Seasonal changes and decrease of suicides and suicide attempts in France over the last 10 years

The objective of this study was to examine the evolution of suicides and suicide attempts over the last 10 years in France. We analyzed the number of deaths by suicide and suicide attempts in metropolitan France (2009–2018) from French national databases, on a daily, weekly or monthly basis. Seasonal variation in suicide rates was modeled using a cosinor function. Based on this model, we determined the association of suicides and suicide attempts with geographic characteristics, age, gender, means used and psychiatric disorders. The number of suicides and suicide attempts decreased over the last ten years in France (mean decrease of 14.49% and 11.69% per year). We observed a significant and recurrent seasonal pattern of suicides and suicide attempts in France, with a peak in spring. The suicide and suicide attempt rates were higher in the northern departments of France. Suicides were more frequent for men (75%) and middle-age individuals (45–54 years old), while suicide attempts were more frequent for women (62%) and young adults and middle-age individual (15–19 and 40–49 years old). Nearly two-thirds of the patients who attempted suicide suffered from comorbid psychiatric disorders. Mood disorders was the most frequent comorbid psychiatric disorder (54%). Voluntary drug intoxication was the most common means of suicide attempt (80%), and hanging was the most common means of suicide (54%). The number of suicides and suicide attempts decreased in France over the last 10 years, with large and recurrent seasonal variations. These findings could be used to alert and adjust prevention policies, as well as developing preventive strategies such as chronotherapeutics.

www.nature.com/scientificreports/ In France, the National Public Health Agency (Santé Publique France) has implemented an epidemiological surveillance for suicide attempts and suicides in France. The objective of this study was to examine the evolution of suicides and suicide attempts over the last 10 years in metropolitan France.

Methods
Study design and population. We examined the number of death by suicide and suicide attempts in metropolitan France for the last 10 years (2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018) in French national databases provided by Santé Publique France. As the national public health agency, Santé Publique France has full legal access to national health databases 7 .
More specifically, the data on deaths by suicide and suicide attempts were provided on a daily basis, from two databases: -A database for suicide attempts, with data from January 1, 2009, to December 31, 2018, grouped by department (French local administrative division). These data corresponded to all medical visits (consultations, hospitalizations, emergencies) for suicide attempts over the period of the study. These data were collected from the database using PMSI-MCO (Programme de Médicalisation du Système d'Information-Médecine Chirurgie Obstétrique) coding and correspond to ICD-10 codes X60 to X84 (detailed below). -Deaths by suicide were extracted from the death certificates database (CepiDc-Inserm, medical research institute in France). These were daily data, by departments, over a period from January 1, 2009, to December 31, 2015. Data after this date were not available at the time of the extraction for analysis.
We calculated a rate of suicide and suicide attempts per department and year, using the number of inhabitants per department and year provided on a database available on the INSEE website (https:// www. insee. fr; National institute for statistics and economic studies).
All the methods were performed in accordance with relevant guidelines and regulations.
Analysis. Daily data were grouped by week or month. Suicide rates were calculated by department per 100,000 inhabitants, per day, week and month. We made a statistical description of the different variables. We statistically examined the seasonal effect of deaths by suicide and suicide attempts by using the cosinor model, which captures a seasonal pattern using a sinusoidal wave (analysis made with the "Season" R package, made by Barnett & Dobson). Analyses were carried out using R Studio software, version 1.1.463.

Results
The complete observation period was of 3652 days for suicide attempts and of 2556 days for deaths by suicide, reporting 68,316 deaths by suicide and 913,938 suicide attempts.
Demographic data. Patients who attempted suicide were mostly female [564,338 women (74.8%) and 349,600 men (25.2%)]. An opposite ratio was observed for deaths by suicide, with mostly men (74.8%, n = 51,148) compared to women (25.2%, n = 17,168) ( Fig. 1A1,B1). Suicide attempts were more frequent in younger patients, with two peaks: 15-19 years old and 40-49 years old. Deaths by suicide increased progressively with age, with a peak between 45 and 54 year-old and with a flattening at older ages see Fig. 1A2,B2).
Suicide and suicide attempt rates. The average number of suicides and/or suicide attempts per day per department in France was 3.068 over the entire study period. The department with the highest rate of suicide attempts per month per 100,000 inhabitants was the Somme in the North (6.63 suicide attempts per week per 100,000 inhabitants). Other departments with the highest rate of suicide attempts per week per 100,000 inhabitants were: Côte d' Armor, Manche, Finistère, Seine Maritime (North-West), Nord, Pas de Calais, Oise (North), Vosges and Territoire-de-Belfort (North-East).
Detailed rates of suicide and suicide attempts per day, per week, and per month for each department in metropolitan France are given in Table S1 in Supplementary Material.
Suicide rates and suicide attempts rates are higher in the northern departments of France than in the south. Figure 1A4 and B4 shows the rates of suicide attempts (A4) and suicide (B4) by month, based on the latitude  Psychiatric disorders. On average, 63.27% of suicide attempts were associated with a diagnosed psychiatric disorder. Mood disorders were the most frequent psychiatric disorders associated with suicides and suicide attempts. Then, substance use disorders and personality disorders were also often associated with suicide attempts ( Table 1). Figure 3 shows the evolution of the number of suicides and suicide attempts over the years. The number of suicides and suicide attempts has decreased over the last ten years. Before 2012, the number of suicide attempts was over 100,000 and the number of suicide was over 10,000. After 2012, a continuous decrease was observed with 90,000 suicide attempts (2018) and less than 9000 suicides (2015). Thus, between 2009 and 2015, the total number of suicides decreased by 14.49%, and the total number of attempted suicides decreased by 11.69%. Figure 1A5 and B5 shows the evolution of the number of suicide attempts and the number of suicides according to the month of the year, and shows a seasonal pattern of suicidal behaviour that replicates itself every year over the 10 years. A decrease is observed in the number of suicide attempts in the winter (off-peaks in December and-February), with an increase in early spring (April-May). The peak of suicide attempts is reached in May-June, before decreasing to a minimum in August. There is also a drop in the number of visits to the emergency room for suicide attempts during the last month of the year (mainly the last week).

Seasonality of suicide attempts and suicides.
Using an adjusted significance level of 0.025, we found a significant seasonality effect with the cosinor function, for the numbers of deaths by suicide and suicide attempts for each year during the 10 years follow-up (p < 0.0001). The period of months with the highest number of events was 4.3 (i.e. between April and May) for suicide attempt, and 5.1 (i.e. at the beginning of May) for suicide, and the period with the lowest number of events was 10.3 (i.e. between October and November) for suicide attempt and 11.1 (i.e. at the beginning of November) for suicide.

Discussion
This study reports two important observations. First, a decrease of suicides and suicide attempts occurred in France since 2010. Secondly, a seasonality is repeatedly observed over years, regarding both suicides and suicide attempts, with a peak in spring.
Regarding the decrease of suicides and suicide attempts from 2010, it may be the consequence of several suicide prevention programs launched in France and worldwide since 2000. The four axes of suicide prevention defined by the Ministry of Social Affairs and Health 8 are: promoting prevention, reducing access to lethal means, promoting access to healthcare, and improving knowledge of the epidemiological situation 9 . In a review of the literature of efficient interventions on suicide prevention, the three most efficient categories of interventions www.nature.com/scientificreports/ were the limitation of access to lethal means, the preservation of contact with the patients hospitalized for a suicide attempt after hospitalization, and the implementation of emergency call centers 10 . Over the last 10 years in France, this kind of measures to prevent suicidal repetitions have been implemented, through postcards, recalls and home visits 11 . These interventions have shown a tendency to reduce suicidal repetitions 11 . In addition, the VigilanS system has been implemented in many regions in France, combining brief contact interventions and a sanitary monitoring of suicide attempts 12,13 . The repeated seasonal peak in suicidal behavior observed in spring and early summer in France confirmed in France previous observations from other international studies [14][15][16][17] . A possible pathophysiological hypothesis is that after cold nights during spring and early summer, the warming of the human body during the day would cause hyper activation of the brown fat tissue, causing increased anxiety and mental activity, and promoting suicidal behavior [18][19][20] . Another possible mechanism would be that the increase in temperature could promote violent behavior and impulsivity by increasing serotonin levels [21][22][23][24][25] . An Australian study reported decreased serotonin levels in winter, and increased levels in summer, suggesting a sensitivity of serotonin to temperature and duration of sunshine 21 . These hypotheses are also consistent with our observations of a lowest rate of suicides and suicide attempts in February, corresponding to the time when the temperature is the coldest. However, we observed a second lowest rate in the number of suicides and suicide attempts in August. This drop in the number of suicidal behaviors can be explained in part by the fact that the month of August corresponds to the major school vacations for the majority of French people, which means time spent with family or friends, which is a protective factor against suicidal behavior 4 . Indeed, the increase of temperature, including at very elevated temperatures, rather leaded to a plateau of increased suicidal behaviors 26,27 . The same hypothesis could be made for the fall of the month of December, corresponding to the end of year celebrations and family reunions. Finally, the photoperiod changes in spring, with an increase of daylight length, may also lead to suicidal behaviors as a consequence of circadian rhythm and sleep-wake cycle alterations 28 . Further studies are expected to better understand underpinning physiological mechanisms of the meteorological factors impact on suicidal behaviors.
These findings highlighting a clear seasonal pattern of suicidal behaviors warrant to impact the existing public health measures mentioned previously, such as reinforcing the link during the spring and summer with patients with a history of suicide attempts or suicidal thoughts, via brief contact interventions for instance 13,29 In addition, broad public health campaigns of prevention and awareness should be delivered during the at-risk period. At the individual level, practitioners should be more vigilant during the spring and summer for at-risk patients with suicidal ideation, or history of suicidal behaviors, or a psychiatric disorder at high risk of suicide.
Approximately one-third of patients who committed suicide and two-thirds of patients who attempted suicide had an associated psychiatric disorder. The most common psychiatric disorder was mood disorders. Those with a substance use disorder were more likely to commit suicide attempts, but not completed suicides. Psychiatric disorders are one of the most important risk factors for suicide 4,29,30 . As such, about one-third of patients with bipolar disorder will attempt suicide in their lifetime 31 . Moreover about one quarter of patients with bipolar disorder have a seasonal pattern of their depressive recurrences, and one out of six patients present with seasonal manic recurrences 32,33 . Patients with schizophrenia are up to 20 times more likely to die by suicide than the general population use 34 . A particular shared vulnerability between suicidal behaviors and psychiatric disorders has been confirmed in our study. This observation calls for systematically screen for suicidal behaviors in these at-risk psychiatric populations, may be at particular at-risk periods with targeted public health interventions or programs.
Limitations. Some limitations should be acknowledged. First, the suicide database extracted from the death certificates has an underestimation of about 10% for the national data, due in particular to deaths for which the cause remains undetermined or to those for which a forensic examination has been carried out and the conclusion has not been reported 35 . Similarly, a probably non-negligible and difficult to estimate number of suicide attempts are not medicalized, and therefore have not been recorded in our databases. However, this rate of nonmedicalized suicide attempts seems to be stable over time 36 . This is also a common caveats, already reported by numerous studies examining suicide behaviors. Second, only two-third of patients who attempted suicide had a reported psychiatric comorbidity. This rate suggests that either a significant proportion of patients with suicidal behavior have no psychiatric history or that such history is under-diagnosed or sought in practice. It may also be that there is a lack of coding or reporting of these psychiatric disorders in hospital departments. In addition, we were unable to explore psychiatric comorbidities associated with deaths by suicide because of a lack of information on the death certificates. A review about psychological autopsy studies estimated that more than 90% of suicide deaths are associated with a comorbid psychiatric disorder 37 . This under-reporting / under-coding of psychiatric comorbidities after a suicide attempt or a death by suicide shows that there is a need to improve the screening for associated psychiatric disorders in France. In our analyses, we were interested in the distribution of the different psychiatric disorders rather than their prevalence. Despite these limitations, this work presents a good epidemiological synthesis of the state of suicide and suicide attempts in France over the last years.

Conclusion
The number of suicides and suicide attempts seems to have decreased in France over the last 10 years, suggesting the effectiveness of the public health prevention measures. A seasonality in suicidal behaviors exist with a peak in spring, which requires further studies to better understand underpinning physiological mechanisms and the different meteorological factors that may influence these variations.

Data availability
Data available on request from the authors.