Effects of Accumulated And Transient Heat Exposure On Schizophrenia Hospitalizations: A Time-Series Analysis On Hourly Temperature Basis

Chao Tang Anhui Medical University School of Public Health Yifu Ji Anhui Mental Health Center: Fourth People's Hospital of Hefei Qingru Li Anhui Medical University School of Public Health Zhenhai Yao Anhui Public Meteorological service center, Hefei, Anhui Jian Cheng Anhui Medical University School of Public Health Yangyang He Anhui Medical University School of Public Health Xiangguo Liu Anhui Medical University School of Public Health Rubing Pan Anhui Medical University School of Public Health Qiannan Wei Anhui Medical University School of Public Health Weizhuo Yi Anhui Medical University School of Public Health Hong Su (  271244914@qq.com ) Anhui Medical University

found a positive correlation between temperature and schizophrenia in 1992 (Gupta et al. 1992). Thereafter, some studies revealed that monthly or daily high temperature can increase hospitalizations for schizophrenia (Clarke et al. 1999 Assessing the effect of heat on disease with daily temperature exposure data is admittedly important for disease warning (Cheng et al. 2017). However, daily mean temperature cannot consider hourly temperature change, which is also a potential risk factor for health (Hu et al. 2019). Some heat-related assessments usually focus on days when the daily mean temperature is above the temperature threshold, which may not examine the practical effects of heat on disease, as hourly extreme high temperatures of these days can also have an adverse effect (Jiao et al. 2019). Thus, many studies suggested that the usage of hourly temperature data to assess the risk of morbidity and mortality is necessary (Lin et    ). Additionally, this 3A ( rst-class) hospital has an advanced electronic data storage system that records all admitted patients. The information of patients mainly includes the date of admissions, residence address, age, gender, date of birth, etc. Only schizophrenic inpatients living in Hefei were included in this study. The diagnosis of schizophrenia was con rmed based on the 10th version of International Classi cation of Diseases (ICD-10), and the code was F20.0-F20.9.
Finally, a total of 53,288 cases were collected from 2005 to 2019.

De nition and calculation of hourly excess heat exposure within a day
We used a novel index-DEHH (daily excess hourly heat) in this study. DEHH is a cumulative temperature variable generated by adding the temperature values that exceed the speci c heat threshold per hour within a day ( where i referred to the hour time of observation; t i was the temperature at the observation time; t h was the heat threshold; △t i was the difference between t i and t h . The heat threshold is the temperature above which hospitalization risk begins to increase, which is conceptually the same as the minimum risk temperature ( Log[E(y t )] = α + cb(DEHH t,l ,β) + ns(humidity,3) + ns(DTR,3) + ns(time,3)+ γDOW+ ŋHoliday (2) where t meant the calendar day of observation; E(y t ) was the predicted counts of schizophrenia admissions on day t; α was the intercept of the model; cb represented the "cross-basis matrix" in DLNM; β referred to the regression coe cient; l represented the lag days; ns meant natural cubic spline function in the model. The potential in uence of relative humidity and diurnal temperature range (DTR) were controlled with ns of 3 df. We used a ns with 3 df to control for long-term trend and seasonality. Day of the week (DOW) and public holidays (Holiday) were incorporated into the model as dummy variables. Based on the lowest Q-AIC values of model, the maximum lag of 7 days was selected to capture the effects of heat on hospitalization for schizophrenia. Previous studies also suggested that the health impacts of heat last for approximately a week (Xiao et al. 2015;Guo. 2016;Lin et al. 2019). In order to identify potentially vulnerable groups, we also conducted subgroup analysis by gender (male, female) and age (0-40 years old, > 40 years old) (Wei et al. 2020).

Sensitivity analyses
We tested the robustness of our ndings by the following analysis. First, we varied the degree of freedom of daily relative humidity (3-5 df) and daily DTR (3-5 df) in the model. Second, we adjusted for seasonality and long-term trend by changing the df for time (3-5 df). Third, we added the daily air pollutants (PM 2.5 , NO 2 , SO 2 , CO, O 3 ) into the model to examine the impact of pollutants on the main results. In addition, in the time-strati ed study, we also set the time interval as 4-y period and 5-y period to check the stability of our results.
We used the "dlnm", "ggplot2", "plyr" and "splines" packages in R software (version 3.6.3) for statistical analysis and making gures. When two-sided p value was less than 0.05, the statistical tests were deemed statistically signi cant. Measurements of the association were expressed as the relative risk (RR) estimates and 95% con dence interval (CI) for per interquartile range (IQR) increment of DEHH and 24 hourly temperature measurements on each day, respectively. Table 1 Table 2). Subgroup analyses by gender illustrated that the effect of DEHH on male schizophrenics was from lag 0 (RR = 1.059, 95% CI: 1.005, 1.116) to lag 3 (RR = 1.036, 95% CI: 1.006, 1.067) (Fig. 2b). However, we did not nd a signi cant association between DEHH and female schizophrenics (Fig. 2c). For cases over 40 years old, DEHH presented signi cant effect from lag 1 (RR = 1.049, 95% CI: 1.011, 1.087) to lag 3 (RR = 1.039, 95% CI: 1.011, 1.077) (Fig. 2d). For cases under 40 years old, we didn't observe a signi cant association (Fig. 2e). The exposure-response relationship of DEHH on total schizophrenic admissions and subgroups were shown in Supplementary Fig. 2. Table 3 showed the cumulative effects of DEHH on total schizophrenia hospitalizations and subgroups from lag 0-0 to lag 0-7. For total cases, the greatest RR of the cumulative effects was 1.189 (95% CI: 1.051, 1.279) at lag 0-5. Meanwhile, the cumulative effects of DEHH on schizophrenia hospitalizations was found only in males and cases over 40 years old in subgroup analyses.    Fig. 3 indicated that DEHH was more strongly associated with hospitalization for schizophrenia in the middle stage of the entire study period (data were shown in Supplementary Table  1). In total, the cumulative effects of DEHH represented a complex change, showing a tendency of early upward and later downward. Figure 4 presented the effects of temperature observation time on heat-schizophrenia relationships in warm season from lag 1 to lag 5. The RR estimates by each IQR increment were statistically signi cant from 0 am at lag 1 (RR = 1.034, 95% CI:

Results
1.013, 1.054) to 3 am at lag 3 (RR = 1.024, 95% CI: 1.010, 1.039). Overall effects of temperature observation time on heatschizophrenia were gradually decreased from lag 1 to lag 5, and almost no statistical signi cance was observed after lag 4 (the RR estimates were not shown in the gure at lag 6 and lag 7). In short, Fig. 4 showed the impacts of heat on hospitalization for schizophrenia varied with the time from lag 1 to lag 5, and the strongest association occurred at 5 am at lag 1 (RR = 1.045, 95% CI: 1.025, 1.066). Our results suggested that there may be a stronger association between heat and schizophrenia at dawn (0 am-6 am) (data were shown in Supplementary Table 2). In addition, in order to re ect the integrity of the effect of heat on the admissions for schizophrenia, we presented the RR estimates of lag 0 in Supplementary Fig. 3. However, the hourly temperature effect pattern on the day of onset (lag 0) should be carefully interpreted, as temperature "exposure" measurement on that day may occur after hospitalizing (Davies, et al. 2016).
Our ndings remain stable through varying the degrees of freedom for relative humidity (3-5 df), DTR (3-5 df), the seasonality and long-term trend (3-5 df) ( Supplementary Fig. 4). We found similar estimates of heat effects before and after controlling for the air pollutants (Supplementary Table 3). Furthermore, in the time-strati ed analysis, the overall trend was consistent after splitting the time period into an interval of 4 and 5 years ( Supplementary Fig. 5 and Fig. 6).

Discussion
On the basis of hourly temperature data, this study employed a novel heat exposure indicator, daily excess heat hourly (DEHH), to initially examine the relationship between accumulated heat exposure within a day and schizophrenia hospitalizations. The results showed that DEHH was associated with an increased admissions risk of schizophrenia from lag 1 to lag 4. Subgroups analyses showed that males and people over 40 years old were more susceptible to DEHH. Timestrati ed analysis observed complex uctuations in schizophrenia hospitalizations risk associated with DEHH, rather than a monotonous upward or downward trend. We also found the effects of transient heat exposure on schizophrenia hospitalizations at different time points within a day.
In the context of global warming, extreme high temperature events due to climate change are considered to be closely Consistent with previous studies that used daily mean temperature as exposure variable, we applied DEHH to identify the association of heat with schizophrenia admissions from a more precise perspective. In the analysis of different individual characters, we found that males were more susceptible to DEHH, which has also been recognized in previous investigations (Wang et al. 2018;Pan et al. 2019). Men are usually more involved in physical work than women, thereby they have more chances to be exposed to high temperatures (Xu et al. 2020 Some strengths are evident in this study. First, based on hourly temperature measurements, it is the rst study to apply DEHH to explore the effect of heat on schizophrenia hospitalizations. DEHH synthesizes hourly temperature exposure information to make exposure assessments more accurate. Second, time-strati ed design was also used for the rst time in the study of heat and schizophrenia, giving the public a better grasp on temporal trend of heat effect on schizophrenia.
Third, we analyzed the heat exposure at a xed time point on schizophrenia hospitalizations. Our ndings can be used to prompt local mental health institutions of the time to arrange emergency preparedness and provide adequate medical resources before the high risk of morbidity.
It is necessary to mention the limitations of this study. First, the present study is an ecological research and will inevitably cause ecological bias. Second, a single city study has limited the extrapolation of the conclusion, as climate characteristics and population acclimatization across different regions are heterogeneous (Cheng et al. 2019). Further research on the relationship between DEHH and schizophrenia in different study settings is needed. Finally, in the analysis of the in uence of time points on heat-schizophrenia relationships, it should be noted that only the temperature varies at the hourly level-hospitalization varies daily rather than hourly. We can only make a rough estimate of the speci c time point when hospitalization risk of schizophrenia is higher. Further analyses need be made to explore the in uence of hourly temperature on schizophrenia by obtaining more detailed hospitalizations data.

Conclusions
In conclusion, we applied a novel index-daily excess heat hourly (DEHH) to explore the in uence of heat on schizophrenia hospitalizations. The results suggested that DEHH increased the risk of schizophrenia hospitalizations. DEHH may be a surrogate indicator for the precise assessment of the relationship between heat and schizophrenia. Furthermore, by exploring the effect of exposure timing on heat-related schizophrenia hospitalizations, it may help health authorities and policymakers to better formulate relevant protection measurements to reduce the risk of schizophrenia. This study revealed the effect of accumulated and transient heat exposure within a day on schizophrenia from different perspectives. We recommend that hourly weather observations should be taken into account when assessing heat impact on schizophrenia hospitalizations.

Declarations
Declaration of competing interest