Changes in the pattern of suicides and suicide attempt admissions in relation to the COVID-19 pandemic

The present analysis is part of the suicide prevention program “FraPPE” (Frankfurt project to prevent suicides by evidence-based measures), funded by the German Health Ministry from 2017 to 2021 (Grant no. ZMVI1-2517FSB136). The program used a multi-level communal intervention in the city of Frankfurt am Main (inhabitants in 2020: ca. 765.000; the number of inhabitants did not meaningfully change from 2019 to 2020). All psychiatric hospitals in Frankfurt/Main participated in the project, as well as the communal health authority, and the Institutes of General Practice and Legal Medicine of the Goethe-University Frankfurt. Interventions comprised establishment of a 24/7 suicide prevention hotline, implementation of a specialized psychotherapy program focusing on prevention of recurrence after SA (ASSIP), education and training of GPs, communal outreach to emergency services to increase the rate of referrals to mental health services after SA and strengthening the communal suicide prevention network. The project started in 2017 and the measures ran until December 2020. Method monitoring and geo-analyses were important further parts of the project. The study will be described in detail in a later publication (in preparation); further information can be retrieved from the website frappe-frankfurt.de. Part of the project was as systematic assessment of routine data on suicide attempts and completed suicides. For the latter, forensic pathologists were called to the sites where completed suicides took place. The data collected on site and at the Institute of Legal Medicine were, on one hand, compared with the police investigation results and supplemented where necessary. On the other hand, the completed suicide cases were compared with the data from death certificates collected at the Communal Health Authority. In doing so, we could obtain the most complete data on completed suicide in Frankfurt/Main. To record suicide attempts, a structured documentation was implemented at all four psychiatric hospitals that provide inpatient mental health services in Frankfurt am Main (University Hospital Frankfurt, Agaplesion Markus Krankenhaus, Städtisches Klinikum Hoechst, Klinik Hohe Mark). Routine demographic and medical information of every in- and outpatient presenting to one of the hospitals after a suicide attempt was collected and entered into a case report form (CRF). Completeness of measurements was supported by integrating this assessment in the clinical SOPs, regular supervision in clinical routine and manual comparison of the clinical information system with the CRFs. The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008. As only routine data were collected, no Ethical Approval and Consent is required according to the Medical Association's professional code of conduct.

Suicide attempt (SA) was defined as deliberate self-harm with intend to die, irrespective of fatality probability [14]. Deliberate self-harm without intend to die, i.e., accidental drug overdose or self-harm for emotional tension relief without suicidal ideation e.g., in the context of borderline personality disorder, were not considered as suicide attempts. This was operationalized by development of a 24-item questionnaire to record SA in the psychiatric hospitals. The questionnaire and a study folder including all necessary information were presented and explained as part of in-house training courses in all participating hospitals. Immediately after admission, the clinician who treated and diagnosed the patient, filled out the case report form (CRF) and forwarded it to the responsible on-site FraPPE-team member (RG, TG, HS or LW). After quality checking, the anonymized CRFs were sent to the Department of Psychiatry of the University Hospital Frankfurt, where one of the authors (CS) centrally controlled every CRF regarding completeness and quality of the data and, if necessary, approached the responsible clinician in case of problems. Additionally, one percent of all CRFs which were drawn randomly from the overall sample were quality checked by the Institute of General Practice. Socioeconomic, individual and clinical variables were extracted from routine data in analogy to the European MONSUE project [15]. In the CRF, sociodemographic information, time and place of suicide attempt, main underlying psychiatric diagnosis, possible proximal triggers of SA, method of SA, consequences of SA (especially further medical and psychiatric treatments) and legal basis for admission were recorded. Completion of the CRF within 48 h upon admission was strived for; the CRF was completed by the clinician who saw the patient upon admission. Data are presented for all hospitals combined.

Completed suicide (CS) was defined as intentional self-harm with fatal outcome [14]. All CS cases from the Frankfurt metropolitan area were included, regardless of whether the person died immediately at the suicide scene or was still hospitalized and died in the course of the consequences of the intentional self-harm. Intentional self-harm was diagnosed in collaboration with the criminal investigators and the forensic pathologist, who were called to the suicide scene, and was primarily determined by existing suicide notes and interviews with family members, friends, or acquaintances, as well as medical records found. Deaths that could not be clearly attributed to intentional self-harm were excluded. A CRF was completed. When possible, the corpse was autopsied at the Institute of Legal Medicine: either on the order of a judicial postmortem examination by a local court, or, if no order was issued, with the consent of the relatives. In all autopsied cases, additional chemical-toxicological examinations were performed to detect medications, alcohol, and drugs in various body fluids and tissue samples. The number of CS was compared and cross-checked between the Municipal Health Authority and the Institutes of Legal Medicine and General Practice.

To examine the impact of the first COVID-19-related lockdown in 2020, we compared the data from March 2019–December 2019 (hereafter termed “baseline”) to the data from March 2020–December 2020 (hereafter termed “pandemic”). An overview on lockdown measures, beginning in March 2020, is given here [5]. All outcomes are presented using descriptive statistics; for continuous, normally distributed variables, mean, median and for binary and categorical variables frequency and percentages will be provided. For statistical testing, a significance level of alpha = 0.05 (5%) and two-sided hypothesis testing (if not specified otherwise) are applied. If variables are not normally distributed, non-parametric tests are conducted.

For data management, statistical analysis, and graphical visualization R version 3.6.1 (R Core Team, 2019) and RStudio version 1.2.5 (RStudio Team, 2020) were used.

During baseline conditions (March–December 2019), 430 SA were documented (thereof 210 females, 49%). In contrast, within the pandemic, i.e., March–December 2020, 296 SA were recorded (thereof 145 female, 49%). The incidence rate (total number of SA per inhabitants) differs significantly between baseline and the pandemic with a decrease of almost 30% observed during the pandemic (Incidence rate ratio (IRR) = 0.69, p < 0.0001). Regarding CS cases, 86 were recorded during baseline (thereof 32 females, 37%). During the pandemic (March–December 2020), 81 CS cases were documented (thereof 22 females, 27%). Thus, in contrast to SA, the incidence rate of CS did not differ significantly between baseline and the pandemic (IRR = 0.94, p > 0.05).

Table 1 provides an overview on the age and sex distribution during baseline and the pandemic. For SA, neither sex distribution nor average age was significantly different between baseline and the pandemic. From the 430 SA cases in 2019, 37 patients (8.6%) were under 18 years of age, while in 2020, 18 patients out of 296 (6.1%) were under 18 years old. The change in the percentage of minors was not significant (Pearson’s Chi-squared test, Chi² = 1.6, p > 0.05, OR = 0.69). The mean age of CS was not significantly different (54.8 years before the pandemic vs. 53.1 years during the pandemic). Two persons under 18 years of age died from suicide in either period. A detailed overview on the age distribution is given in Supplementary Figs. 1 and 2.

All following data relate to adults only. Regarding the marital status, there were no significant changes between both periods: 46 vs. 47% of patients were single, 7 vs. 8% were divorced, 5 vs. 5% were widowed, 21 vs. 19% were married, 6 vs. 6% were cohabitating in 2019 vs. 2020, respectively (remainder to 100%: unknown, other, or not documented; 15 vs. 14%). A significant difference was noted in the composition of the household (Pearson’s Chi-squared test, Chi² = 23.9, p = 0.013; Supplementary Table 1) purporting to an increased percentage of patients attempting suicide living alone. Moreover, the occupational status at baseline and during the pandemic was significantly different (Pearson’s Chi-squared test, Chi² = 27.0, p = 0.0003; Supplementary Table 2). However, the values are hard to interpret since the most pronounced change occurred in the category “other”. There were no significant changes in citizenship. While in 2019, 95 patients attempting suicide had their principal residence outside of Frankfurt (24%), this number declined to 53 in 2020 (19%).

Regarding CS, no significant differences were found regarding household composition and occupations status, however, the power of this sample was likely too low to detect changes. The incidence rate of CS in individuals with non-German citizenship increased from baseline to the pandemic from 4.4/100.000 inhabitants to 8.8/100.000 inhabitants (IRR = 1.99, p > 0.05), especially in the group of non-German EU citizens (1.0/100.000 to 13.2/100.000) (IRR = 13.3, p < 0.001), while the rate for German citizens decreased from 14.3/100.000 to 11.5/100.000 (IRR = 0.8, p > 0.05).

There were no significant changes in the main psychiatric diagnoses underlying SA between baseline and the pandemic in adult patients (Supplementary Table 3). In completed suicide, the rate of individuals where no psychiatric diagnosis could be established was comparatively high; no significant changes in the distribution of diagnoses were found.

With respect to the methods used in SA, we found a highly significant effect that the rate of intoxication-related SA significantly increased in adult patients during the pandemic (Pearson’s Chi-squared test, Chi² = 50.1, p < 0.0001; Table 2). In CS, there was slight, but non-significant, decrease in fatal suicidal intoxications, which was mirrored by a numeric increase in falls from height and collisions with a railway, as well as drowning and sharp violence.

Regarding the day of the week when a SA occurred, we observed a reversed pattern between the two datasets: while at baseline, the trough in SA was during the middle of the week, whereas Wednesday was the peak SA day during the pandemic (Pearson’s Chi-squared test, Chi² = 19.3, p = 0.004; Fig. 1a). No significant changes were observed with respect to the hour of SA (Pearson’s Chi-squared test, Chi² = 31.1, p > 0.05; Fig. 1b); most suicides during both periods occurred between 20.00 and midnight, although during the pandemic, almost no SA were documented between midnight and 8.00. No significant change was observed for the month of SA (not shown). No reliable data are available for time and weekday of completed suicides. The seasonal pattern of CS is shown in Supplementary Fig. 3.

A significant change was found in the location of SAs, as significantly less people attempted suicide elsewhere vs. at home during the pandemic (49 vs. 59%; Pearson’s Chi-squared test, Chi² = 6.6, p = 0.01, OR = 0.69). If SA was attempted outside home, there was also a shift in the SA location, in that less suicides occurred in train stations or railways (Pearson’s Chi-squared test, Chi² = 30.9, p = 0.02; Table 3). The number of CS committed at the registered address decreased from 69.8 to 53.1%, while the rate of CS at transport routes (railway, road) increased from 10.5 to 17.3%, as did CS in rivers/ponds (3.5–7.4%; all p > 0.05).

Certainly, there are several limitations to our study. First and foremost, only SA seen in psychiatric hospitals and attached emergency rooms were documented, but not SA seen exclusively outside of hospitals or in one of the other six Frankfurt hospitals that provide emergency, but not psychiatric care (BG Unfallklinik, Bürgerhospital, Hospital zum Heiligen Geist, Krankenhaus Nordwest, St. Katharinen-Krankenhaus and St. Elisabethen-Krankenhaus), which might explain a share of the dark figure of around 50% given above. However, patients admitted after SA are referred to one of the institutions that participated in the present study. There is no reason to assume that referral patterns changed during the pandemic.

More uncertainty exists regarding the cases seen by practitioners only; however, even if reduced referral to psychiatric hospitals would fully account for the difference between baseline and the pandemic, this would not invalidate our conclusion that mental health care provision to those with greatest need was reduced in 2020. Our data should, therefore, draw attention to improve public mental health services to prevent an increase in suicidal behaviors in the aftermath of the COVID-19 pandemic. Further studies that monitor secular trends in suicidal behaviors from all healthcare sectors are urgently needed to allow preventive measures; the fact that no effective surveillance system for SA/CS is in place despite the high death toll is pointing to a huge unmet medical need that needs to be addressed.