Journal of Affective Disorders 302 (2022) 332–351

Contents lists available at ScienceDirect

Journal of Affective Disorders

journal homepage: www.elsevier.com/locate/jad

Predictors of suicidal ideation, suicide attempt and suicide death among

people with major depressive disorder: A systematic review and
meta-analysis of cohort studies
Xinming Li a, b, Fuqin Mu c, Debiao Liu a, Jin Zhu a, d, e, f, Song Yue a, g, Min Liu a, Yan Liu a, d, e, f, *,
JianLi Wang a, h, i, *
a
School of Mental Health, Jining Medical University, Jining 272013, China
b
Department of Psychiatry, Shandong Mental Health Center, Shandong University, Jinan 250014, China
c
School of Basic Medicine, Jinzhou Medical University, Jinzhou 121002, China
d
Shandong Key Laboratory of Behavioral Medicine, School of Mental Health, Jining Medical University, Jining 272013, China
e
Shandong Collaborative Innovation Center for Diagnosis & Treatment & Behavioral Interventions of Mental Disorders, Institute of Mental Health, Jining Medical
University, Jining 272013, China
f
Center of Evidence-Based Medicine, Jining Medical University, Jining 272013, China
g
Department of Pathology, Weifang Medical University, Weifang 261053, China
h
School of Epidemiology, Public Health and Preventive Medicine, University of Ottawa, Ottawa K1Z 7K4, Canada
i
The Royal’s Institute of Mental Health Research, University of Ottawa, Ottawa K1Z 7K4, Canada

1. Introduction Identifying individuals with MDD who are at high risk of suicide is a
public health challenge and has important clinical implications (May
Suicide is a global public health challenge nowadays. Every 40 s, one et al., 2012; Spijker et al., 2010; Witte et al., 2009; Harris and Barra­
person may die because of suicide (Nock et al., 2008a). Suicide-related clough, 1997). In terms of prediction, the predictors of suicidal ideation,
screening or assessment tools such as Tool for Assessment of Suicide Risk suicide attempt and suicidal death among MDD individuals may be
(TASR) which has been widely used in clinical practice, could help different (Maniam et al., 2014; Nock et al., 2008b). For example, sui­
professionals effectively identify people at suicide risk, thus contributing cidal ideation is episodic, and has quick onset and short duration. So far,
to suicide prevention and suicide deaths (Wong, 2018; Large, 2018). The it is difficult to monitor and intervene suicidal ideation in real time.
crisis interventions for suicide (e.g. help hotline, psychiatric emergency Factors which could predict suicide attempt or deaths may not be pre­
room and cognitive behavioral therapy, etc.) could reduce suicide dictive for suicidal ideation (Kleiman and Nock, 2018).
deaths temporarily, and these interventions might also reduce the Awareness of risk factors for suicide in MDD is crucial for informing
prevalence of suicide ideation and attempt (Nordentoft, 2011; Wong, clinical practice and suicide interventions (Hawton et al., 2013; May
2018). However, the causes of suicidal ideation, suicide attempt and et al., 2012). A meta-analysis in 2013 including 19 studies showed that
suicide death may be different and the etiology can be multifactorial, male gender, family history of psychiatric disorder, previous attempted
including biological factors, psychological factors, cognitive factors, and suicide, anxiety comorbid and severe depression were predictors for
environmental factors (Sudol and Mann, 2017; Coentre et al., 2017; suicide death in people with depression which included not only MDD,
Gournellis et al., 2018). but melancholia and other mood disorders (Hawton et al., 2013).
Many people who had died of suicide have mental disorders, espe­ Although the correlations between suicide and depression had been
cially major depressive disorder (MDD) which occurred in half to two discussed in many studies (Braun et al., 2016; Maniam et al., 2014; Nock
thirds of suicide cases (Hegerl, 2016). The pooled lifetime prevalence of et al., 2008b; Nordentoft, 2011; Wong, 2018), the predictors of suicide
suicide attempt was 31% (95% CI: 27%− 34%) and the pooled 1-year in depressed patients are inconsistent. There was no meta-analysis that
prevalence was 8% (95% CI: 3%− 14%) (Dong et al., 2019) in 27,340 illustrate predictors of suicidal ideation, suicide attempt and suicide
individuals with MDD. 1.45 suicide deaths and 2.76 suicide attempts per death among people with MDD, respectively. The aim of this study was
1000 patient-years were reported among 6934 MDD individuals even to conduct a systematic review and meta-analysis of the predictive
during the long-term antidepressants treatment (Braun et al., 2016). factors for suicidal behaviors (suicidal ideation, suicide attempt and

* Corresponding authors.
E-mail addresses: [email protected] (Y. Liu), [email protected] (J. Wang).

https://doi.org/10.1016/j.jad.2022.01.103
Received 22 September 2021; Received in revised form 25 January 2022; Accepted 27 January 2022
Available online 29 January 2022
0165-0327/© 2022 Elsevier B.V. All rights reserved.
X. Li et al. Journal of Affective Disorders 302 (2022) 332–351

Fig. 1. Flowchart of selection processing

suicide death, respectively) among people with MDD. 2.3. Inclusion and exclusion criteria

2. Methods Inclusion criteria for this study were as follows: (1) Participants had
confirmed diagnosis of MDD; (2) Study outcomes included suicidal
2.1. Protocol and registration ideation, suicide attempt and/or suicide death; (3) Studies used cohort
or longitudinal design; (4) No restrictions on languages and published
The systematic review and meta-analysis was registered in PROS­ periods.
PERO (International Prospective Register of Systematic Reviews) with Exclusion criteria for this study were as follows: (1) No reports on
the online protocol number of CRD42019123324 and was conducted in suicidal ideation, suicide attempts and suicide death; (2) No reports on
accordance with the Preferred Reporting Items for Systematic Reviews MDD; (3) Reports on depressive symptoms; (4) Cross-sectional studies,
and Meta-analyses (PRISMA) guidelines (Moher et al., 2009). case-control studies, randomized clinical trials; (5) Did not report or
could not calculate hazard ratios (HR), risk ratios or relative risks (RR)
2.2. Source and literature searching and its 95% confidence interval (CI) information; (6) Animal studies or
cell studies; (7) Case reports, letters, comments, etc.
Six English electronic databases (PubMed, Embase, Web of Science,
SpringerLink, PsycINFO and Cochrane Library) and three Chinese elec­ 2.4. Measurements of major depressive disorder
tronic databases (China Knowledge Resource Integrated Database,
Wanfang database, and Weipu database) were searched from their in­ MDD was diagnosed by clinicians or using fully structured, e.g.,
ceptions until March 15, 2021. We performed Boolean search technique, Composite International Diagnostic Interview (CIDI), or semi-
e.g., (depression OR depressive disorder* OR major depressive disorder* structured, e.g., Structured Clinical Interview for DSM (SCID) in­
OR MDD OR depressi* state) AND (suicide OR self-injury OR suicid* struments based on Diagnostic and Statistical Manual of Mental Disor­
thought* OR suicid* ideation* OR suicid* attempt* OR self-injurious OR ders (DSM) or the International Statistical Classification of Diseases
self-destruction OR self-mutilation OR self-inflicted OR self-slaughter) (ICD) and Related Health Problems criteria.
AND (cohort OR follow-up OR followed-up OR panel OR longitudinal
OR prospective). Relevant references were searched as well. The flow­
chart was shown in Fig. 1. 2.5. Measurements of suicide related behaviors

Suicidal ideation and suicide attempt were measured by CIDI, ISC

333
 X. Li et al.
Table 1
Characteristics of 24 studies included (NOS≥6) for systematic review in 954,822 participants with major depressive disorder.
No. Authors Published Country Follow-up Sample MDD Age Inventory of Outcome Inventory of No. of Predictors NOS Meta-
year period size patients (years) Depression* outcome** suicide score analysis
related included
population

1 Aaltonen 2019 Finland 24 years 56,826 56,826 ≥18 Clinical suicide ICD − 9 2587 male, depression severity, suicide 6 Yes
et al. diagnosed death attempts, depression, alcohol
dependence, income, education
2 Alexopoulos 1999 United 6 months 354 354 74.17 Clinical suicidal HDRS 155 previous suicide attempts with 6 Yes
et al. States to 7 years (mean) diagnosed ideation serious intent, poor social support,
(mean= impaired instrumental activities of
1.8 years) daily living, limited impairment in
activities of daily living, severity of
depression
3 Bradvik et al. 2008 Sweden 50 years 687 424 0–92 DSM-IV suicide ICD 6–9 28 male, severe depression 7 Yes
death
4.1 Coryell et al. 2019 United 16.6 years 838 838 35.9 Clinical suicide Death 23 anxiety symptoms, obsessive- 8 Yes
States (mean) diagnosed death registration compulsive, somatic,
depersonalization, anxiety disorder,
obsessive-compulsive disorder,
generalized
4.2 Coryell et al. 2019 United 16.6 years 838 838 35.9 Clinical suicide Self-report 90 anxiety symptoms, obsessive- 8 Yes
States (mean) diagnosed attempts compulsive, somatic,
depersonalization, anxiety disorder,
obsessive-compulsive disorder,
generalized
5.1 Fergusson 2003 New 21 years 1063 184 14–21 DSM-IV suicidal Self-report 59 family factors, individual factors, 8 Yes
334

et al. Zealand ideation school factors

5.2 Fergusson 2003 New 21 years 1063 184 14–21 DSM-IV suicide Self-report 17 family factors, individual factors, 8 Yes
et al. Zealand attempts school factors
6 Grunebaum 2010 United 1 year 136 136 39.2 DSM-IV Suicide SSI 12 Less secure/more avoidant 7 No
et al. States (mean) attempt attachment, poor social adjustment
7.1 Handley 2018 Australia 1, 3 and 5 1051 364 53.71 CIDI-3.0 suicide CIDI 3.0 76 depression severity, age, age of 7 Yes
et al. years (mean) attempts depression onset, comorbidities,
help-seeking
7.2 Handley 2018 Australia 1, 3 and 5 1051 364 53.71 CIDI-3.0 suicidal CIDI 3.0 254 depression severity, age, age of 7 Yes
et al. years (mean) ideation depression onset, comorbidities,
help-seeking
8 Hansen et al. 2003 Denmark 10 years 210 210 48.4 DSM-III suicide Death 15 personality disorder 7 No
(mean) death registration
9 Holma et al. 2010 Finland 5 years 249 249 39.6 DSM–IV suicide Beck Scale 36 age, gender, psychotic features, 8 No
(mean) attempts alcohol dependence, marital status,

Journal of Affective Disorders 302 (2022) 332–351

suicide attempt prior to study entry,
social support
10.1 Holma et al. 2020 Finland 5 years 269 269 20–60 SCID-II suicidal SSI 15 alcohol abuse/ dependence 6 Yes
ideation
Holma et al. 2020 Finland 5 years 269 269 20–60 SCID-II suicide SSI 15 alcohol abuse/dependence 6 Yes
10.2 attempts

11 Høyer et al. 2004 Denmark 20 years 53,466 22 >15 ICD-8 suicide Death 1487 gender, diagnosis of alcohol/drug 6 No
death registration abuse, number of psychiartic
admission, admission conditions,
duration of illness, time after
admission to psychiatric hospital,
time after discharge from psychiatric
hospital
(continued on next page)
 X. Li et al.
Table 1 (continued )
No. Authors Published Country Follow-up Sample MDD Age Inventory of Outcome Inventory of No. of Predictors NOS Meta-
year period size patients (years) Depression* outcome** suicide score analysis
related included
population

12 Ilgen et al. 2009 United 5 years 887,859 885,967 >18 NARDEP suicide ICD-9/ ICD- 1892 any substance use disorder, those 7 No
States death 10 with a substance use disorder (Non-
African American), those with a
substance use disorder who are
not African American (any inpatient
treatment in the
prior 12-months), Those without a
substance use disorder(male),those
without a substance use disorder who
were male(non-African American)
13.1 Kovacs et al. 1993 United 12 -month 183 134 11.2 DSM-III suicidal ISC items 62 major depressive disorder, comorbid 7 No
States (mean) ideation major depressive and conduct/
substance use disorder
13.2 Kovacs et al. 1993 United 12 -month 183 134 11.2 DSM-III suicide ISC items 18 major depressive disorder, comorbid 7 No
States (mean) attempts major depressive and conduct/
substance use disorder
14 Mattisson 2007 Sweden 30–49 3563 436 35 DSM-IV suicide Death 12 gender, age at onset, severe 6 Yes
et al. years (median) death registration impairment, alcohol disorder
15 May et al. 2012 Canada 10 years 49 49 30 DSM-IV suicide SCID 13 anxiety disorder at baseline, 7 Yes
(mean) attempts substance abuse at baseline,
substance abuse lifetime, personality
disorder, hopelessness, social
adjustment difficulty, poor maternal
335

relationship
16 Pan et al. 2017 Taiwan 1 year 1710 140 13–14 Adolescent suicidal Self-report 70 BMI category 6 No
Depression ideation
Inventory
17.1 Petteri 2003 Finland 15 months 269 269 39.6 DSM-III suicidal SSI 156 sex, age, HAM-D score, HS score, 8 Yes
Sokero et al. (mean) ideation alcohol dependence/abuse, PSSS-R
score, SOFAS score
17.2 Petteri 2003 Finland 15 months 269 269 39.6 DSM-III suicide SSI 41 sex, age, HAM-D score, HS score, 8 Yes
Sokero et al. (mean) attempts alcohol dependence/abuse, PSSS-R
score, SOFAS score
18 Riihimaki 2014 Finland 5 years 134 134 45.3 DSM-IV suicide SSI 14 gender, age, follow-up time, co- 6 Yes
et al. (mean) attempts morbid substance use disorder,
suicide attempts prior to study entry
19 Schneider 2001 Germany 5 years 278 278 39.8 DSM-III-R suicide Self-report 16 hypochondriac delusions and 7 No
et al. (mean) death preoccupations, initial insomnia,

Journal of Affective Disorders 302 (2022) 332–351

recurrent depressive episodes
20 Seo et al. 2014 South Korea 2 years 344 344 46. DSM-III-R suicide SSI-B 5 Temperament dimension (novelty 7 No
(mean) attempts seeking, harm avoidance, reward
dependence, persistence), character
dimension (self-directedness,
cooperativeness, self-transcendence)
21 Seo et al. 2015 South Korea 9 years 1003 1003 48.1 DSM-IV suicidal SSI-B 565 recurrent depression, history of 7 Yes
(mean) ideation suicide attempt, living alone, single
22.1 Spijker et al. 2010 The 2 years 586 248 18–64 CIDI suicidal CIDI 97 gender, living without partner, 8 Yes
Netherlands ideation comorbid anxiety, major depressive
disorder, previous suicidal ideation,
use of professional care, dysthymia
22.2 Spijker et al. 2010 The 2 years 586 248 18–64 CIDI suicide CIDI 19 8 Yes
Netherlands attempts
(continued on next page)
 X. Li et al.
Table 1 (continued )
No. Authors Published Country Follow-up Sample MDD Age Inventory of Outcome Inventory of No. of Predictors NOS Meta-
year period size patients (years) Depression* outcome** suicide score analysis
related included
population

living without partner, comorbid

anxiety, previous suicidal ideation,
previous suicidal acts
23.1 Wang et al. 2015 United 3 years 6004 6004 ≥18 DSM-IV suicidal Self-report 1249 stressful life events 8 Yes
States ideation
23.2 Wang et al. 2015 United 3 years 6004 6004 ≥18 DSM-IV suicide Self-report 169 stressful life events 8 Yes
States attempts
24 Zivin et al. 2007 United 5 years 887,859 885,967 >18 NARDEP suicide ICD-9/ ICD- 1683 gender,race,Hispanic,any substance 7 Yes
States death 10 abuse,age(>65,45–64,18–44), PTSD,
previous inpatient stay for psychiatric
disorder, service connection, regin,
336

Charison score
Fourth Edition criteria
.
DSM-III: The Diagnostic and Statistical Manual of Mental Disorders
.
Third Edition criteria
.
WMH-CIDI 3.0: World Mental Health Composite International Diagnostic Interview version 3.0
.
ICD-8: The International Statistical Classification of Diseases and Related Health Problems 8
th Revision.
ICD-9: The International Statistical Classification of Diseases and Related Health Problems 9
th Revision.
ICD-10: The International Statistical Classification of Diseases and Related Health Problems 10
th Revision.
SCID-II: Structured Clinical Interview for DSM-III-R personality disorders
.
NARDEP: the VA’s National Registry for Depression.
* DSM-IV: The Diagnostic and Statistical Manual of Mental Disorders
**
SSI: Scale for Suicide Ideation; SSI-B: Beck Scale for Suicide Ideation; ISC: Symptom-oriented Interview Schedule for Children; HDRS: Hamilton Depression Rating Scale; CIDI: Composite International Diagnostic
Interview; ICD: The International Statistical Classification of Diseases and Related Health Problems; SCID: Structured Clinical Interview for DSM-III-R.

Journal of Affective Disorders 302 (2022) 332–351

X. Li et al. Journal of Affective Disorders 302 (2022) 332–351

Table 2
The summary of meta-analysis results of predictors for suicide related behaviors in people with major depressive disorder.
Suicide Items Predictors References Statistical Combined 95% CI Inverse- No. of Egger’s
related index effects Lower Upper squared test included test p value
behaviors level level p value studies *

Suicidal ideation
Demographic
factors
Living alone Not living alone RR 1.76 1.24 2.50 0.803 2 –
History of
diseases
Recurrent No recurrent RR 3.65 0.72 18.55 0.108 2 –
depression depression
Previous suicidal No previous RR 3.63 2.57 5.13 0.634 4 –
behaviors suicidal
behaviors
Severe No severe RR 1.68 1.17 2.43 <0.001 6 0.247
depression depression
Alcohol use No alcohol use RR 1.37 0.70 2.68 0.070 2 –
related disorders related disorders
Behaviors
Help-seeking No help-seeking RR 1.65 1.01 2.69 0.623 3 –
Environmental
factors
Negative school No negative RR 2.13 1.70 2.69 0.743 2 –
factors school factors
Negative family No negative RR 1.43 1.28 1.59 <0.001 9 0.002
factors family factors
Stressful life No stressful life RR 1.15 1.07 1.24 0.001 16 0.122
events events
Positive school No positive RR 0.66 0.36 1.22 <0.001 2 –
factors school factors
Suicide attempts
Demographic
factors
Male Female RR 0.99 0.40 2.47 0.267 2 –
History of
diseases
Substance No substance RR 9.29 3.60 23.99 0.396 3 –
misuse misuse
Previous suicidal No previous RR 6.38 2.97 13.72 0.704 3 –
behaviors suicidal
behaviors
Anxiety No anxiety RR 4.08 1.71 9.72 0.872 2 –
disorders disorders
Alcohol use No alcohol use RR 3.63 2.03 6.50 0.523 4 –
related disorders related disorders
Severe No severe RR 2.45 0.88 6.85 <0.001 4 –
depression depression
Hopelessness No hopelessness RR 1.06 0.97 1.16 0.185 3 –
Anxiety No anxiety HR 1.03 1.00 1.05 0.650 8 0.438
symptoms symptoms
Anxiety No anxiety HR 1.03 0.79 1.34 0.729 5 –
disorders disorders
Environmental factors
Negative school No negative RR 3.87 2.61 5.75 0.761 2 –
factors school factors
Negative family No negative RR 1.65 1.42 1.93 <0.001 10 0.001
factors family factors
Stressful life No stressful life RR 1.34 1.12 1.60 0.001 18 0.225
events events
Positive school No positive RR 0.45 0.15 1.33 <0.001 2 –
factors school factors
Suicide death
Demographic factors
Male Female HR 3.27 1.55 6.90 0.057 3 –
History of diseases
Previous suicidal No previous HR 2.10 1.92 2.30 0.844 4 –
behaviors suicidal
behaviors
Anxiety No anxiety HR 1.77 1.11 2.81 0.787 5 –
disorders disorders
Severe No severe HR 1.52 1.16 1.98 <0.001 4 –
depression depression
Alcohol or No alcohol or HR 1.44 1.08 1.91 <0.001 4 –
substance abuse subatance abuse
Anxiety No anxiety HR 1.02 0.97 1.07 0.363 8 0.773
symptoms symptoms

337
X. Li et al. Journal of Affective Disorders 302 (2022) 332–351

*
Egger’s test was conducted while eight and over studies were included for meta-analysis of each predictor.

Fig. 2. Pooled estimates with 95% CI for associated factors of suicide related behaviors (suicidal ideation, suicide attempts and suicide, respectively) in people with
major depressive disorders (A: Suicidal ideation; A1: Recurrent depression; A2: Previous suicidal behaviors; A3: Negative school factors; A4: Living alone; A5: Severe
depression; A6: Help-seeking; A7: Negative family factors; A8: Alcohol use related disorders; A9: Stressful life events; A10: Positive school factors; B: Suicide at­
tempts; B1: Substance misuse; B2: Previous suicidal behaviors; B3: Anxiety disorders; B4: Negative school factors; B5: Alcohol use related disorders; B6: Severe
depression; B7: Negative family factors; B8: Stressful life events; B9: Hopelessness; B10: Male; B11: Positive school factors; B12: Anxiety symptoms; B13: Anxiety
disorders; C: Suicide; C1: Male;

338
X. Li et al. Journal of Affective Disorders 302 (2022) 332–351

Fig. 2. (continued).

339
X. Li et al. Journal of Affective Disorders 302 (2022) 332–351

Fig. 2. (continued).

340
X. Li et al. Journal of Affective Disorders 302 (2022) 332–351

Fig. 2. (continued).

341
X. Li et al. Journal of Affective Disorders 302 (2022) 332–351

Fig. 2. (continued).

342
X. Li et al. Journal of Affective Disorders 302 (2022) 332–351

Fig. 2. (continued).

(Symptom-oriented Interview Schedule for Children), SSI (the Scale for were constructed to examine the potential publication bias.
Suicide Ideation), SSI-B (Beck Scale for Suicide Ideation) or self-reports. Egger’s test (α=0.05) was conducted to test publication bias if the
Suicide death was ascertained by ICD codes, death registration or number of included studies for meta-analysis of predictors were 10 and
reported by others. over.

2.6. Quality assessment (NOS criteria) 3. Results

The quality of selected studies was assessed by Newcastle–Ottawa 3.1. Characteristics of included studies
Scale (NOS) (Stang, 2010). It consisted of 8 items which was divided into
3 dimensions: (1) Selection of the study (4 items); (2) Comparability (1 Overall, 6750 articles were identified based on the key words and
item); (3) Outcomes for cohort studies (3 items). The total score of NOS search strategies and 6696 articles were excluded after title and abstract
is 9. According to the NOS criteria, studies with the scores of 6 and over screening due to ineligible research outcomes and study designs.
were regarded as high quality and were included in this study. Articles Totally, 54 cohort articles were collected for full text review. 13 articles
with the NOS scores less than 6 were recognized as low quality and were were excluded due to the low NOS scores (NOS<6) and 17 articles on
excluded. general population with depressive symptoms were also excluded.
Finally, 24 cohort studies covering 954,822 participants with MDD
2.7. Data extraction without age and sex restrictions were selected for systematic review and
only 15 studies were included for further meta-analysis based on the
After the full text reading and NOS assessment, data were extracted availability of relevant data. The publication periods ranged from 1993
from the qualified reports and evaluated by two independent in­ to 2020. The sample size ranged from 22 to 885,967. The follow-up
vestigators (XML & DBL). The extracted information included authors, periods ranged from 1 year to 50 years and the median was 5 years.
published year, countries, languages, study sample sizes, inventories, Seven studies were conducted in United States (29%), followed by five in
and predictors. If there were any disagreements between the two in­ Finland (21%), two in South Korea (8%), Sweden (8%) and Denmark
vestigators, the third person’s (YL) suggestions would be considered. (8%), respectively. One study was from the Netherlands, Germany,
Canada, Taiwan, New Zealand and Australia, respectively. The PRISMA
2.8. Statistical analysis diagram is in Fig. 1. The basic information of all the included 24 cohort
studies is presented in Table 1.
The extracted data were analyzed using Stata 15.1 (StataCorp LLC).
In terms of heterogeneity test, if I2≤50%, a fixed effect model was used. 3.2. Predictors for suicide related behaviors in people with MDD
Subgroup analyses and sensitivity analyses of leave-one-out method
were performed to examine the sources of heterogeneity while I2>50%. Table 2 shows the results of predictive factors associated with suicide
Random effect model was used if I2>50%. RR (for suicidal ideation, related behaviors among people with MDD. The combined effects of
suicide attempt) or HR (for suicide attempt, suicide death) and their predictors for suicidal ideation were: recurrent depression (RR=3.65,
95% CIs were calculated separately for the pooled effects of predictors 95%CI:0.72–18.55), previous suicidal behaviors (RR=3.63, 95%CI:
for suicidal behaviors in people with MDD. Funnel plots of predictors 2.57–5.13), negative school factors (e.g., truancy, suspension)

343
X. Li et al. Journal of Affective Disorders 302 (2022) 332–351

Fig. 3. Funnel plots of associated factors with suicide related behaviors (suicidal ideation, suicide attempts and suicide, respectively) in people with major depressive
disorders (A: Suicidal ideation; A1: Recurrent depression; A2: Previous suicidal behaviors; A3: Negative school factors; A4: Living alone; A5: Severe depression; A6:
Help-seeking; A7: Negative family factors; A8: Alcohol use related disorders; A9: Stressful life events; A10: Positive school factors; B: Suicide attempts; B1: Substance
misuse; B2: Previous suicidal behaviors; B3: Anxiety disorders; B4: Negative school factors; B5: Alcohol use related disorders; B6: Severe depression; B7: Negative
family factors; B8: Stressful life events; B9: Hopelessness; B10: Male; B11: Positive school factors; B12: Anxiety symptoms; B13: Anxiety disorders; C: Suicide; C1:
Male; C2: Previous suicidal behaviors; C3: Anxiety disorders; C4: Severe depression; C5: Alcohol or substance abuse; C6: Anxiety symptoms).

344
X. Li et al. Journal of Affective Disorders 302 (2022) 332–351

Fig. 3. (continued).

345
X. Li et al. Journal of Affective Disorders 302 (2022) 332–351

Fig. 3. (continued).

346
X. Li et al. Journal of Affective Disorders 302 (2022) 332–351

Fig. 3. (continued).

(RR=2.13, 95%CI: 1.70–2.69), living alone (RR=1.76, 95%CI: (HR=2.10, 95%CI: 1.92–2.30), anxiety disorders (HR=1.77, 95%CI:
1.24–2.50), severe depression (RR=1.68, 95%CI: 1.17–2.43), help- 1.11–2.81), severe depression (HR=1.52, 95%CI: 1.16–1.98), alcohol or
seeking (e.g., use of professional care) (RR=1.65, 95%CI: 1.01–2.69), substance abuse (HR=1.44, 95%CI: 1.08–1.91) and anxiety symptoms
negative family factors (e.g., childhood sexual abuse, parents’ remar­ (HR=1.02, 95%CI: 0.97–1.07). The forest plots were displayed in Fig. 2
riage, etc.) (RR=1.43, 95%CI: 1.28–1.59), alcohol use related disorders (C1~C6).
(RR=1.37, 95%CI: 0.70–2.68), stressful life events (e.g., loss, victimi­ Positive school factors (pooled RR=0.66, 95%CI: 0.36–1.22)
zation, financial stress, etc.) (RR=1.15, 95%CI: 1.07–1.24) and positive appeared to be protective for suicidal ideation, and male sex (pooled
school factors (e.g., school certificate passes, school enjoyment) RR=0.45, 95%CI: 0.15–1.33) was associated with a lower risk of suicide
(RR=0.66, 95%CI: 0.36–1.22). The forest plots were displayed in Fig. 2 attempt. However, the associations were not statistically significant.
(A1~A10).
The combined effects of the predictors for suicide attempts were:
3.3. Sensitivity analysis and subgroup analysis
substance misuse (RR=9.29, 95%CI: 3.60–23.99), previous suicidal
behaviors (RR=6.38, 95%CI: 2.97–13.72), anxiety disorders (RR=4.08,
For the heterogeneity test results of I2>50%, we conducted the leave-
95%CI: 1.71–9.72), negative school factors (e.g., truancy, suspension)
one-out sensitivity analysis. However, we did not find any study that
(RR=3.87, 95%CI: 2.61–5.75), alcohol use related disorders (RR=3.63,
significantly affected the overall meta-analysis results in each group.
95%CI: 2.03–6.50), severe depression (RR=2.45, 95%CI: 0.88–6.85),
Therefore, random effect models were used for the pooled effects
negative family factors (e.g., childhood sexual abuse, parents’ remar­
calculation. The number of included studies was too small. As such,
riage, etc.) (RR=1.65, 95%CI: 1.42–1.93), stressful life events (e.g., loss,
subgroup analysis was not conducted.
victimization, financial stress, etc.) (RR=1.34, 95%CI: 1.12–1.60),
hopelessness (RR=1.06, 95%CI: 0.97–1.16), male (RR=0.99, 95%CI:
0.40–2.47), positive school factors (e.g., school certificate passes, school 3.4. Funnel plots and Egger’s tests for publication bias
enjoyment) (RR=0.45, 95%CI: 0.15–1.33), anxiety symptoms
(HR=1.03, 95%CI: 1.00–1.05) and anxiety disorders (HR=1.03, 95%CI: Fig. 3 displays the publication bias for each predictive factor. The
0.79–1.34). The forest plots were displayed in Fig. 2 (B1~B13). data showed that there was publication bias associated with recurrent
The combined effects of the predictors for suicide death were: being depression and stressful life events for suicidal ideation, negative school
male (HR=3.27, 95%CI: 1.55–6.90), previous suicidal behaviors factors, positive school factors and negative family factors for suicidal
ideation and attempts, previous suicidal behaviors for suicide, anxiety

347
X. Li et al. Journal of Affective Disorders 302 (2022) 332–351

Fig. 3. (continued).

symptoms and disorders for suicide and suicide attempts. The publica­ 2018). The long-term suicide risk for medium and severe depression was
tion bias may affect the reliability of the results. 3.1% and 11.4%, respectively (Brådvik et al., 2008). Severe depression
According to Egger’s test, there was publication bias associated with patients have more significant changes in pathology such as
negative family factors for suicidal ideation (p = 0.002) and suicide hypothalamic-pituitary-adrenal axis changes, structural and functional
attempts (p = 0.001) (Table 2). Egger’s test found no publication bias brain changes and so on, which may contribute to suicide attempt
associated with predictors of severe depression (p = 0.247) and stressful (Malhi and Mann, 2018). People who have previous suicidal behaviors
life events (p = 0.122) for suicidal ideation, predictors of anxiety were also at high risk to perform future suicide attempts and committed
symptoms (p = 0.438) and stressful life events (p = 0.225) for suicide suicide (Aaltonen et al., 2019; Holma et al., 2010; May et al., 2012;
attempts, predictor of anxiety symptoms (p = 0.773) for suicide death Miranda and Shaffer, 2013; Seo et al., 2015). Individuals’ suicidal
(Table 2). thoughts or behaviors would get strengthened in MDD patients’ mem­
ory, reduce the threshold of suicide and then lead to suicide action
4. Discussion (Rudd, 2006; Beck, 1996). Besides, cognitive inflexibility, defined as
failing to handle changes in the external environmental properly, could
4.1. Main findings explain why people attempting suicide before seemed to be more
vulnerable to commit suicide again in the future (Schotte and Clum,
This study presented the predictive factors for suicidal ideation, 1982, 1987; Miranda et al., 2012).
suicide attempt and suicide death in individuals with MDD. Severe MDD is strongly associated with environmental factors, such as
depression, previous suicidal behaviors and alcohol/substance disorders stressful life events (victims of crime, having trouble with boss or co­
were common risk factors for suicidal ideation, suicide attempt and workers, interpersonal stresses, etc.) from family context (family history
suicide death. Individuals with MDD who had suffered from negative of suicide, childhood sex abuse, etc.) and school context (bad interper­
environmental factors, such as stressful life events, negative school sonal affiliations, bad school achievement, etc.) (Wang et al., 2015;
factors and negative family factors, were more likely to have suicidal Fergusson et al., 2003). As one of the stressful life events, poor inter­
ideation and attempt suicide. Furthermore, comorbid anxiety was a personal relationship indicates low social support which is closely
predictor for suicide attempt and suicide death in MDD individuals. associated with suicide related behaviors in MDD individuals (Lew­
Severe depression should be seen as a syndrome or a degree of insohn et al., 1997). Specific genetic factors make some people more
impairment and it was a strong predictor for suicide (Handley et al., vulnerable to stressful life events and more susceptible to depression

348
X. Li et al. Journal of Affective Disorders 302 (2022) 332–351

Fig. 3. (continued).

(Kendler et al., 1999). Furthermore, MDD patients who have financial comorbidity) deserve further research. Environmental factors such as
stress were prone to suffer from treatment resistant, which might in­ school and family experiences and lifetime grown-up environment
crease the risk of suicidal attempts (Amital et al., 2008; Stegenga et al., context need to be taken into account for the prevention of suicide in
2012). The diversity of individual’s reaction to stressful life events and clinical practice. Effective school and family supports might play an
behavior could explain part of the correlations (Wang et al., 2015). important role in prevention against future suicide death among people
Alcohol use related disorders should be taken into account in the with MDD.
assessment of suicide risk in MDD inpatients, which has important sig­
nificance for clinical guidance (Hansen et al., 2003). The purpose of 4.2. Limitations
drinking is to pursue the pleasant experience. Besides, when people feel
bad, drinking may be a kind of adjustment method. However, when an There were several limitations in this study. First, the literature da­
alcohol dependent failed to acquire enough alcohol they need, he/she tabases were limited, although we had searched different English and
would resort to make some suicidal related behaviors (Ribeiro et al., Chinese databases regardless of published periods and language re­
2016). By reducing natural anxiety towards self-mutilation or affecting strictions. Second, diverse diagnostic criteria had been used for the
drug metabolism, alcohol use related disorders could still increase sui­ definition of suicide related behaviors and MDD. This could lead to high
cide risk in depressed patients characterized by low mood (Boenisch and severe heterogeneity which would impact the reliability of results in
et al., 2010). Anxiety comorbidity could also increase the risk of suicide this study. Third, there were no enough published studies with high
behaviors. MDD patients have deficits in basic attentional level. On the quality for subgroup analysis which should be conducted to explore the
another hand, comorbid anxiety could cause impairments in higher level potential sources of heterogeneity, such as published year, age of par­
executive attention functions among individuals with MDD (Lyche et al., ticipants, follow-up periods, diagnosis criteria of MDD and diagnosis
2011). Anxiety disorders in MDD patients is important for future criteria of suicide related behaviors. Finally, gender played an important
research in suicide prevention. role in suicidal related behaviors among people with MDD. However,
Predictors indentified in this study should be considered while only two articles for suicide attempts and three articles for suicide death
developing suicide risk prediction models and risk assessment tools for were included on gender differences in this meta-analysis. More re­
estimating future suicide in people with MDD (Hawton et al., 2013). The searches should be conducted to display the exact correlations between
common risk factors for the three outcomes of suicided related behaviors sex and suicide related behaviors among people with MDD.
(e.g. previous suicidal behaviors, environmental factors and

349
X. Li et al. Journal of Affective Disorders 302 (2022) 332–351

Fig. 3. (continued).

5. Conclusions MDD based on these predictive factors.

Previous suicide behaviors, mental disorders (severe depression, Ethical standards

anxiety, alcohol/substance disorders) and environmental factors (school
and family context, social support, stressful life events, etc.) could The manuscript does not contain animals, patients, or population
contribute to suicide related behaviors in people with MDD. Men data.
suffering from MDD were at higher risk of suicide death than women.
The screening and identification of suicide risk individuals among Author contributions
people with MDD would be extremely meaningful for population health.
Due to the highly complex course of suicide, it is important to develop YL & JLW contributed to the study design. XL, FM, DL & JZ did
theoretical frameworks for suicide interventions among individuals with literature searches and screening. XL, ML & SY extracted the data. XL &

350
X. Li et al. Journal of Affective Disorders 302 (2022) 332–351

FM analyzed the data. XL, DL, FM & JZ wrote this manuscript and they Kendler, K.S., Karkowski, L.M., Prescott, C.A., 1999. The assessment of dependence in
the study of stressful life events: validation using a twin design. Psychol. Med. 29 (6),
contributed equally to this study. YL & JLW were responsible for English
1455–1460.
revision of the manuscript. All authors reviewed the manuscript. Kleiman, E.M., Nock, M.K., 2018. Real-time assessment of suicidal thoughts and
behaviors. Curr. Opin. Psychol. 22, 33–37.
Declaration of Competing Interest Large, M.M., 2018. The role of prediction in suicide prevention. Dialogues Clin. Neurosci.
20 (3), 197–205.
Lewinsohn, P.M., Gotlib, I.H., Seeley, J.R., 1997. Depression-related psychosocial
The authors declared no conflicts of interests. variables: are they specific to depression in adolescents? J. Abnorm. Psychol. 106
(3), 365–375.
Lyche, P., Jonassen, R., Stiles, T.C., Ulleberg, P., Landrø, N.I., 2011. Attentional functions
Acknowledgements in major depressive disorders with and without comorbid anxiety. Arch. Clin.
Neuropsychol 26 (1), 38–47.
This study was funded by the Natural Science Foundation (Grant Malhi, G.S., Mann, J.J., 2018. Depression. Lancet 392 (10161), 2299–2312.
Maniam, T., Marhani, M., Firdaus, M., Kadir, A.B., Mazni, M.J., Azizul, A., Salina, A.A.,
number: 81901391), Natural Science Foundation of Shandong Province, Fadzillah, A.R., Nurashikin, I., Ang, K.T., Jasvindar, K., Noor Ani, A., 2014. Risk
China (Grant number: ZR2019MH095), Research Fund for Lin He’s factors for suicidal ideation, plans and attempts in Malaysia–results of an
Academician Workstation of New Medicine and Clinical Translation in epidemiological survey. Compr. Psychiatry 55, S121–S125.
May, A.M., Klonsky, E.D., Klein, D.N., 2012. Predicting future suicide attempts among
Jining Medical University (Grant number: JYHL2019MS05, depressed suicide ideators: a 10-year longitudinal study. J. Psychiatr. Res. 46 (7),
JYHL2018ZD01), Key Research and Development Plan of Jining (Grant 946–952.
number: 2019SMNS033) and Taishan Scholars Program of Shandong Miranda, R., Gallagher, M., Bauchner, B., Vaysman, R., Marroquín, B., 2012. Cognitive
inflexibility as a prospective predictor of suicidal ideation among young adults with
Province (Grant number: tsqn201909145). All these funders had no role
a suicide attempt history. Depress. Anxiety 29 (3), 180–186.
in the design and conduction of this study. Miranda, R., Shaffer, D., 2013. Understanding the suicidal moment in adolescence. Ann.
N. Y. Acad. Sci. 1304, 14–21.
References Moher, D., Liberati, A., Tetzlaff, J., Altman, D.G., 2009. Preferred reporting items for
systematic reviews and meta-analyses: the PRISMA statement. PLoS Med 6 (7),
e1000097.
Aaltonen, K.I., Isometsa, E., Sund, R., Pirkola, S., 2019. Risk factors for suicide in Nock, M.K., Borges, G., Bromet, E.J., Alonso, J., Angermeyer, M., Beautrais, A.,
depression in Finland: first-hospitalized patients followed up to 24 years. Acta Bruffaerts, R., Chiu, W.T., de Girolamo, G., Gluzman, S., de Graaf, R., Gureje, O.,
Psychiatr. Scand. 139 (2), 154–163. Haro, J.M., Huang, Y., Karam, E., Kessler, R.C., Lepine, J.P., Levinson, D., Medina-
Amital, D., Fostick, L., Silberman, A., Beckman, M., Spivak, B., 2008. Serious life events Mora, M.E., Ono, Y., Posads-Villa, J., Williams, D., 2008a. Cross-national prevalence
among resistant and non-resistant MDD patients. J. Affect. Disord. 110 (3), 260–264. and risk factors for suicidal ideation, plans and attempts. Br. J. Psychiatry 192 (2),
Beck, A.T., 1996. Beyond belief: a theory of modes, personality, and psychopathology. In: 98–105.
Salkovskis, P (Ed.), Frontiers of Cognitive Therapy. Guilford Press, New York, Nock, M.K., Borges, G., Bromet, E.J., Cha, C.B., Kessler, R.C., Lee, S., 2008. Suicide and
pp. 1–25. suicidal behavior. Epidemiol. Rev. 30, 133–154.
Boenisch, S., Bramesfeld, A., Mergl, R., Havers, I., Althaus, D., Lehfeld, H., Niklewski, G., Nordentoft, M., 2011. Crucial elements in suicide prevention strategies. Prog.
Hegerl, U., 2010. The role of alcohol use disorder and alcohol consumption in suicide Neuropsychopharmacol. Biol. Psychiatry 35 (4), 848–853.
attempts–a secondary analysis of 1921 suicide attempts. Eur. Psychiatry 25 (7), Ribeiro, D.B., Terra, M.G., Soccol, K.L., Schneider, J.F., Camillo, L.A., Plein, F.A., 2016.
414–420. Reasons for attempting suicide among men who use alcohol and other drugs. Rev.
Bradvik, L., Mattisson, C., Bogren, M., Nettelbladt, P., 2008. Long-term suicide risk of Gaucha. Enferm. 37 (1), e54896.
depression in the Lundby cohort 1947-1997–severity and gender. Acta Psychiatr. Rudd, M.D., 2006. Fluid vulnerability theory: a cognitive approach to understanding the
Scand. 117 (3), 185–191. process of acute and chronic suicide risk. In: Ellis, TE (Ed.), Cognition and Suicide:
Braun, C., Bschor, T., Franklin, J., Baethge, C., 2016. Suicides and suicide attempts Theory, Research, and Therapy. America Psychological Association, Washington,
during long-term treatment with antidepressants: a meta-analysis of 29 placebo- DC, pp. 335–368.
controlled studies including 6934 patients with major depressive disorder. Seo, H.J., Song, H.R., Yim, H.W., Kim, J.B., Lee, M.S., Kim, J.M., Jun, T.Y., 2015. Age-
Psychother Psychosom 85 (3), 171–179. related differences in suicidality between young people and older adults with
Coentre, R., Talina, M.C., Góis, C., Figueira, M.L., 2017. Depressive symptoms and depression: data from a nationwide depression cohort study in Korea (the
suicidal behavior after first-episode psychosis: a comprehensive systematic review. CRESCEND study). Compr. Psychiatry 56, 85–92.
Psychiat. Res. 253, 240–248. Schotte, D.E., Clum, G.A., 1982. Suicide ideation in a college population: a test of a
Dong, M., Zeng, L.N., Lu, L., Li, X.H., Ungvari, G.S., Ng, C.H., Chow, I., Zhang, L., model. J. Consult. Clin. Psychol. 50 (5), 690–696.
Zhou, Y., Xiang, Y.T., 2019. Prevalence of suicide attempt in individuals with major Schotte, D.E., Clum, G.A., 1987. Problem-solving skills in suicidal psychiatric patients.
depressive disorder: a meta-analysis of observational surveys. Psychol. Med. 49 (10), J. Consult. Clin. Psychol. 55 (1), 49–54.
1691–1704. Spijker, J., de Graaf, R., Ten, H.M., Nolen, W.A., Speckens, A., 2010. Predictors of
Fergusson, D.M., Beautrais, A.L., Horwood, L.J., 2003. Vulnerability and resiliency to suicidality in depressive spectrum disorders in the general population: results of the
suicidal behaviours in young people. Psychol. Med. 33 (1), 61–73. Netherlands Mental Health Survey and Incidence Study. Soc. Psychiatry Psychiatr.
Gournellis, R., Tournikioti, K., Touloumi, G., Thomadakis, C., Michalopoulou, P.G., Epidemiol. 45 (5), 513–521.
Christodoulou, C., Papadopoulou, A., Douzenis, A., 2018. Psychotic (delusional) Stang, A., 2010. Critical evaluation of the Newcastle-Ottawa scale for the assessment of
depression and suicidal attempts: a systematic review and meta-analysis. Acta the quality of nonrandomized studies in meta-analyses. Eur. J. Epidemiol. 25 (9),
Psychiatr. Scand. 137 (1), 18–29. 603–605.
Handley, T., Rich, J., Davies, K., Lewin, T., Kelly, B., 2018. The challenges of predicting Stegenga, B.T., Nazareth, I., Grobbee, D.E., Torres-González, F., Svab, I., Maaroos, H.I.,
suicidal thoughts and behaviours in a sample of rural australians with depression. Xavier, M., Saldivia, S., Bottomley, C., King, M., Geerlings, M.I., 2012. Recent life
Int. J. Environ. Res. Public Health 15 (5), 928. events pose greatest risk for onset of major depressive disorder during mid-life.
Hansen, P.E., Wang, A.G., Stage, K.B., Kragh-Sorensen, P., 2003. Comorbid personality J. Affect. Disord. 136 (3), 505–513.
disorder predicts suicide after major depression: a 10-year follow-up. Acta Psychiatr. Sudol, K., Mann, J.J., 2017. Biomarkers of suicide attempt behavior: towards a biological
Scand. 107 (6), 436–440. model of risk. Curr. Psychiat. Rep. 19 (6), 31.
Harris, E.C., Barraclough, B., 1997. Suicide as an outcome for mental disorders. A meta- Wang, Y., Sareen, J., Afifi, T.O., Bolton, S.L., Johnson, E.A., Bolton, J.M., 2015.
analysis. Br. J. Psychiatry 170, 205–228. A population-based longitudinal study of recent stressful life events as risk factors for
Hawton, K., Casañas, I.C.C., Haw, C., Saunders, K., 2013. Risk factors for suicide in suicidal behavior in major depressive disorder. Arch Suicide Res 19 (2), 202–217.
individuals with depression: a systematic review. J. Affect. Disord. 147 (1–3), 17–28. Witte, T.K., Timmons, K.A., Fink, E., Smith, A.R., Joiner, T.E., 2009. Do major depressive
Hegerl, U., 2016. Prevention of suicidal behavior. Dialogues Clin. Neurosci. 18 (2), disorder and dysthymic disorder confer differential risk for suicide? J. Affect. Disord.
183–190. 115 (1–2), 69–78.
Holma, K.M., Melartin, T.K., Haukka, J., Holma, I.A., Sokero, T.P., Isometsa, E.T., 2010. Wong, J.C., 2018. Predicting suicide and its prevention. Ann. Acad. Med. Singap. 47 (9),
Incidence and predictors of suicide attempts in DSM-IV major depressive disorder: a 357–359.
five-year prospective study. Am. J. Psychiatry 167 (7), 801–808.

351