Impact of pesticide regulations on mortality from suicide by pesticide in China: an interrupted time series analysis

Background Pesticide bans and regulatory restrictions have been shown to be effective strategies for preventing suicide in several countries. Suicide and suicide by pesticides have decreased significantly in China over the past two decades. However, whether the reduction was associated with pesticide regulation is unknown. Methods The monthly data on suicide and suicide by pesticide from 2006 to 2018 were obtained from China's Disease Surveillance Point (DSP) system. Information on China's pesticide regulations since 1970 was obtained from Pesticide Action Network International (PAN International), Joint Meeting on Pesticide Management Highly Hazardous Pesticides (JMPM HHP) lists, the website of the Ministry of Agriculture of China, Pesticide Information Network of China, and the Wan Fang database. Change point detection and policy analysis were combined to identify the time of any trend change breakpoint of suicide and suicide by pesticide. Interrupted time series analysis was used to investigate the pre- and post-breakpoint trends of monthly standardized rates in suicide and suicide by pesticide. Results The standardized pesticide suicide rate decreased by 60.5% from 6.50 in 2006 to 2.56 per 100,000 in 2018. Larger declines were evident among people in urban areas (67.3%), female individuals (63.5%), and people aged 15–44 years (68.1%). The effect of policies banning highly hazardous organophosphorus pesticides (HHOP) [rate ratio (RR) = 0.993, 95% CIs (0.991–0.994)] in December 2008 and stopping domestic sales and use of paraquat aqueous solution (RR = 0.992, 95% CIs: 0.990–0.994) in July 2016 were more pronounced than regulating the paraquat-related products (RR = 1.003, 95% CIs: 1.002–1.004) in April 2012. Conclusion Declines in suicide by pesticide in China occurred contemporaneously with regulatory bans and restrictions implemented on several pesticides, particularly in urban areas, among female individuals, and the relatively low age profile. These findings indicate the potential influence of these bans on trends of suicide by pesticides.


. Introduction
Suicide is a serious and important global health problem. There were approximately 703,000 suicides worldwide in 2019 (an agestandardized rate of 9 per 100,000 people), accounting for 1.3% of all deaths and ranking as the 17th leading cause of death globally (1,2). Importantly, 79% of these suicides occurred in low-and middle-income countries (3). To effectively address this issue, multi-sectoral organizations and relevant stakeholders have been encouraged to take concerted action through comprehensive national strategies and measures (4)(5)(6). The World Health Organization's (WHO) Comprehensive Mental Health Action Plan 2013-2030 and Sustainable Development Goal 3 (target 3.4 and indicator 3.4.2) set out the global goal of reducing suicide rates by one-third by 2030 (3,7).
Despite declines in the rate of suicide in the past two decades, the number of suicides in China is still very large compared to the rest of the world (3,(8)(9)(10)(11)(12). Research suggests that a substantial proportion of suicidal acts in China are impulsive, following acute psychosocial stress or crisis (13). The leading method of suicide in China is pesticide poisoning, accounting for approximately half of all suicide deaths (9,14). It is widely recognized that one of the most effective approaches for preventing suicide is restricting the availability of commonly used, high-lethality methods (15). In countries where pesticides account for a high proportion of suicides, such as China, the WHO, and the Food and Agriculture Organization of the United Nations (FAO) currently recommend the pesticide regulator identify, and withdraw from the sale, the pesticides most commonly used in fatal self-poisoning (16). These recommendations are supported by a systematic review of the impact of pesticide bans on suicides by pesticide poisoning as well as overall suicide rates, which is best exemplified by evaluations of bans in South Korea and Sri Lanka, which have both enacted bans on highly hazardous pesticides in recent years (17)(18)(19). Cost-effectiveness analysis shows that national bans of highly hazardous pesticides are highly cost-effective for reducing suicide by pesticides (20).
In China, pesticide regulation began in the 1970s (Table 1) (21). Local evaluations on the association between pesticide exposure and suicide attempt patterns have been undertaken in several provinces in China (22)(23)(24)(25). The impact of these bans has not been investigated at a national level. This study (1) collates information on the array of pesticide bans and regulatory changes since 1970 and (2) investigates the association between pesticide bans and regulations on national trends of suicide in China.
. Materials and methods . . Suicide and population data Suicide data were obtained from the China National Disease Surveillance Point (DSP) system for the period 2006-2018. The DSP system of cause-of-death data developed into a national and regionally representative sample vital registration system in 1990 with 145 DSPs. In 2004, the number of DSPs increased from 145 to 161, and the cause-of-death data were collected through a household survey before being reported to the local CDC. In 2013, the DSP system further expanded and the number of DSPs increased to 605, covering approximately 24% of the Chinese people with provincial representativeness (26). Cases of death in the DSP system are collected from hospitals and from household surveys to capture non-hospital deaths. All the data obtained are exchanged and certified with relevant departments, including civil affairs, public security, and the maternal and child healthcare sector (26,27). In this study, 158 DSPs that are consistent in the surveillance system from January 2006 to December 2018 were included in the analysis (with 3 DSPs excluded since 2013). Population data were derived from the National Statistical Bureau of China, and suicide rates were age-standardized using the standard population of China's 2010 census. Suicide counts in this study were defined based on the International Classification of Diseases 10th Revision (ICD-10) codes X60-X84, Y10-Y34, W75-W76, and X48. ICD-10 codes for counts of suicide by pesticide were X68 (intentional pesticide suicide), X48 (accidental pesticide suicide), and Y18 (intentional unknown pesticide suicide). Deaths by accidental pesticide (X48) poisoning were also included, as research indicates many of these deaths may be missed suicides (28).

. . Pesticide regulation data
Data on China's pesticide bans and regulations were collated using several sources. The key source was data from the Pesticide Action Network International (PAN International) (29-31). Information was also obtained from the FAO/WHO Joint Meeting on Pesticide Management Highly Hazardous Pesticide (JMPM HHP) lists (31). The website of the Ministry of Agriculture of China, the China Pesticide Information Network, and the Wan Fang database were used as additional sources as well as the personal notes of one of the authors (CJT). The search period for pesticide bans and regulations related to the earliest available period (1970) to 31 December 2021.

. . Definition of the intervention point
The intervention points were defined by combined change point detection (CPD) and content analysis of the pesticide policy. Altogether, 54 pesticides have been banned and restricted in China since 1970, with almost half [n = 24 (44.4%)] of them being banned or restricted in 2002 (Table 1). However, for many of the key highly hazardous organophosphorus pesticides (HHOP) used for self-harm, the years 2002-2004 represented only restrictions on registration. The pesticides were available (although at likely lower levels) for agriculture until their ultimate ban in December 2008. Therefore, December 2008 was defined as the primary intervention point for subsequent time-series models. The CPD method estimates the model by fitting iteratively the linear model and was implemented by the segmented package in R (32,33) and explored the monthly location of suicide by pesticides breakpoints, assuming that 1-4 breakpoints existed in 2006-2018 (Appendix Table 1). Considering the content of the policy and its implementation, we identified the potentially effective   Revoking: the registrations, production licenses, and production approval certificates of methamidophos, parathion-methyl, parathion, monocrotophos, and ammonium phosphate; Prohibits: their domestic production, circulation, and using on their own or mixed with other substances. Stops: processing and approving the field trial; Revoking: registration (including formal registration, repacking registration and provisional registration of pesticide formulations containing fipronil ingredient, and repeals their registration and production approval certificate and domestic sales, except for health, for corn seed coating agent, and for the export products.)

. . Statistical analysis
An interrupted time series (ITS) regression method was used in this study to evaluate the impact of the enactment of pesticide regulations on suicide in China (34)(35)(36). Our null hypothesis was that the intervention point was not associated with any changes in suicides by pesticide among adults aged 15 years and older in China. Segmented regression analysis was conducted for the analysis of three intervention policies (37). Generalized least squares method and fitting linear segmented model (function gls in R package nlme) were used to investigate the relationship between the standardized monthly number of deaths, interventions, and time (38). We considered January 2006 as the starting time, December 2008 as intervention 1, April 2012 as intervention 2, and July 2016 as intervention 3 to investigate whether trends of suicide changed after the pesticide regulations. The final regression equation is as follows: where S t is the standardized monthly suicide number in time t, people t is the number of standard population (in personyears) with log-transformed and was used as an offset term in time t, β 0 is the intercept, time is the corresponding month and year calculated in cumulative months, and β 1 represents the underlying pre-intervention trend. intervention 1 , intervention 2 , and intervention 3 were coded as a binary variable to reflect the pesticide policy enactment (pre-intervention period: 0; began and post-intervention period: 1), and the corresponding regression coefficients β 2 , β 4 , and β 6 indicate the level change and reflect the immediate effect following interventions 1-3, respectively. T 1 , T 2 , and T 3 were the time points when interventions started. (time − T 1 ) * intervention 1 , (time − T 2 ) * intervention 2 , and t(time − T 3 ) * intervention 3 were interaction terms between intervention and year, and β 3 , β 5 , andβ 7 , respectively, indicate the slope change following the intervention. Fourier terms consisting of sine/cosine three pairs were used to adjust seasonal variations. ε t is the error term.
We conducted two further analyses to evaluate the intervention effects. First, models were stratified to investigate any differences in trends by gender, the urbanization rate (≤45.14 vs. >45.14 %), age group, and the number of pesticide vs. non-pesticide suicides. The urbanization rate in all surveillance sites was dichotomized at the median as the cutoff point (high-urbanization rate sites as urban or low-urbanization rate sites as rural) based on China's 2010 census data. Age was categorized into three groups (15-44, 45-64, and ≥65 years old). Suicides involving methods other than pesticides were calculated by subtracting suicides by pesticide from suicides by all other methods. Second, due to the potential lag effect of the intervention, we repeated the ITS analysis in suicide by pesticide using the time lags between 1 and 18 months after introducing the policy in December 2008, April 2012, and July 2016, respectively.
We calculated the rate ratio (RR), the monthly percent change (MPC), the average monthly percent change (AMPC), and its 95% confidence intervals (CIs) for the periods preintervention and post-intervention. The details are found in the Appendix. The deseasonalized curve was used to depict the trend of change in suicide. We checked for autocorrelation by the autocorrelation function and the partial autocorrelation function (Appendix Figure 3). The model-selection strategies in this study were the smallest Bayesian Information Criterion (BIC) and optimal interpretability. R 4.  Table 2). The lag effect test showed that the trend of suicides by pesticide was significantly changed and tended to steepen after interventions 1 and 3 lagged from 0 to 14 and from 0 to 18 months, respectively, and tended to flatten after intervention 2 lagged from 0 to 6 months (Appendix Figure 1).
Before the policy intervention period in December 2008, suicide by pesticide and non-pesticide MPC was not statistically significant, but a significant decrease was evident after December 2008, and the decreasing trend significantly flattened after the policy in April 2012 ( Figure 2, Table 3, Appendix Table 3). After July 2016, the decreasing trend of suicide by pesticide became significantly steeper (RR = 0.992, 95% CIs: 0.990-0.994), while the trend change in suicide by non-pesticide methods was not significant (RR = 0.999, 95% CIs: 0.997-1.001) ( Figure 2, Table 3, Appendix Table 3).

. . Trends by sex
The standardized suicide rate for male and female individuals decreased by 40 Table 3).

. . Trends by age group
The standardized rate of suicide among those aged 15-44, 45-65, and ≥65 years declined by 43.5, 40.9, and 49.7%, respectively, from 6.68, 13.01, and 45. After the intervention in July 2016, the trend was decreasing and significant in all age groups. Meanwhile, the non-pesticide suicide rate significantly decreased in those aged 15-44 and 45-64 years after the intervention in December 2008, the trend change tended to flat significant only in those aged 15-44 years after the intervention in April 2012 and July 2016. All changes in suicide by non-pesticide for those aged 65 years and older were not significant after interventions 1, 2, and 3 (       decreasing trend tended to be flattened in high-urbanization areas after the intervention in April 2012 (Table 3, Figures 1, 2,  Appendix Table 3).

. Discussion
This study showed that China's regulations of banning and restricting pesticides were associated with significant contemporaneous declines in total suicide from 2006 to 2018, largely due to reductions in suicide by pesticide poisoning. A decline in total suicide and suicide by pesticide was evident in the period after the policy intervention in December 2008 (when HHOP pesticides were fully banned rather than restricted in use), with a flat decreasing trend after the policy intervention in April 2012 (when the new registration and production of paraquat parent drug and aqueous solution stopped and the existing paraquat production and after-sales services were regulated), and a steep decreasing trend after the policy intervention in July 2016 (when paraquat aqueous solutions were entirely banned rather than restricted in use). The decreases were more prominent in people living in urban areas, among female individuals, and people aged 15-44 years. Meanwhile, a lagged analysis suggests an increase in the association after the implementation of China's pesticide regulations in July 2016.
A study from South Korea also showed that the impact of a pesticide (paraquat) ban was more significant for both men and the elderly on suicide prevention and control, but the study only observed the short-term association over /fpsyt. .

FIGURE
The isolated trend cycle curve was smoothed of monthly suicide, pesticide suicide, and non-pesticide suicide rate after excluding the seasonal variations, including the counterfactual in China from January to December . Intervention : banned five highly hazardous organophosphorus (HHOP) pesticides implemented in December . Intervention : stopped the new registration and production of paraquat parent drug and aqueous solutions and regulated the existing paraquat production and after-sales services in April . Intervention : stopped domestic sales and use of paraquat aqueous solution in July .
the suicide rate and the pesticide self-poisoning suicide rate in rural China from 2009 to 2014 (41). Given the inconsistent time periods and locations of these several studies, we were unable to compare the trend and speed of the national suicide rate decline. The current study extends this regional study (25,39) and is based on an assessment of national pesticide regulations over a longer period and evaluates whether there were any contemporaneous declines in suicide over a much longer period and the firsttime studied paraquat pesticide policies' effect on suicide in mainland China. Similar to the research performed in other countries (42, 43), the obvious seasonality of suicide and pesticide suicide in China was present in our study, with the suicide rate peaking in spring and declining in winter (43), although seasonality tended to diminish (44). Between 2006 and 2018, the rate of suicide by pesticides in China declined at an average annual rate of 4.7%, with 4.4, 4.4, and 5.2% for male individuals, people living in rural areas, and people aged 15-44 years, respectively. At present, the suicide rate of male individuals in China is still higher than that of female individuals and remains much higher in rural areas than in urban areas (14).
Considering that the suicide rate in China has declined dramatically over the past for more than a decade and the proportion of suicide by pesticides dropped from over half to approximately a third, we consider the pesticide ban and restriction policy to have been implemented effectively in reducing the burden of suicide.
Our findings suggest that the potential impact of pesticide prohibition and restriction on suicide by pesticide was greater in rural areas. Simultaneously, the reduction in suicide by the nonpesticide method implies that the trend change in this period may be attributed to natural variation or other non-pesticide national suicide intervention policies, but the natural variation assumption can be discredited by comparing urban areas with rural areas and male suicide changes with female suicide changes. Moreover, the trend showed a different shift in the suicide and non-pesticide suicide groups. The association can be observed after the policy of banning and restricting pesticides was introduced intensively after December 2008, as the intervention point, when the five important HHOPs (methamidophos, monocrotophos, methyl parathion, parathion, and ammonium phosphate) were wholly removed from the market in China (45). Before 2008, the . /fpsyt. . prohibition and restriction policies on these high-toxicity pesticides had focused on revoking the relevant certificates of production, sale, and use (46,47), which implies that it was only in 2008 that these organophosphorus pesticides were removed entirely from the Chinese market. Therefore, it is likely that the policy of banning and restricting the sale and use of pesticides had lagged impacts on the prevention of suicide by pesticides in the period post-2008, which was consistent with the stronger association evident in the lagged analysis in this current study.
Our results showed an unlikely effect of the regulated paraquatrelated product in April 2012. The trend of suicide by pesticide tended to slow down significantly. Compared with the nonpesticide suicide group, a similar trend changed in both groups, prompting us to infer that a risk factor of suicide flattened the decrease of suicide, and even in rural areas where the intervention effect was more evident, the policy restriction effect may be masked by the risk factor. China has the largest periodically floating population of individuals migrating from rural to urban regions, who tend to have low socioeconomic and health status (48). The inter-provincial migrants increased to the highest level in 2012 and faced more life challenges than others (49,50). We hypothesize that the social phenomenon of migration influenced suicide and suicide by pesticides in China, particularly in 2012. However, stopping domestic sales and the use of paraquat aqueous solutions in July 2016 effectively decreased suicide by pesticide.
China is the world's largest pesticide user, is the second-largest pesticide producer, and has the largest agricultural population. Many suicides in China are believed to be impulsive as a reaction to acute psychological distress (13); the availability of highly hazardous pesticides facilitates such impulsive suicides (making low-intent acts of self-poisoning lethal). Research from other countries indicates regulations banning highly toxic pesticide regulations can lead to falls in method-specific and, in some cases (e.g., Sri Lanka, Bangladesh, South Korea, and China Taiwan), overall suicides (39, 51). Page et al. analyzed the national suicide mortality trends of China from 2006 to 2013 and found that hanging, as a proportion of all suicides, increased from 27 to 31%, while the contribution of pesticide poisoning to total suicides declined from 55 to 49%, although the absolute numbers of suicides using both methods reduced substantially (9). The current study also found that the proportion of suicide by pesticide has decreased in recent decades. The results of this study are consistent with those from previous studies and further provide evidence for timely and effective intervention in low-and middle-income countries where suicide by pesticide remains the main method of suicide.
In order to improve the quality and safety of agricultural products and reduce environmental pollution, the production, use, and trade of high-toxic and high-risk pesticides is increasingly controlled worldwide. The formulation and implementation of pesticide bans and policies of restriction is a process of multisectoral integration and cooperation with relevant stakeholders. This multi-sectoral collaboration is indispensable in the context of suicide prevention. Since the Stockholm Convention on Persistent Organic Pollutants (POPs) was signed by the Chinese delegation comprising representatives from the Ministry of Health, the Ministry of Agriculture, the State Economic and Trade Commission, and the General Administration of Environmental Protection in 2002, China's pesticide ban and restriction has been much strengthened, institutionalized, and standardized (45,52). With the introduction of a series of prohibition and restriction policies, the Ministry of Agriculture has intensified the research and development of low-toxic biological pesticides.
One key strength of this study was the use of a 15-year period of monthly suicide data collected by the DSP system, a continuous and consistent source of data that included certification of death with relevant local departments (civil affairs, public security, and the maternal and child healthcare sector) (26). This strongly supports the accessibility, comprehensiveness, and quality of suicide information in China (27). Another strength of our study concerns the comprehensive analyses of the effect on multiple key pesticide policies to avoid pesticide suicide decrease misattributed to concerned ban but neglected potential policy or elusive factor effects; thus, we further discussed the pesticides policies on suicide and non-pesticide suicide. Additionally, an ITS analysis was designed to evaluate the trends of population-level intervention (34), including at various lag periods, which accounted for seasonality and autocorrelation in time series.
This study had some limitations. First, similar to all other surveillance systems worldwide, the problem of accuracy and completeness is inevitable in the DSP system, where the potential sources of system bias remain. The surveillance data included in the analysis also did not adjust for potential under-reporting of suicide, which may cause information bias (11). However, its impact on trend analysis would be small. Second, the primary purpose of this study was to evaluate the single pesticide policy effect, but multiple policy efforts to ban different pesticides occurred at different time points during 2006-2021. Despite the substantial evidence collected, the performance of these policies and the effect between these policies could not be definitely determined. Furthermore, the multiple interventions also make it challenging to define the intervention time points in the interrupted time series regression analysis. Our approach to define the intervention time points was informed by the change point detection method implemented by segmented regression, and this may potentially lead to false positive findings in the interrupted time series regression analysis, although our final decision of the intervention time points was based on an analysis of the content of the pesticide ban policies. Third, only 2 years of suicide data are available before the HHOP pesticides ban in 2008 in our analyses, even though no specific number of time points was demanded, a short pre-intervention period may raise the uncertainty in the post-intervention trend judgment (34,36), but the monthly data would make up for this deficiency, to a certain extent. Fourth, ITS analysis is a quasi-experimental design method that assumes the omitted variable effect consistent across the study period; however, the facts may be different (12).

. Conclusion
Recent declines in suicide in China occurred contemporaneously with regulatory bans and restrictions on five key HHOPS in 2008 and a paraquat aqueous solution ban in 2016. Similar declines were observed for suicide by pesticide poisoning and for suicides using other methods since 2008. Larger declines were observed for suicide by pesticide poisoning .
than that for suicides using other methods, indicating the potential influence of pesticide bans on pesticide suicide trends since 2008.

Data availability statement
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

Author contributions
SL designed and supervised the study. YY, YJ, and RL analyzed the data and wrote the first draft. All authors have interpreted the data and contributed to intellectual content.