Short-term exposure to ambient temperature and the mortality burden of suicide in Japan

Background. Previous studies suggest that temperature may trigger suicide, with higher ambient temperatures associated with increased risk globally. However, quantification of the attributable suicide burden due to non-optimal temperatures, particularly across specific subgroups like sex, age, and method of suicide, is limited. This study aimed to quantify this attributable burden across Japan and assess variations in these subgroups to enhance understanding and identify high-risk populations. Methods. Data on daily suicide counts and mean temperatures were collected for all 47 prefectures in Japan (1973–2015) for a comprehensive, nationally representative analysis. Using time-stratified case-crossover analyses and a distributed lag non-linear model (DLNM), we estimated temperature-suicide associations, accounting for long-term trends, seasonality, and day-of-the-week effects, and capturing both nonlinear and delayed impacts of temperature. Overall attributable fractions and fractions attributable to cold (0th-10th percentile), cool (10th-50th), warm (50th-90th), and hot (90th-100th) temperatures were computed. Subgroup analyses were conducted by sex, age (0 – 64 years and ≥65 years), and method of suicide (violent or nonviolent). Results. Approximately 19.9% of suicides (95% empirical confidence interval (eCI): 18.8, 20.9) could be attributed to non-optimal temperatures, representing the proportion of cases that would not have occurred under optimal temperature conditions, identified at the temperature corresponding to the minimum risk of suicide. The highest fraction (9.9%, 95% eCI: 9.4, 10.4) was observed for warm temperatures (50th-90th percentile). Higher burdens were observed in females (23.7%), individuals aged ≥65 years (31.9%), and violent suicides (22.4%). Conclusions. Nearly 20% of suicides in Japan are associated with non-optimal temperatures, particularly during warmer conditions. These findings highlight the importance of integrating temperature considerations into suicide prevention strategies, particularly for vulnerable populations such as the elderly. While focused on Japan, these results advocate future research to explore similar burdens globally, enhancing our understanding of environmental impacts on suicide and informing targeted public health interventions.


Introduction
Suicide is a major public health concern globally.There are an estimated 703,000 global suicide deaths every year, and suicide is one of the leading causes of death worldwide, responsible for more deaths than HIV/AIDS, malaria, or homicide [1].Suicide was the fourth leading cause of death worldwide in 2019 among people aged 15-29 years for both sexes [1].
Globally, during the period 1990-2015, suicide rates showed downward trends in most countries [2].Despite showing a downward trend, some Eastern European countries and Japan remain among the countries with the highest levels of suicide deaths [2].
In 2021, there were a total of 21,007 suicides in Japan, with 13,933 suicides by men and 7,068 suicides by women [3].The suicide rate was 16.7 suicides per 100,000 people in 2021 (22.8 for men and 11.0 for women) [3].In 2019, among individuals aged 10 to 29 in Japan, the incidence of suicide emerged as the leading cause of death, with rates of 5.9 per 100,000 for those aged 10-19 and 17.2 per 100,000 for the 20-29 age bracket [3].Globally, Japan had the sixth-highest suicide rate and the second-highest rate of female suicide in 2019, although the rate for males was higher than that for females [3].
While the factors contributing to suicide are complex and multifaceted [4][5][6], emerging research has begun to explore the potential influence of environmental variables, such as ambient temperature, on suicide risk.Previous studies have suggested that ambient temperature may be associated with suicide occurrence [7].A study in the US found a significant association between temperatures above the median temperature, regardless of season, and an increased risk of suicide [8].Similarly, in California, increased suicide risk was associated with higher daily ambient temperatures [9].Additionally, a longitudinal study conducted across the United States and Mexico found that local ambient temperatures were robustly associated with suicide rates, with a generally linear association [10].For instance, between 1968 and 2000 in the United States, it was found that a rise of +1 °C in the average monthly temperature was associated with a 0.7% increase in the monthly suicide rate [10].Studies conducted across various European countries, including Belgium, Switzerland, France, and Spain, have consistently reported associations between increased ambient temperature and suicide mortality [11][12][13][14].Furthermore, a meta-analysis found the overall relative risk for suicide deaths per 1 °C rise in temperature was 1.01 (95% CI: 1.00, 1.02) [15].A multi-country study of 341 locations in 12 countries found that higher ambient temperature was associated with a higher risk of suicide, with a nonlinear association; the relative risk for the maximum suicide temperature versus the minimum suicide temperature was 1.33 (95% CI: 1.30, 1.36) [16].The nonlinear association was observed most clearly in Japan, South Korea, and Taiwan, while in Western countries such as Canada, Spain, Switzerland, the United Kingdom, and the United States, the associations were more linear [16].In Japan, another study identified a nonlinear, inverted J-shaped association between temperature and suicide rates.Notably, the study observed variations in this association based on gender, with females showing a more pronounced association than males.Additionally, a stronger association was found in individuals aged 65 years compared to younger age groups [17].However, the study did not evaluate whether the temperature-suicide association differs by the method of suicide, specifically violent and nonviolent suicide.
Although previous studies have quantified the association between ambient temperature and suicide in terms of relative risk [7], very few have quantified estimates of attributable burden, such as attributable number or attributable fractions.A study quantified the attributable fraction of emergency ambulance dispatches for suicide due to temperature in Shenzhen, China, with a total attributable fraction of 13.8% [18].However, to our knowledge, no study has assessed the fraction of suicide attributable to temperature in Japan or for different methods of suicide.Attributable burden or attributable risk measures offer better information on the actual impact of temperature than ratio measures such as relative risk, which are more helpful for policymakers [19,20].There are various suicide prevention measures in Japan after the national Basic Act of Suicide Prevention was established in 2006 [21].For instance, March, an early spring month with typically the highest number of suicide cases, is designated as 'Suicide Awareness Month' by the Ministry of Health, Labour and Welfare to promote awareness about suicide through messages, posters, videos, and online advertisements [22].The government's priority measures include enhancing support for local-level practical initiatives, bolstering efforts to prevent youth suicide, and addressing suicides stemming from work-related issues [21].Despite the growing body of evidence on the association between ambient temperature and suicide, there remains a significant gap in the literature regarding the precise quantification of the burden of suicide attributable to nonoptimal temperatures, particularly in Japan.This gap extends to the investigation of how this burden varies by gender, age, and method of suicide, which are critical for tailoring public health interventions and policies.This information is particularly useful for assessing how much of the suicide burden could theoretically be mitigated if the identified risk factor were addressed.Such quantification could inform the development of health policies and interventions by highlighting the importance of environmental factors in suicide prevention efforts.This approach could potentially allow for the allocation of resources and the design of preventive strategies that take into account the specific contributions of ambient temperature to suicide, aiming to contribute to efforts to lower suicide rates in Japan.
Addressing the current gap, our study aims to contribute to the field by analyzing the association between daily mean temperature and suicide mortality in Japan, quantifying the attributable fraction of suicide due to non-optimal temperatures, and exploring the differential impact by gender, age, and method of suicide.This approach not only advances our understanding of the environmental determinants of suicide but also potentially supports the development of targeted, evidence-based prevention strategies that consider the nuanced effects of temperature on suicide risk.Therefore, quantifying the burden of suicide attributable to temperature is important for informing health policy and public health interventions.
The objectives of this study are as follows: first, to analyze the association between daily mean temperatures and suicide mortality in Japan; second, to quantify the fraction of suicides that can be attributed to non-optimal temperatures; and third, to explore how this attributable fraction varies across different demographics, specifically by gender and age, and by the method of suicide.

Data
We obtained daily time-series data for suicide counts and weather variables across Japan's 47 prefectures for the period from 1973 to 2015.We selected this specific time frame because the datasets we had were both complete and available for all 47 prefectures during this period.This 43-year period allows for a robust longitudinal analysis.

Weather data
The weather data, comprising daily mean temperature (°C), relative humidity (%), and daily total sunshine duration (hours), were collected from the Japan Meteorological Agency.The Japan Meteorological Agency records and provides hourly measurements of weather variables for weather stations across Japan.The daily weather data were computed by averaging hourly data captured by a single weather station in each capital city of the prefectures, except in Shiga and Saitama, where the station is near the capital city.The missing data amounted to 0.03% for temperature.

Suicide data
Daily time-series data for suicide counts were obtained from the Ministry of Health, Labour and Welfare of Japan, which maintains computerized death certificate data covering all prefectures in Japan.Suicide was defined as intentional self-poisoning and self-harm using the eighth, ninth, and tenth revisions of the International Statistical Classification of Diseases and Related Health Problems [ICD-8 [23], ICD-9 [24], ICD-10 [25]]: codes E950.0-E958.9 for ICD-8 and −9 and codes X60-X84 for ICD-10.In the literature, whether suicide is violent or not is generally defined based on the method, in which poisoning and drug overdoses are considered to be nonviolent [26,27].Codes E953.0-E958.9 for ICD-8 and −9 and codes X70-X84 for ICD-10 were used to identify violent suicide such as hanging, jumping from a high place, and firearm discharge, while nonviolent suicide [ICD-8/9: E950-E952 and ICD-10: X60-X69] included self-poisoning and drug overdose, in line with a previous study [17].

Statistical analysis
Estimation of the association between temperature and suicide We first estimated the association between temperature and suicide by analyzing daily mean temperature series and suicide counts from the period 1973-2015.Our approach involved a two-stage analysis.
In the first stage, to estimate prefecture-specific temperature-suicide associations, we employed a timestratified case-crossover design, which was conducted using a conditional Poisson regression model to allow for overdispersion [28], following an approach detailed in a previous study [16].We defined a stratum through a three-way interaction of calendar year, month, and day-of-week.This enabled the matching of each suicide case with control days that fall on the same weekday within the same month and year, thereby facilitating exposure level comparisons within these strata.The case-crossover design inherently adjusted for potential confounders such as long-term time trends, seasonality, and day-of-the-week effects, under the assumption that any unmeasured time-varying confounders are constant within a stratum [29].Strata without suicide events were excluded from the analysis.
To model the temperature-suicide association, we utilized a distributed lag nonlinear model (DLNM).DLNM allowed for the simultaneous modeling of both the nonlinear exposure-response and the delayed effects of temperature [30].Specifically, our model employed a cross-basis function using a natural cubic B-spline basis with a degree of freedom (df) of 4 for both the temperature and the lag.The lag was extended up to three previous days.These modeling choices were guided by a previous study [17].
In the second stage, we pooled the prefecture-specific estimates through a multivariate meta-regression model [31].Details on the estimation of the exposure-response association and the multivariate meta-regression model are described in the following paper [17].This model included both prefecture-specific average temperatures and temperature ranges as meta-predictors, whose significance was evaluated using a multivariate Wald test.Additionally, we tested residual heterogeneity using a multivariate extension of the Cochran Q test and I 2 statistic.This stage of analysis yielded the best linear unbiased predictions for the overall cumulative exposure-response estimates for each prefecture, drawing on the fitted multivariate meta-regression model [31].
This method allows prefectures with fewer daily suicide counts to borrow information from larger populations with similar characteristics [31].

Quantification of the attributable burden of suicide due to temperature
This analysis aimed to quantify the number and fraction of suicide cases attributable to temperature.First, for each prefecture, we identified the minimum suicide temperature, defined as the temperature corresponding to the minimum relative risk of suicide between the first and the 99th percentiles of the temperature dataset.This was achieved using the best linear unbiased prediction of the overall cumulative temperature-suicide association in each prefecture.
Employing this minimum suicide temperature as a reference, we then calculated the temperatureattributable number and attributable fraction of suicide.This was accomplished by re-centering the natural cubic spline that models the exposure-response association, using an approach described in detail in a previous study [20].Briefly, in each location, the overall cumulative relative risk corresponding to each day's temperature was used to calculate the attributable number and attributable fraction for the subsequent 3 days, using the 'attrdl' function in R [20].The attributable fraction AF x t , at a day t in the series was calculated by the following equation [32]: where l and L represent the minimum and maximum lag, respectively, and b å x l t, denotes the overall cumulative risk for temperature x t for day t, obtained by the best linear unbiased prediction of the overall cumulative exposure-response association.
We summed the contributions across all days of the series to ascertain the total attributable number of suicides due to non-optimal temperatures, namely, temperatures greater than or less than the minimum suicide temperature.The total attributable fraction was subsequently calculated as the ratio of this total attributable number and the total number of suicide deaths.Furthermore, we segmented the attributable fraction by temperature category-cold, cool, warm, and, based on cutoffs at the 10th, 50th, and 90th percentiles of prefecture-specific temperatures.The specific temperature range definitions are as follows: cold (0th-10th percentile), cool (10th-50th), warm (50th-90th), and hot (90th-100th) temperatures.These cutoffs are consistent with a previous study [33].The 95% empirical confidence intervals (eCIs) were calculated using 1000 Monte Carlo simulations, assuming a multivariate normal distribution for the best linear unbiased predictions of the reduced coefficients [20].

Subgroup analysis
Subgroup analyses were conducted for age (0-64 years and 65 years of age), sex, and violent and nonviolent suicide.

Sensitivity analysis
Sensitivity analysis was performed by controlling for the averages of relative humidity and sunshine duration over the current day to the day before, using a natural cubic B-spline with three degrees of freedom.

Software and statistical packages
All analyses were conducted with R version 4.0.3(R Foundation for Statistical Computing, Vienna, Austria), with packages 'gnm' and 'dlnm' for the time-stratified case-crossover analysis and 'mvmeta' for metaregression.

Results
Table 1 shows the descriptive statistics for the number of suicides based on sex and age and the weather variables.A total of 1,049,592 suicides were recorded in Japan, of which 67.0% were by males and 73.7% were by individuals aged 0-64 years.Most suicides (84.9%) were violent.
The cumulative temperature-suicide association curves for the overall population and subgroups are shown in figure 1.
The exposure-response curves are generally nonlinear with an inverted J shape, with the risk of suicide increasing as the temperature increases but plateauing for temperatures above a threshold, namely, the maximum suicide temperature, which was identified as 23.1 °C (the 83rd percentile of the temperature distribution) in case of overall suicide.These inverted J-shaped nonlinear associations, in which the relative risk reaches a maximum suicide temperature before plateauing after a threshold, apply to overall suicide and subgroups, except for nonviolent suicide.In the case of nonviolent suicide, the curve leveled off at relatively lower temperatures, resembling more of an inverted U shape.The relative risk curve was higher for females than for males and was considerably higher for the group aged 65 years than for the group aged 0-64 years.For females, the relative risk increased until it reached the maximum suicide temperature of around 29 °C, then  plateaued, while the maximum suicide temperature for males was lower, around 24 °C.For the group aged 65 years, there was a steeper nonlinear increase in relative risk, with the curve plateauing at higher temperatures after reaching a maximum suicide temperature of around 29 °C, compared to the group aged 0-64 years, for whom the maximum suicide temperature was around 23 °C.The relative risk curve was higher for violent suicide than for nonviolent suicide.For violent suicides, the relative risk continued to increase until reaching the maximum suicide temperature of around 27 °C, then plateaued.In contrast, for nonviolent suicide, the relative risk reached its maximum at around 4 °C, then plateaued before decreasing at higher temperatures.The relative risks for overall and each subgroup are shown in table S1.
The main results of this study were the attributable fractions of suicide due to temperature, which are summarized in table 2.
The total attributable fraction of suicide due to temperature (both heat and cold) was 19.9% (95% eCI: 18.8, 20.9).This attributable fraction corresponds to an attributable number of 208,437 (95% eCI: 197096, 219291) excess suicide deaths due to temperature (table 2).The attributable burden was also calculated as separate components: cold, cool, warm, and hot temperatures (figure 2).
Results from the multivariate meta-regression analysis indicated that residual heterogeneity (representing the variability among prefecture results not explained by the model) is low, as evidenced by a nonsignificant Cochran Q test (table S12).Specifically, we observed a decrease in residual heterogeneity, with the I 2 value dropping from 11.0% in the intercept-only model to 4.6% when we included average temperature and temperature range as meta-predictors.This reduction in I 2 and the nonsignificant result of the Cochran Q test suggest that the variance among the outcomes for different prefectures is minimal and is well-accounted for by our model parameters.In practical terms, the drop in I 2 signifies that the inclusion of average temperature and temperature range as factors in our model substantially aids in explaining the variations in suicide observed across prefectures.Essentially, this indicates a stronger and more consistent relationship between ambient temperature and suicide, taking into account these specific variables, which improves the overall interpretability and fit of our statistical model.The sensitivity analysis corroborated the robustness of our findings, showing that the pooled temperature-suicide association (figure S2) and attributable fractions due to temperature (table S13) remained similar after adjusting for relative humidity or sunshine duration, and the findings were robust.

Discussion
Principal findings The current study was a nationwide analysis to quantify the attributable burden of suicide due to temperature across all prefectures in Japan, assessing variations by sex, age group, and method of suicide.We identified an inverted J-shaped nonlinear association between temperature and overall suicides, as well as for subgroups, except for nonviolent suicides.A total of 19.9% of suicides during the study period  could be attributable to non-optimal temperatures, with a larger burden attributed to warm temperatures (50th-90th percentile) than to extreme heat or cold.Notably, the burden was more pronounced in females, individuals aged 65 years, and violent suicide.

Comparison with other studies
The comparability to previous studies quantifying the attributable burden of suicide due to temperature is restricted by the scarcity of such studies.A study conducted in Shenzhen, China, reported a total attributable fraction of 13.8% (95% eCI: −6.6, 28.2) for emergency ambulance dispatches (EADs) for suicide [18].The study found a higher fraction attributable to hot temperatures, 13.8% (95% eCI: −6.1, 28.1), compared to 0.1% (95% eCI: −1.2, 1.3) for cold [18].However, the estimated attributable fractions were largely uncertain, unlike our findings [18].Although it is difficult to directly compare their findings with the results from the current study due to differences in the classification of cold and heat, and because their study investigated morbidity rather than mortality, our study also found that a low fraction of suicides was attributable to cold temperatures.
Another study in Jiangsu Province, China, estimated the burden of cause-specific mortality attributable to temperature but reported only the burden for external causes as a whole (15.3%) despite including suicide data and commented that they did not observe significant temperature-suicide association, likely due to low daily suicide counts in the study areas [34].
Unlike the current study, a Global Burden of Disease (GBD) study found a negative population attributable fraction (PAF) due to cold temperature of −4.3 (95% Uncertainty Interval (UI): −5.4,−3.3) for suicide in 2019, demonstrating the protective effects of cold [35].The GBD study estimated the burden of suicide due to nonoptimal temperatures in 2019, based on data from nine countries.The disparity between the results of the GBD study and those of our current research may be attributed to the methodology used in the GBD study.Specifically, the GBD study computed the PAFs for both high and low temperatures, relative to a reference value representing the theoretical minimum risk exposure level.It is important to note that this reference value was established based on the temperature associated with the lowest mortality across all included causes, encompassing non-external causes of cardiorespiratory and metabolic mortality, not solely limited to suicide mortality.The theoretical minimum risk exposure levels ranged from approximately 18 °C to 27.5 °C for study locations in mean annual temperature zones of 6 °C to 28 °C.
Our study expands on existing knowledge by indicating that the attributable burden of suicide is higher among older people, females, and violent suicides.The attributable fraction for older individuals (65 years) was twice that for younger people (0-64 years).The attributable fraction due to warm temperatures (50th-90th percentiles) was the highest, not extreme hot temperatures or cold.This phenomenon could potentially be due to the observed nonlinear J-shaped temperature-suicide association, where higher effects are seen at warmer temperatures compared to cooler ones until reaching a certain threshold, beyond which the effect plateaus at more extreme temperatures.This is in line with previous research in Japan which showed an increased relative risk with rising temperatures but leveled off above a threshold temperature, namely the maximum suicide temperature [16,17].Additionally, the larger attributable fractions for warm temperatures over hot temperatures or extreme may be explained by warm temperatures accounting for a greater number of days.Warm temperatures were responsible for suicides across different sex and age groups; however, for nonviolent suicides, cool temperatures (10th-50th percentiles) were associated with the highest fraction.This might be due to uncertainties in estimates in regions with fewer nonviolent suicides or differences in exposure-response curve, suggesting distinct mechanisms for nonviolent and violent suicides.
The findings of the current study show a nonlinear temperature-suicide association in Japan, with the risk leveling off at extremely high temperatures.This is consistent with a broader body of existing studies investigating the association between ambient temperature and suicide in terms of relative risk in Japan [16,17,36].Our findings are also consistent with a previous study in Tokyo, Japan, which indicated that an increase in daily mean temperature was nonlinearly associated with an increased risk of drug overdose, the most common method of unsuccessful suicide attempts in Japan [37].Our findings also align with previous studies that showed higher relative risks for older people and women in Japan [16,17].Similarly, the previous study investigating the association between daily mean temperature and drug overdose using emergency ambulance dispatch data in Tokyo also found higher relative risks among older people and women [37].However, the results from a multi-country study showed that the observed differences by sex and age group, with nonoverlapping confidence intervals, occur only in Japan and South Korea [16].Specifically, Japan shows differences for both sex and age groups, while South Korea shows them only for age groups [16].One partial explanation for the higher risk among the elderly in Japan may be the lower use of air conditioning even in hot weather [38].In contrast, other studies reported a higher relative risk for suicide among younger people in countries such as Switzerland and China [14,39].
In agreement with the findings of the current study, some studies conducted in other countries have found a similar larger risk for females [14,40], while other studies have found a higher risk for males in countries such as the UK, Spain, Mexico, and France [10,11,16].This observed discrepancy in vulnerability across sex groups could be attributed to culture-specific factors [17].For example, lower status and social welfare of women compared to men in Japan may heighten the risk of suicide events in this subgroup in response to changes in environmental conditions, such as temperature fluctuations [17].
Furthermore, the current study found a statistically significant attributable burden and association between temperature and both violent and nonviolent suicide, consistent with some previous studies [14,41], while other studies found no association between temperature and nonviolent suicide [42][43][44][45].However, unlike the present study, a previous study in Switzerland found a larger relative risk of 1.7 (95% CI: 1.2, 2.5) for nonviolent suicide than violent suicide of 1.3 (95% CI:1.2, 1.5), despite a higher proportion of violent suicide (90.2%) than nonviolent suicide (9.8%) [14].The reasons for the discrepancy are not yet clear, as few studies have estimated the relative risk for different methods of suicide.These seemingly contrasting findings in different locations illustrate the complexity of the temperature-suicide association, in which the underlying mechanism remains unclear.

Potential mechanisms
Although there is no scientific consensus on the exact mechanism of the temperature-suicide association yet, there is a hypothesis that people with pre-existing mental illness may be more vulnerable to the mental health impacts of high temperatures [46].For example, heat can intensify mental health issues in individuals with preexisting conditions, potentially leading to an increased risk of suicide [47].One proposed pathway is that high temperatures may trigger changes in neural activity in the brain, potentially worsening pre-existing mental health conditions, such as anxiety, and possibly predisposing individuals with risk factors to suicide [48,49].In addition, sleep disturbances, known to escalate with rising temperatures, represent another plausible pathway through which ambient heat may negatively impact mental health outcomes [50,51].Another commonly proposed hypothesis involves changes in serotonergic innervation [16].Some evidence suggests that a deficiency in serotonin or 5-hydroxytryptamine (5-HT) is associated with completed suicide and suicidal behaviors [50,52].The serotonergic system may be linked with impulsive-aggressive behavior, a risk factor for suicide [53].Studies have suggested that ambient temperature may influence changes in the serotonergic system, with a negative correlation between biomarkers of serotonin and temperature [54,55], which may theoretically lead to impulsive aggression and suicide.It should be noted that this serotonin hypothesis was initially proposed as a potential explanation for the seasonality of suicide as the serotonergic system appears to have a circannual rhythm [56].However, whether acute increases in temperature can immediately affect serotonin levels in humans remains largely unstudied.Similarly, dose-response relationships concerning fluctuations of daily mean temperature, serotonin, and suicide remain unexplored in the existing literature.In contrast, nonviolent suicide tended to be associated with substance use disorders [27], suggesting there may be additional mechanisms for the association between temperature and nonviolent suicide.For example, some researchers have hypothesized that the accessibility of methods may be relevant [57].Poisoning by carbon monoxide, used in nonviolent suicide attempts, may be more successful in enclosed, poorly ventilated spaces-a scenario that is more frequent in winter due to heating systems and reduced ventilation.However, it is important to note that nonviolent suicide methods such as carbon monoxide from charcoal burning, exposure to gas exhaust, and selfpoisoning by drugs are available throughout the year.Moreover, although the reasons behind the observed differences in the shapes of the exposure-response curves for violent (inverted J-shaped) and nonviolent (inverted U-shaped) suicides found in this study are not definitively known, the difference in the curves suggests potentially distinct physiological or psychosocial mechanisms underlying these two types of suicides in response to temperature variations.For instance, higher temperatures may exacerbate mood instability or impulsivity, which are often associated with violent methods of suicide [54,55].Conversely, nonviolent methods, which may include planned overdoses, might not follow the same temperature-related fluctuations in mood or behavior.Furthermore, the differences might also reflect socio-demographic and psychological factors that distinguish individuals who choose violent methods from those who opt for nonviolent means.For example, access to means, cultural factors, and individual mental health conditions (such as the presence of substance abuse disorders, which have been associated with nonviolent suicide methods) might interact differently with environmental stressors like temperature.To summarize, potential underlying mechanisms for the short-term temperature-suicide association remain largely theoretical, and further investigations may be needed.Moreover, it remains unclear whether the mechanisms for the association between temperature and suicide differ by sex, age, and method of suicide.A deeper understanding of the underlying mechanisms could strengthen the case for public health initiatives.

Strengths and limitations
The current study has several strengths.It is the first study to nationally quantify the attributable burden of suicide due to non-optimal temperature in Japan.Covering multiple prefectures, it examines over 1 million suicides across the country.Furthermore, it pioneers the quantification of temperature-related suicide burden based on the method of suicide.Some limitations should be acknowledged.This study's focus was on the attributable burden of suicide due to temperature in Japan.Japan has a nonlinear temperature-suicide association similar to Taiwan and South Korea, whereas other countries may have a more linear association [16].Therefore, the findings should not be interpreted as globally representative.Future research could explore the attributable burden of suicide due to temperature in a diverse range of geographical settings to discern whether a universal pattern exists.This approach could help in developing a more nuanced understanding of the global impact of temperature on suicide burden.Further multi-country studies should provide a more global estimation of the attributable burden of suicide, especially in less studied regions such as low-and middle-income countries.In addition, the ambient temperature data collected from a single monitoring station were used to represent the exposure for the entire prefecture, similar to previous studies [16,17], potentially causing Berkson-type measurement errors and leading to uncertainty in the estimates [58].Future studies should aim to incorporate data from multiple monitoring stations within each prefecture or employ satellite-derived temperature data, which can provide more accurate and spatially comprehensive coverage.This approach could reduce the measurement error.

Implications
The findings of the present study suggest that mental health and suicide prevention services in Japan could be strengthened during periods of warm weather, for example, 50th to 90th percentile of prefecture-specific mean temperature, as the results suggest that warm temperatures contribute to a higher burden of suicide attributable to temperature than hot or cold temperatures.Heat warning systems and public health campaigns could be designed to target vulnerable populations, particularly the elderly, given that the findings suggest they have a higher burden of temperature-attributable suicide.The findings are consistent with previous literature that the elderly have the highest heat-related vulnerability [59].In addition, climate change adaptation plans may need to consider the mental health impacts of temperature.As global temperatures continue to rise, the impact of temperature on mental health and suicide may become even more pronounced [10].Therefore, climate change adaptation plans may benefit from incorporating strategies to address the potential mental health impacts of temperature, particularly for vulnerable populations.It is important to note, however, that these recommendations are preliminary and should be pursued in conjunction with ongoing research to fully understand the complex relationship between ambient temperature and suicide burden.Raising public awareness about the potential mental health impacts of temperature changes may also be beneficial.

Conclusions
In conclusion, our study found that almost 20% of suicides in Japan are attributable to non-optimal temperatures, and a greater burden was due to the warm temperatures (9.9%), rather than extreme heat or cold.Some groups, such as females (23.7%) and people aged 65 years and over (31.9%), had a higher burden.This information could assist public health officials and medical professionals in planning targeted interventions and policies, considering the rapidly changing climate and aging population.
Further research is needed to understand the underlying mechanisms linking ambient temperature and suicide, to inform specific policy recommendations, and to identify effective interventions to mitigate the burden.

Figure 1 .
Figure 1.Overall cumulative exposure-response relationships for overall suicide and subgroups for Japan.Nationwide overall cumulative temperature-suicide associations as the best linear unbiased prediction (BLUP) (with 95% CIs, shaded grey) for the maximum suicide temperature versus the minimum suicide temperature (−2.3 °C, which corresponded to the 1st percentile of observed temperature distribution) for (a) overall suicides, (b) males and females, (c) 0-64 years population and 65 years population, and (d) violent and nonviolent suicides.The vertical dashed line indicates the maximum suicide temperature.

Table 1 .
Descriptive statistics of daily suicide and meteorological variables in Japan from 1973 to 2015.
N = 1,049,592, representing the total number of suicides recorded in Japan during the study period.

Table 2 .
Attributable fraction and attributable number of suicides due to temperature in Japan during the study period.