.01°C/day. These combined changes signal a trend toward more sudden and severe temperature drops. Conversely, event duration shortened, with the downward trend over the past 30 years more than twice that of the overall period (Fig 4). In a warming world, Hong Kong’s winter warm-to-cold temperature flip events are becoming more frequent. These events are evolving with a shorter duration and rapid temperature drop, posing a new challenge for public health, particularly for the elderly in an aging society.
3.2. Warm-to-cold temperature flips in winter pose significant mortality risks
From 1995 to 2021, the cause-specific mortality rates per 100,000 individuals during the winter seasons in Hong Kong were as follows: 243.82 for non-external causes; 166.91 for non-cancer, non-external causes; 66.36 for cardiovascular deaths; 0.36 for influenza; 36.47 for pneumonia; 53.25 for respiratory causes; and 5.22 for symptoms, signs, and abnormal clinical or laboratory findings not classified elsewhere.
The impact of warm-to-cold temperature flip events on mortality risk was detailed in Table 1. Compared to non-event days, days with flip events were associated with an increased risk of overall non-external mortality (RR: 1.08, 95% CI: 1.05–1.11). Similar elevated risks were observed for non-cancer non-external (RR: 1.10, 95% CI: 1.06–1.14), cardiovascular (RR: 1.11, 95% CI: 1.06–1.17), influenza (RR: 4.35, 95% CI: 1.64–11.51), pneumonia (RR: 1.09, 95% CI: 1.01–1.16), and respiratory deaths (RR: 1.10, 95% CI: 1.04–1.17). Stratified analyses revealed that all seven causes of death experienced increased risks when the daily minimum temperature of an event finally reached or fell below the cold weather warning threshold, regardless of the specific characteristics of these events.
Compared to males, the overall impact of temperature flip events was more pronounced among females, particularly for influenza-related mortality (RR: 6.35, 95% CI: 1.26–32.00) and cardiovascular mortality (RR: 1.15, 95% CI: 1.07–1.23). The elderly population, especially those aged 75 years and older, experienced the most significant increase in mortality risk. Similar patterns of increased risk were observed during events with Tmin dropped to the cold weather warning threshold or below. Conversely, exposure to temperature drops above cold weather warning threshold appeared to have protective effects in certain age and gender groups (Table A in S1 Appendix).
Characteristics of starting temperature, drop magnitude, cooling rate, and duration for winter events categorized into high/low levels based on the 75th percentile are shown in Table 2, while the corresponding mortality risks are presented in Table 3. Compared to non-event scenario, events that began at relatively cold temperatures were associated with higher mortality risks than those starting at warmer temperatures, particularly for influenza-related deaths (cold starting: RR: 4.83, 95% CI: 1.44–16.20). Events characterized by larger temperature drops and shorter durations were also linked to more remarkable increase in mortality risks than those with smaller drops and longer durations. When these two factors were combined into the cooling rate, higher cooling rates corresponded to more substantial increases in mortality risk for non-external causes (RR: 1.20, 95% CI: 1.14–1.28) and other causes, except for death of influenza and symptoms, signs and abnormal clinical and laboratory findings.
When combining the starting temperature and cooling rate of each event, those with a warming starting point followed by a rapid temperature decline were associated with the most significant increases in pneumonia (RR: 3.17, 95% CI: 2.14–4.68) and respiratory (RR: 1.92, 95% CI: 1.39–3.65) related mortality. Cold-starting events with rapid cooling were linked to higher non-external mortality risks (RR: 1.21, 95% CI: 1.14–1.29) compared to those with slower cooling rates (RR: 1.08, 95% CI: 1.03–1.13). Cold-slow events also showed elevated mortality risks across several causes (Fig 5). Both males and females were vulnerable to some types of temperature drop events. Compared to males, females experienced a higher risk of non-external mortality during cold-rapid events (RR: 1.22, 95% CI: 1.12–1.32) and warm-rapid events (RR: 1.25, 95% CI: 1.01–1.54). The gender differences were also observed in other cause-specific mortality risks.
Across different age groups, the elderly, particularly those aged 75 years and older, generally experienced the highest non-external mortality risks during cold-slow (RR: 1.11, 95% CI: 1.04–1.18) and cold-rapid (RR: 1.20, 95% CI: 1.11–1.30) temperature flip events. Similar patterns were observed for non-cancer non-external, cardiovascular, and respiratory related deaths. Individuals aged 17 years and below were observed with increased non-cancer-non-external mortality risk (RR: 1.64, 95% CI: 1.01–2.67) during cold-slow flips (Fig D in S1 Appendix).
When categorizing each flip event that reached the cold weather warning threshold (12°C) into above- and below-threshold components, increased mortaility risks were observed in both groups (Fig 6). For the above-threshold component, nearly all flip events were associated with an increased non-external mortality risk, particularly among the elderly. However, warm-rapid flip events were linked to a declined but not significant risk, indicating less robust estimates due to the rarity of these warm-rapid flip events. In the below-threshold component, increased non-extrenal mortality risk were observed across all flip events, except for warm-rapid flips, where the small sample size restricted scientific estimations. Similarly patterns were observed for other cause-specific mortalities. See Table B in S1 Appendix for more details.
Risks were estimated for the above-threshold (A, B, C) and below (D, E, F) components of the temperature flip events.
Sensitivity analysis showed robust relationships between temperature flip events and mortality risks. When the maximum lag was set to 21 days, the cause-specific mortality risks associated with temperature flip events increased as well, especially for events reaching the cold weather warning threshold temperature. The highest mortality risks were observed in events that began with warm starting temperatures and rapid cooling rate, followed by those starting from cold points with rapid cooling rates (Table C-D and Fig E in S1 Appendix).
4. Discussion
This study provides a comprehensive evaluation of the evolving patterns of winter warm-to-cold temperature flips and their associated health impacts in Hong Kong, a subtropical city with an increasingly aging population. The findings revealed a rising frequency of winter warm-to-cold temperature flip events, with larger drop magnitudes, shorter durations, and faster cooling rates. Regarding health effects, these temperature flip events, especially those triggerred cold weather warnings, were significantly linked to increased cause-specific mortality risks. The risks varied between genders, but generally increased with age, with individuals over 75 years being the most vulnerable to sudden warm-to-cold temperature flips.
4.1. New pattern of warm-to-cold temperature flips
Climate change is leading to warmer winters with more frequent and intensified warm-to-cold temperature flips in subtropical climate regions [16]. Since stronger synoptic variations in mid-latitudes are associated with eddy activities and frontal systems, which occasionally bring cold air masses from high latitudes and warm air masses from low latitudes [37], thereby causing more frequent abrupt temperature changes in these subtropical climate regions. Such rapid temperature drop events can lead to unpredictable health risks for vulnerable populations given growing adaptation to warmer winters in subtropical regions [38,39]. Compared to existing research [16], our study not only demonstrates an increasing trend in the frequency and intensity of warm-to-cold temperature flip events in Hong Kong, but also uncover a new pattern that such flip events are no longer limited to the traditional winter period of December to February, they now also occur in the winter’s shoulder months (March) in Hong Kong.
Rapid temperature drops starting at a higher temperature are becoming a critical component in defining cold weather exposure during warming winters, differing from the literature focus on prolonged exposure to low daily temperatures [3,40,41]. Our research results suggest that, in addition to previously demonstrated health risks associated with prolonged periods of low daily temperatures, the sudden temperature drops in a warming winter could also be a crucial factor in human environmental adaptation. Due to limited time to adapt to rapid temperature shifts, warm-to-cold temperature flip events are likely to exert greater stress on the human body, especially for vulnerable groups such as the elderly and those with pre-existing health conditions. Such events occur over a few days, often go unnoticed by the public because the starting weather is still warm even in winter time and such events may not trigger the cold weather warning systems. This may lead to adverse health outcomes, particularly in Hong Kong, where the vulnerability could be magnified by rare installation of heating systems and thermal insulation in residential building. This is evidenced by several recent rapid temperature drops in Hong Kong’s winters that did not prompt a warning but resulted in negative health consequences among the elderly and those with poor living conditions [5,42]. Our study highlights that 60% of warm-to-cold temperature flip events could trigger the cold weather warning, their health impacts at the beginning of the event are neglected, i.e., the first 1–2 days of such events, since the daily minimum temperature has not dropped below the trigger point. Therefore, beyond focusing solely on prolonged low-temperature events, cold weather warning systems urgently need to be improved by adding the pre-caution weather alter to capture these increasing, intensifying, and accelerating warm-to-cold temperature flips in subtropical cities. This is crucial for enabling early cold weather warnings and public health interventions for these high-risk events, especially in subtropical cities with growing elderly populations.
4.2. Rising mortality risks in older adults during temperature flips
Previous studies reveal that temperature variations over multiple days are associated with changes in mortality risks, especially among the elderly and for certain cause-specific mortalities. Some studies analyzed the temperature fluctuations between two neighboring days, and reported varying health impacts. For example, Guo et al. found increased risks of non-external and cardiovascular mortality when the temperature dropped by more than 3.0°C between neighboring days in two cities in Australia, with the elderly being at higher risk compared to other groups [43]. In contrast, a nationwide multi-location study in the United States reported an overall protective effect of temperature drops between days [44]. However, this relationship would be modified by seasonal and regional factors, with the elderly and individuals with respiratory and pneumonia disease being especially vulnerable to such temperature variations, particularly in spring. This study examined sudden temperature flip events and found that exposure to these events significantly increased mortality risks, especially for influenza-related deaths, largely due to the relatively low baseline mortality rate and small sample size for this cause. Another study evaluated extreme diurnal temperature ranges and observed a notably higher mortality risk in spring [45]. Traditionally, February marks the end of winter in Hong Kong, while our research indicates that increasingly frequent temperature fluctuations in March, a transitional month leading into early spring, may also contribute to significant yet often underestimated health burdens. Regarding gender differences, previous research suggested that females are more vulnerable to substantial temperature drops between days than males [43,46], while this study indicates that gender differences may vary depending on the cause of death and the type of temperature flip event. Since most current studies primarily focus on intra- and inter-day temperature fluctuations, which are likely to be less impactful due to people’s acclimatization to such temperature changes in the warming winters, few studies have specifically investigated the health impacts of extreme, multi-day warm-to-cold temperature flips. This study provides new evidence that the elderly are generally the most susceptible to such rapid temperature flips, particularly for those aged 75 years and older.
The physiological mechanisms underlying the impact of sudden temperature changes on mortality are complex. Cold temperatures activate thermoregulatory responses that primarily regulate skin and core body temperatures [8]. Previous research suggests that there is a threshold temperature at which these thermoregulatory responses are initially triggered [8]. In our study, temperature flip events that begin at warmer temperatures with slow temperature declines tend to have a less noticeable impact on mortality, which may be partly explained by this threshold effect. In addition, populations in subtropical regions are adapting to warming winters due to climate change. On the other hand, intense temperature drops that exceed the body’s initial tolerance limits may overwhelm the body’s automatic temperature regulation systems, impairing their ability to adapt effectively in the short term. This can lead to increases in blood pressure, immune activation, and inflammatory responses, as well as disruptions to other physiological functions [47].
The gender differences in vulnerability to different kinds of temperature flips could be partially attributed to the complex interactions between biological, behavioral, and socioeconomic factors. Physiologically, variations in thermoregulation, body composition, and hormonal profiles can influence how each gender responds to temperature stress [48]. Behaviorally, males are frequently exposed to extreme temperatures through outdoor activities, while socioeconomic factors like higher rates of poverty, social isolation, and inadequate housing can disproportionately affect women and men, particularly older individuals. Older populations are more vulnerable to sudden temperature drops compared to other age groups [49,50]. Aging can impair thermoregulatory capacity through several mechanisms: it may affect cardiac function by reducing cardiac output and increasing systemic vascular resistance, both of which are vital for temperature regulation [51]. Additionally, aging can directly impact vascular and muscular systems, thereby disrupting the body’s ability to maintain thermal balance [52]. Furthermore, pre-existing chronic health conditions can exacerbate the adverse effects of sudden temperature drops on mortality [53].
4.3. Implications
Given increasingly frequent and intense warm-to-cold temperature flip events in Hong Kong, the current extreme cold weather warning systems, which are triggered by 12 °C of daily minimum temperature, may inadequately capture the health risk faced by the public, especially for the city’s large and increasing aging population. For the older adults, the primary driver of winter health threats is shifting from cold extremes to a combination of abrupt temperature changes and absolute low temperatures, which poses significant new risks to their cardiovascular and respiratory systems. Our study found that 40% of identified temperature flips that do not trigger the local warning system were associated with increased mortality risk. For temperature drops that could trigger the warning system, declines occurring before reaching 12 °C also pose health risks. These findings suggest a critical need to update the cold weather warning system by incorporating the previously overlooked factor of warm-to-cold temperature flips. By referring to the China Grade of Cold Wave, daily minimum temperature drops of 8 °C in 24 hours, 10°C in 48 hours, or 12°C in 72 hours, with the absolute temperature lower than 4°C, could be defined as a cold wave [54]. Similarly, based on our study, daily minimum temperature drops starting from 21.0°C with a cooling rate of 3.0 °C per day could be detected as a sudden temperature warm-to-cold flip event. Given that Hong Kong’s current 9-day weather forecast is sufficient for the local meteorological office to monitor potential temperature drops, it could be used to issue early alerts and special weather advisories, thereby better protecting public health.
The findings of this study offer potential for extrapolation to other subtropical urban centers globally. Although the baseline climates and socioeconomic profiles of other subtropical cities may differ, the fundamental biological and environmental mechanisms linking abrupt temperature transitions to health outcomes remain comparable. The framework established in this study could provide a scalable template for assessing temperature-related health risks. By incorporating detrended minimum temperatures and climatological reference periods, our method ensures that event identification remains sensitive to anomalous transitions even within the context of global warming. Future research in other cities may adapt this methodology by utilizing alternative metrics, such as daily mean or maximum temperatures, and varying standard deviation thresholds to define warm and cold days. These localized calibrations facilitate a more comprehensive understanding of how diverse populations respond to sudden temperature shifts across different geographical contexts.
The findings of this research also underscore significant public health implications, particularly in the context of climate change. The increased mortality risks associated with sudden temperature drops, especially those with a rapid cooling rate and could reach to the cold weather warning threshold temperature, highlights the urgent need for targeted interventions to protect vulnerable populations: the elderly. Compared to the general population, old adults are more vulnerable to extreme temperature fluctuations and lack of strength in thermal regulations, resulting in higher demand for support. However, limited access to heating systems exacerbates their vulnerability and worsens the situation. Critical supports such as community support programs and adequate cold-weather preparations would help mitigate the adverse health impacts of abrupt temperature declines. In addition, long-term initiatives are necessary to enhance resilience against temperature flip-related health risks, especially as the aging population grows. Policymakers may consider integrating these findings into social and community services to better support vulnerable populations. Developing preventive action plans such as improving thermal regulation assistance would help enhance people’s resilience against temperature drop-related health risks, ultimately reducing preventable deaths during winter seasons.
4.4. Limitations
This study has several limitations that should be acknowledged. First, the weather data used to identify temperature flip events were retrieved from a single urban representative station, which may not fully capture intra-urban variations of such flip events within the city. However, the station’s 140-year continuous weather records ensure the results are highly representative of long-term historical trends in temperature flip events. Additionally, this choice is consistent with most environmental health research using urban stations for city-wide exposure [55,56]. Second, outdoor temperature data were used to represent human exposure, potentially leading to inaccuracies since individuals tend to spend most of their time indoors. While factors like building insulation and behavioral adaptations have been discussed [57,58], limited sample sizes and non-standardized indoor data collection complicate a clear link between outdoor-indoor temperature difference and individual perception. As the primary contribution of this study is to establish a methodological framework for identifying warm-to-cold flips and their associated health risks, using outdoor temperature allows the findings to be more readily applied to large-scale epidemiological studies. Third, the study focused only on mortality as a health outcome, thereby limiting the scope of the findings, as health impacts of temperature drops can range from behavior changes, discomfort and mild symptoms to hospitalizations and deaths, highlighting the need for further research into the broader spectrum of health effects. Nonetheless, the findings of this study offer scientific evidence supporting the potential increasing trend in mortality risk associated with sudden warm-to-cold temperature flip events.
5. Conclusion
Sudden warm-to-cold temperature flip events have become more frequent in recent decades, with larger drop magnitude and faster cooling rates. These events are associated with a significant rise in mortality risks, particularly among older populations. Enhanced cold weather warning system and early preventive actions are crucial strategies to help reduce the health risks of sudden temperature flips in subtropical regions under the context of climate change and aging society.
Supporting information
S1 Appendix. Supporting Information for “Increased Mortality Risks of Winter Temperature.
https://doi.org/10.1371/journal.pclm.0000859.s001
(DOCX)
Acknowledgments
Special thanks to Dr. Janice Ho for the proofreading and language editing of this manuscript. Additionally, we acknowledge the data contributions from the Hong Kong Census & Statistics Department and the Hong Kong Observatory.
References
- 1. Costello A, et al. Managing the health effects of climate change: Lancet and University College London Institute for Global Health Commission. The Lancet. 2009;373:1693–733.
- 2. Watts N, Adger WN, Ayeb-Karlsson S, Bai Y, Byass P, Campbell-Lendrum D, et al. The Lancet Countdown: tracking progress on health and climate change. Lancet. 2017;389(10074):1151–64. pmid:27856085
- 3. Zhu Y, Yang T, Huang S, Li H, Lei J, Xue X, et al. Cold temperature and sudden temperature drop as novel risk factors of asthma exacerbation: a longitudinal study in 18 Chinese cities. Sci Total Environ. 2022;814:151959. pmid:34843761
- 4. Wang Z, Zhou Y, Luo M, Yang H, Xiao S, Huang X, et al. Association of diurnal temperature range with daily hospitalization for exacerbation of chronic respiratory diseases in 21 cities, China. Respir Res. 2020;21(1):251. pmid:32993679
- 5. Li Y, Wang X-L, Zheng X. Impact of weather factors on influenza hospitalization across different age groups in subtropical Hong Kong. Int J Biometeorol. 2018;62(9):1615–24. pmid:29804235
- 6. Lin Y, Qi J, Hu J, Liu J, He G, Yin P, et al. Age inequality in temperature-related fall mortality among old people in China in a warming climate. npj Clim Atmos Sci. 2025;8(1).
- 7. Hess KL, Wilson TE, Sauder CL, Gao Z, Ray CA, Monahan KD. Aging affects the cardiovascular responses to cold stress in humans. J Appl Physiol (1985). 2009;107(4):1076–82. pmid:19679742
- 8. Castellani JW, Young AJ. Human physiological responses to cold exposure: Acute responses and acclimatization to prolonged exposure. Auton Neurosci. 2016;196:63–74. pmid:26924539
- 9. Stewart S, Keates AK, Redfern A, McMurray JJV. Seasonal variations in cardiovascular disease. Nat Rev Cardiol. 2017;14(11):654–64. pmid:28518176
- 10. Halonen JI, Zanobetti A, Sparrow D, Vokonas PS, Schwartz J. Outdoor temperature is associated with serum HDL and LDL. Environ Res. 2011;111(2):281–7. pmid:21172696
- 11. Halonen JI, Zanobetti A, Sparrow D, Vokonas PS, Schwartz J. Relationship between outdoor temperature and blood pressure. Occup Environ Med. 2011;68(4):296–301. pmid:20864465
- 12. Zhang X, Zhang S, Wang C, Wang B, Guo P. Effects of moderate strength cold air exposure on blood pressure and biochemical indicators among cardiovascular and cerebrovascular patients. Int J Environ Res Public Health. 2014;11(3):2472–87. pmid:24583830
- 13. D’Amato M, Molino A, Calabrese G, Cecchi L, Annesi-Maesano I, D’Amato G. The impact of cold on the respiratory tract and its consequences to respiratory health. Clin Transl Allergy. 2018;8:20. pmid:29997887
- 14. Lam HCY, Chan EYY, Goggins WB 3rd. Short-term Association Between Meteorological Factors and Childhood Pneumonia Hospitalization in Hong Kong: A Time-series Study. Epidemiology. 2019;30 Suppl 1:S107–14. pmid:31181013
- 15. Lee W, Choi HM, Lee JY, Kim DH, Honda Y, Kim H. Temporal changes in mortality impacts of heat wave and cold spell in Korea and Japan. Environ Int. 2018;116:136–46. pmid:29679776
- 16. Wu S, Luo M, Lau GN-C, Zhang W, Wang L, Liu Z, et al. Rapid flips between warm and cold extremes in a warming world. Nat Commun. 2025;16(1):3543. pmid:40263258
- 17. Santer BD, Wigley TML, Mears C, Wentz FJ, Klein SA, Seidel DJ, et al. Amplification of surface temperature trends and variability in the tropical atmosphere. Science. 2005;309(5740):1551–6. pmid:16099951
- 18. Cohen J, Agel L, Barlow M, Entekhabi D. No detectable trend in mid-latitude cold extremes during the recent period of Arctic amplification. Commun Earth Environ. 2023;4(1).
- 19. Hong Y, Wang S-YS, Son S-W, Jeong J-H, Kim S-W, Kim B, et al. Arctic-associated increased fluctuations of midlatitude winter temperature in the 1.5° and 2.0° warmer world. npj Clim Atmos Sci. 2023;6(1).
- 20. Woo S, Kim B, Jeong J, Kim S, Lim G. Decadal changes in surface air temperature variability and cold surge characteristics over northeast Asia and their relation with the Arctic Oscillation for the past three decades (1979–2011). J Geophys Res. 2012;117(D18).
- 21. Wijayanto T, Wakabayashi H, Lee J-Y, Hashiguchi N, Saat M, Tochihara Y. Comparison of thermoregulatory responses to heat between Malaysian and Japanese males during leg immersion. Int J Biometeorol. 2011;55(4):491–500. pmid:20824480
- 22.
Khatun A. Adaptation to cold of tropical people by local cold tolerance and linguistic approaches. Nagasaki University. 2017.
- 23. Conlon KC, Rajkovich NB, White-Newsome JL, Larsen L, O’Neill MS. Preventing cold-related morbidity and mortality in a changing climate. Maturitas. 2011;69(3):197–202. pmid:21592693
- 24. Son J-Y, Gouveia N, Bravo MA, de Freitas CU, Bell ML. The impact of temperature on mortality in a subtropical city: effects of cold, heat, and heat waves in São Paulo, Brazil. Int J Biometeorol. 2016;60(1):113–21. pmid:25972308
- 25. Yang J, Ou C-Q, Ding Y, Zhou Y-X, Chen P-Y. Daily temperature and mortality: a study of distributed lag non-linear effect and effect modification in Guangzhou. Environ Health. 2012;11:63. pmid:22974173
- 26. Guo Y, Gasparrini A, Armstrong BG, Tawatsupa B, Tobias A, Lavigne E, et al. Temperature Variability and Mortality: A Multi-Country Study. Environ Health Perspect. 2016;124(10):1554–9. pmid:27258598
- 27. Woo J. Policy implications of population ageing in Hong Kong in the UN decade of healthy aging. American Journal of Medicine and Public Health. 2023;4.
- 28. Guo J, Xia D, Zhang L, Zou Y, Yang X, Xie W, et al. Future indoor overheating risk for urban village housing in subtropical region of China under long-term changing climate. Building and Environment. 2023;246:110978.
- 29. Xia D, Xie W, Guo J, Zou Y, Wu Z, Fan Y. Building Thermal and Energy Performance of Subtropical Terraced Houses under Future Climate Uncertainty. Sustainability. 2023;15(16):12464.
- 30. Li Z, Jin Y, Liang X, Zeng J. Thermography evaluation of defect characteristics of building envelopes in urban villages in Guangzhou, China. Case Studies in Construction Materials. 2022;17:e01373.
- 31. Bai HY, Liu P, Justo Alonso M, Mathisen HM. A review of heat recovery technologies and their frost control for residential building ventilation in cold climate regions. Renewable and Sustainable Energy Reviews. 2022;162:112417.
- 32. Gasparrini A, Guo Y, Hashizume M, Lavigne E, Zanobetti A, Schwartz J, et al. Mortality risk attributable to high and low ambient temperature: a multicountry observational study. Lancet. 2015;386(9991):369–75. pmid:26003380
- 33. Gasparrini A. Distributed Lag Linear and Non-Linear Models in R: The Package dlnm. J Stat Softw. 2011;43(8):1–20. pmid:22003319
- 34. Peng RD, Dominici F, Louis TA. Model Choice in Time Series Studies of Air Pollution and Mortality. Journal of the Royal Statistical Society Series A: Statistics in Society. 2006;169(2):179–203.
- 35. Bao J, Wang Z, Yu C, Li X. The influence of temperature on mortality and its Lag effect: a study in four Chinese cities with different latitudes. BMC Public Health. 2016;16:375. pmid:27146378
- 36. Kekkou F, Economou T, Lazoglou G, Anagnostopoulou C. Temperature extremes and human health in Cyprus: Investigating the impact of heat and cold waves. Environ Int. 2025;199:109451. pmid:40286556
- 37. Pithan F, Svensson G, Caballero R, Chechin D, Cronin TW, Ekman AML, et al. Role of air-mass transformations in exchange between the Arctic and mid-latitudes. Nature Geosci. 2018;11(11):805–12.
- 38. Li J, Sun R, Chen L. A review of thermal perception and adaptation strategies across global climate zones. Urban Climate. 2023;49:101559.
- 39. Hwang RL, Weng YT, Huang KT. Considering transient UTCI and thermal discomfort footprint simultaneously to develop dynamic thermal comfort models for pedestrians in a hot-and-humid climate. Building and Environment. 2022;222:109410.
- 40. Lin Y-K, Ho T-J, Wang Y-C. Mortality risk associated with temperature and prolonged temperature extremes in elderly populations in Taiwan. Environ Res. 2011;111(8):1156–63. pmid:21767832
- 41. Ceccherini G, Russo S, Ameztoy I, Romero CP, Carmona-Moreno C. Magnitude and frequency of heat and cold waves in recent decades: the case of South America. Nat Hazards Earth Syst Sci. 2016;16(3):821–31.
- 42. Qiu H, Sun S, Tang R, Chan K-P, Tian L. Pneumonia Hospitalization Risk in the Elderly Attributable to Cold and Hot Temperatures in Hong Kong, China. Am J Epidemiol. 2016;184(8):555–69. pmid:27744405
- 43. Guo Y, Barnett AG, Yu W, Pan X, Ye X, Huang C, et al. A large change in temperature between neighbouring days increases the risk of mortality. PLoS One. 2011;6(2):e16511. pmid:21311772
- 44. Zhan Z, Zhao Y, Pang S, Zhong X, Wu C, Ding Z. Temperature change between neighboring days and mortality in United States: A nationwide study. Sci Total Environ. 2017;584–585:1152–61. pmid:28162760
- 45. Tang J, Xiao C-C, Li Y-R, Zhang J-Q, Zhai H-Y, Geng X-Y, et al. Effects of diurnal temperature range on mortality in Hefei city, China. Int J Biometeorol. 2018;62(5):851–60. pmid:29224119
- 46. Zha Q, Chai G, Zhang Z-G, Sha Y, Su Y, Wu T. Impact of temperature changes between neighboring days on cardiovascular disease hospital admissions among suburban farmers in Qingyang, Northwest China. Int J Biometeorol. 2022;66(6):1233–45. pmid:35583607
- 47. Zanobetti A, O’Neill MS, Gronlund CJ, Schwartz JD. Summer temperature variability and long-term survival among elderly people with chronic disease. Proc Natl Acad Sci U S A. 2012;109(17):6608–13. pmid:22493259
- 48. Zhang S, Zhu N. Gender differences in thermal responses to temperature ramps in moderate environments. J Therm Biol. 2022;103:103158. pmid:35027194
- 49. Van Someren EJW. Thermoregulation and aging. Am J Physiol Regul Integr Comp Physiol. 2007;292(1):R99–102. pmid:16902181
- 50. Florez-Duquet M, McDonald RB. Cold-induced thermoregulation and biological aging. Physiol Rev. 1998;78(2):339–58. pmid:9562032
- 51. Knight J, Nigam Y. Exploring the anatomy and physiology of ageing. Part 1--The cardiovascular system. Nurs Times. 2008;104(31):26–7. pmid:18727348
- 52. Jung D, Kim H, An J, Hong T. Thermoregulatory responses of young and elderly adults under temperature ramps. Building and Environment. 2023;244:110760.
- 53. Sun S, Tian L, Qiu H, Chan K-P, Tsang H, Tang R, et al. The influence of pre-existing health conditions on short-term mortality risks of temperature: Evidence from a prospective Chinese elderly cohort in Hong Kong. Environ Res. 2016;148:7–14. pmid:26994463
- 54.
Administration CM. Grade of cold wave. 2017.
- 55. Burkart K, Meier F, Schneider A, Breitner S, Canário P, Alcoforado MJ, et al. Modification of Heat-Related Mortality in an Elderly Urban Population by Vegetation (Urban Green) and Proximity to Water (Urban Blue): Evidence from Lisbon, Portugal. Environ Health Perspect. 2016;124(7):927–34. pmid:26566198
- 56. Ho JY, Shi Y, Lau KKL, Ng EYY, Ren C, Goggins WB. Urban heat island effect-related mortality under extreme heat and non-extreme heat scenarios: A 2010-2019 case study in Hong Kong. Sci Total Environ. 2023;858(Pt 1):159791. pmid:36328261
- 57. Hou Y, Cao B, Zhu Y, Zhang H, Yang L, Duanmu L, et al. Temporal and spatial heterogeneity of indoor and outdoor temperatures and their relationship with thermal sensation from a global perspective. Environ Int. 2023;179:108174. pmid:37660634
- 58. Xia X, Chan KH, Niu Y, Liu C, Guo Y, Ho K-F, et al. Modelling personal temperature exposure using household and outdoor temperature and questionnaire data: Implications for epidemiological studies. Environ Int. 2024;192:109060. pmid:39401479