Risk factors for fatal and nonfatal repetition of suicide attempts: a literature review

Abstract

Objectives

This review aimed to identify the evidence for predictors of repetition of suicide attempts, and more specifically for subsequent completed suicide.

Methods

We conducted a literature search of PubMed and Embase between January 1, 1991 and December 31, 2009, and we excluded studies investigating only special populations (eg, male and female only, children and adolescents, elderly, a specific psychiatric disorder) and studies with sample size fewer than 50 patients.

Results

The strongest predictor of a repeated attempt is a previous attempt, followed by being a victim of sexual abuse, poor global functioning, having a psychiatric disorder, being on psychiatric treatment, depression, anxiety, and alcohol abuse or dependence. For other variables examined (Caucasian ethnicity, having a criminal record, having any mood disorders, bad family environment, and impulsivity) there are indications for a putative correlation as well. For completed suicide, the strongest predictors are older age, suicide ideation, and history of suicide attempt. Living alone, male sex, and alcohol abuse are weakly predictive with a positive correlation (but sustained by very scarce data) for poor impulsivity and a somatic diagnosis.

Conclusion

It is difficult to find predictors for repetition of nonfatal suicide attempts, and even more difficult to identify predictors of completed suicide. Suicide ideation and alcohol or substance abuse/dependence, which are, along with depression, the most consistent predictors for initial nonfatal attempt and suicide, are not consistently reported to be very strong predictors for nonfatal repetition.

Keywords:

• suicide
• deliberate self-harm
• suicide attempt
• repetition
• predictors

Introduction

In recent years, suicide-attempt (SA) rates have been widely studied. A World Health Organization community survey reported the lifetime prevalence of SAs at 0.4%–4.2%.¹ Female sex, young age, marital status (divorced or widowed), and having a personality disorder have been associated with an increased risk of attempting suicide.

The incidence rate for completed suicide (S) is 11.2/100,000,² increases with age, and is three times higher in males than in females.³ Suicide accounts for about 1% of all deaths and is the ninth-leading cause of death in the US and the third in ages 15–24 years.^3,4 Rates in Caucasians are twice those of non-Caucasian populations, and married people are less likely than single, divorced, or widowed to commit suicide.² For those in bereavement, the risk is higher in the first year after loss. Rates are higher in Protestants (31.4/100,000) than in Catholics (10.9/100,000) or Jews (15.5/100,000). Unemployment increases the rate of suicide by 50%.³

The suicide risk is higher in psychiatric patients compared to nonpsychiatric populations. More specifically, lifetime risk of suicide has been reported as high as 15% for affective disorders, 10% for schizophrenia, and 2%–3% for alcohol abuse.⁴ With respect to affective disorders, the risk is higher with increasing severity of depression. Suicides occur more often in patients with a family history of suicide, mood disorders, and alcohol abuse.⁴ Suicidality tends to emerge early in the course of a mood disorder, and increases in association with melancholia and agitation.⁴

Despite these many variables having been associated with suicidal behavior, their usefulness in predicting future suicidal behavior remains undemonstrated. The prospective prediction of later suicide remains difficult.^5,6 A need exists, as underlined by Hughes and Owens,⁶ for more effective monitoring of people who contact hospitals because of SAs, and for more information on patients who carry out SAs but do not attend hospital. The likelihood of a repeated attempt after a first SA has been investigated less extensively. An episode of self-harm is a strong predictor of later suicide, with the risk peaking in the first 6 months after a self-harming episode, but risk persists for many decades.

A recent review estimated the 1-year incidence of repetition at 16% and fatal repetition at 2% of attempters.⁷ After 9 years, the suicide-fatality rate increased to more than 5%. Both fatal and nonfatal repetition rates were reported to be lower in Mediterranean than in Northern European countries.⁸

However, despite the potential importance of studies investigating the risk factors involved in repetition of SA, no systematic reviews of the issue have been reported. Accordingly, the aim of this review was to identify the evidence for predictors of repetition of SA, and more specifically for subsequent S.

Methods

One of the authors (MB) searched both PubMed and Embase systematically for studies carried out between January 1, 1991 and 31 December, 2009 in English, using the keywords repetition/repeated suicide attempt, repetition/repeated self-harm, recurrence/recurrent self-harm, recurrence/recurrent suicide attempt, repetition/repeated self-poisoning, and recurrence/recurrent self-poisoning. Suicides in most primary studies included those that were definite (by verdict of a coroner or equivalent authority) or probable (open verdict or equivalent judgment); definitions were too variable for us to discriminate further, and we have included them all and used this broad definition of suicide. With the terms “suicide attempt” or “SA,” any nonfatal act in which the patient causes self-harm (self-mutilation, poisoning, jumping from high places, firearm shots, hanging, asphyxiation) was considered. The nomenclature has been taken from Silverman et al; we considered all suicidal acts, despite the degree of suicidal intent. With the term “SA” we mean a not-completed suicide (with or without injuries), while with the term “S” we mean a completed suicide.^9,10

For the aim of this study, we included cohort studies, case-control studies, and cross-sectional studies. Since our review focused on environmental risk factors and not on management, we excluded studies investigating self-harm management and/or care. Moreover, some of these studies investigated selected populations at risk, and others had very small samples. Thus, we decided to exclude studies investigating selected populations (childhood/adolescence, elderly, males/females only, minorities only, patients with a specific personality disorder), studies with small samples (fewer than 50), or prospective studies with a follow-up shorter than 6 months. We decided to exclude special populations because the aim of the review was the prospective prediction of later suicide in the whole population referring to the emergency room.

Studies on self-poisoning only were included because the self-poisoning method encompasses 80% of females’ and 64% of males’ SAs.⁸ For the same reason, we decided to include studies on adults only.

Data extraction

After a first screening, 211 papers satisfied our inclusion criteria. Six were useful for the introduction and for the discussion. The majority of them were carried out in Europe. Since designs, the variables studied, and the length of follow-up were different among these studies, a formal meta-analysis or direct comparison was not possible. After excluding 53 studies for not fulfilling the scope of the review, five studies for small sample size, 72 studies for a selected sample (37 childhood/adolescence, ten elderly, three females only, one males only, 19 patients with a psychiatric diagnosis, two minorities), we were left with 76 studies: 13 (17%) with a cohort analysis, 45 (59%) with a case-control analysis, and 18 (24%) with a cross-sectional analysis (Figure 1).

Figure 1 Literature review.

Sixty of them (79%) were carried out in Europe, more specifically, 24 in the UK, 20 in Scandinavian countries, five in Ireland, two in France, one in Spain, one in the Netherlands, one in Belgium, and six in three or more countries. The other studies were done in the US (five), Australia (four), Canada (two), the People’s Republic of China (two), Iran (one), Brazil (one), and Uganda (one). SA was investigated in 64, while S was investigated in 18 studies.

All the risk factors investigated in the studies were inserted and then selected by a consensus-based process (by all the authors).

Results

Nonfatal repetition

The strongest predictor for nonfatal repetition was a history of SA, a finding reported as significant in 13 of 16 multivariate analyses and 13 of 14 univariate analyses (Table 1). Also significant were being a victim of a sexual abuse (multivariate 5/9, univariate 4/5), poor global functioning (multivariate 3/4, univariate 4/4), having a psychiatric disorder (multivariate 5/11, univariate 6/11), undergoing psychiatric treatment (multivariate 2/7, univariate 7/8), depression (multivariate 3/10, univariate 8/12), anxiety (multivariate 2/6, univariate 4/5), or alcohol abuse or dependence (multivariate 4/10, univariate 4/8). There were weaker associations for having a personality disorder, repetition for young adult age, unmarried status, alcohol abuse or dependence, psychiatric morbidity or treatment, and unemployment status. For many variables (Caucasian ethnicity, having a criminal record, having any mood disorders, bad family environment, and impulsivity) there are indications for a correlation, but data are very scarce. The results of analyses are in Table 1.

Table 1 Summary of available factors correlated with suicidality

Variable	Number of risk-factor-significant references in multivariate analysis (significant/total) (S)	Number of risk-factor-significant references in multivariate analysis (significant/total) (SA)	Number of risk-factor-significant references in univariate analysis (significant/total) (S)	Number of risk-factor-significant references in univariate analysis (significant/total) (SA)
Age	9 older/16	3 medium age/23 3 young/23	2 older/5	3 older/28 7 young-medium age/28
Sex	4 M/15	4 F/22 3 M/22	1 M/5	4 F/26 1 M/26
Ethnicity	0/1	1 Caucasian/3	0/1	2 Caucasian/4
Marital status	1 married or widowed/5	2 not married/182 divorced/18	1 not married/2	2 divorced/22 4 not married/22 2 married/22
Employment status	3 not employed/7	3 not employed/14	0/4	9 unemployed/18 1 not student/18 1 retired/18
Social class	0/2	0/5	0/1	1 low/4
Education	0/2	1 low/10		2 low/9
Housing status	2 alone/4	0/8	1 alone/2	3 alone/7
Psychiatric morbidity	2/7	5/11	1/4	6/11
Psychiatric treatment	1/3	2/7	0/2	7/8
Psychosis	2/7 1/7 rev	2/7	1/2	2/7
Mood disorders	2/3	3/4		0/1
Depression	0/5	3/10	1/2	8/12
Anxiety	0/3	2/6	0/1	4/5
Alcohol misuse	2/9	4/10	1/4	4/8
Substance misuse	1/7	0/7	0/3	2/8
Somatic diagnosis	3/4	1/41 rev/4	1/2	1/3
Family history of S	0/1	1/2		2/4
Personality disorders	0/7	2/6 5/9		3/6 4/5
History of sexual abuse	7/11 1 rev/11	13/16	1/5	13/14
Method	1 less poison/4	2 penetrating/9	0/2	2 more poison/8
Suicidal ideation	5/9	1/13 1 rev/13	1/2	6/14
Hopelessness	0/2	2/8		5/6
Circumstances	0/3	0/6	0/2	2/8
Family environment		1 bad/1		2 bad/3
Criminal record		1/4		2/3
Global functioning	1 low/2	3 low/4	1 low/1	4 low/4
Impulsivity	1 rev/1	1/1	1 rev/1	2/2

Abbreviations: F, female; M, male; SA, suicide attempt(s); rev, reverse correlation; S, completed suicide.

Completed suicide

The strongest predictors of S are older age (multivariate 9/16, univariate 2/5), a high suicide ideation (multivariate 5/9, univariate 1/2), a history of SA (multivariate 7/11, univariate 1/5). Living alone, male sex, and alcohol abuse are weaker predictors. There is a correlation (but supported by very scarce data) for poor impulsivity and having a somatic diagnosis. There are no data available for sexual and physical abuse during childhood or for the family environment. The syntheses of the available results are in Table 2.

Table 2 Factors correlated with suicide attempts (SAs) and completed suicide (S)

Study	Year	Design	Number of patients	Factors significantly associated with SA
Appleby et al^Citation3	1999	Case-control (S)	149 P, 149 C	Care reduced, history of SA, suicidal thoughts during aftercare, most recent admission at first illness
Asnis et al^Citation11	1993	Cross-sectional	74 repeaters, 90 first attempters	No variables associated
Batt et al^Citation12	1998	Cross-sectional	158 multirepeaters, 164 first repeaters, 310 first attempters	Married status, not a student, not living alone, alcohol dependence, anxiety
Bille-Brahe and Jessen^Citation13	1994	Case-control	773	Self-poisoning, living alone, less hanging/cutting, divorced, unemployment
Boyes^Citation14	1994	5-year follow-up, retrospective cohort (male, female)	1,597	No differences
Brådvik^Citation15	2003	Case-control (S)	98 S, 89 C	No differences
Brezo et al^Citation16	2008	Case-control (previous SA)	Not known	History of SA, compulsivity, anxiety
Carter et al^Citation17	1999	Case-control	1,238	Female sex, single or divorced/widowed/separated, retired, age classes 25–34 or 35–44 years, length of stay
Carter et al^Citation18	2002	Case-control	1,317	History of SA, personality disorders, low social class
Carter et al^Citation19	2005	Case-control (S)	31 P, 93 C	Increased number of drugs ingested, increased dose ingested, drug/alcohol abuse/dependence
Cedereke and Ojehagen^Citation20	2005	Case-control	178	History of SA, history of psychiatric treatment, lower global functioning, suicide ideation
Chandrasekaran and Gnanaselane^Citation21	2008	Case-control	293 (67 repeated)	Hopelessness, history of psychiatric treatment, major depression, lower global functioning
Christiansen and Jensen^Citation22	2007	Case-control (S and SA)	2,614 P, 39,210 C	SA method (self-poisoning), age class 15–24 years, psychiatric morbidity, S, history of SA, method (jumping from high places), ages 15–24 and 25–59 years
Coakley et al^Citation23	1994	Cross-sectional	122 repeaters, 179 first attempters	Older age, history of depression, schizophrenia, or alcohol dependence
Colman et al^Citation24	2004	Case-control	369 (92 repeaters)	History of SA, history of depression, history of schizophrenia, and poor physical health
Conner et al^Citation25	2007	Case-control	277 P 277 C	Depression, acute stress, poor quality of life
Cooper et al^Citation26	2005	Case-control (S)	7,968	S: not living with a close relative, avoiding discovery, alcohol abuse/dependence, method (cutting), history of psychiatric treatment, physical health problem, high risk management at ER, history of SA
Cooper et al^Citation27	2006	5-year follow-up, prospective cohort (ethnicity)	299 South Asians, 6,884 Caucasians	Caucasian ethnicity
Corcoran et al^Citation28	2004	Case-control	1,256	History of SA, age class 45–49 years
Crane et al^Citation29	2007	Cross-sectional	323 repeaters, 285 first attempters	Suicidal ideation, male sex, depression, hopelessness
da Silva Cais et al^Citation30	2009	Cross-sectional	101 repeaters, 102 first attempters	Female sex, suicide ideation, unemployment status or being a housewife, history of emotional, physical, or sexual abuse, criminal record, hopelessness, depression
De Moore and Robertson^Citation31	1996	Case-control (S)	223	History of SA, planned attempt, narcotic overdose, mental illness (dementia, depression, psychosis)
Ekeberg et al^Citation32	1991	Case-control (S)	934	Older age, suicide ideation
Evans et al^Citation33	1996	Cross-sectional	185	Impulsiveness
Evans et al^Citation34	2000	Cross-sectional	421	No differences
Forman et al^Citation35	2004	Cross-sectional	114 repeaters, 39 first attempters	Unemployment, child emotional abuse, family mental illness, family suicide attempt, depression, hopelessness, psychosis, substance abuse, less problem-solving
Gilbody et al^Citation36	1997	Case-control	1,576 P (36% previous SA)	History of SA
Harriss et al^Citation37	2005	2–6 years’ follow-up, prospective cohort study (suicidal ideation) (S and SA)	2,489	S: suicidal ideation SA: no correlation
Haukka et al^Citation38	2008	Case-control (S and SA)	18,199	S: psychosis, mood disorders SA: female sex, age-group 30–40 years, any psychiatric disorder, alcohol abuse/dependence
Haw et al^Citation39	2003	12–16 months’ prospective cohort (suicidal ideation)	118	No correlation
Haw et al^Citation40	2007	Cross-sectional	4,167 (3 groups: first attempters, <4 episodes repeaters, ≥4 repeaters; M and F calculated separately)	M with ≥4 episodes more aged 25–34 years, history of SA, current psychiatric disorders, personality disorders, psychiatric treatment, alcohol/drug abuse, being a victim of violence, criminal record, history of SA less in nonrepeaters
Hawton et al^Citation41	2003	Case-control	150	Psychiatric disorders
Henriques et al^Citation42	2005	5–10 years’ follow-up, prospective cohort (suicidal ideation) (S)	393	Suicidal ideation
Heyerdhal et al^Citation43	2009	Case-control	2,032	Deep coma, age class 30–49 years, use of sedative drugs and opiate agents
Hjelmeland et al^Citation44	1998	Case-control (S and SA)	654	S: older age class SA: history of SA, lower suicidal intent
Hjelmeland and Polit^Citation45	1996	Case-control	1,012 (507 first attempters, 509 repeaters)	In first attempters: history of sexual abuse, psychiatric disorder In repeaters: alcohol abuse and suicide among relatives
Johnston et al^Citation46	2006	Case-control	4,743	History of SA, history of psychiatric treatment, employment status, unmarried, Caucasian ethnicity
Kapur et al^Citation47	2006	Case-control (S and SA)	7,723	S: longer period since the first act, male sex, older age, single status SA: age class 25–54 years, single status, Caucasian ethnicity, unemployed status, current or previous psychiatric treatment, history of SA, alcohol abuse/dependence, psychiatric diagnosis
Keeley et al^Citation48	2003	Case-control	2,287	Male sex, history of SA, dysfunctional family of origin, history of sexual abuse, criminal record
Kinyanda et al^Citation49	2005	Cross-sectional	25 repeaters, 75 first attempters	Single status, have children, live alone or with parents, sexual problems as major precipitant for SA, had more negative life events in childhood and fewer negative life events in the past year
Kiankhooy et al^Citation50	2009	Case-control	156	Self-inflicted injury, penetrating mechanism of injury, length of stay, male sex
Lilley et al^Citation51	2008	18-month prospective cohort (method)	7,344	Self-poisoning
McAuliffe et al^Citation52	2006	Case-control	836	Repeaters scored higher on the passive avoidance factor and on the negative expression factor and lower on the active handling factor on Utrecht coping list
McAuliffe et al^Citation53	2007	Cross-sectional (suicide ideation)	84 repeaters, 52 first attempters	No differences
McAuliffe et al^Citation54	2008	Case-control	152	Older age, history of SA
McEvedy^Citation55	1997	Case-control	628	Married status, older age
Neeleman et al^Citation56	1998	Cross-sectional	120 natural-cause death, 36 accidental death, 11 suicide death	Adolescent emotional instability, conduct problems
Nordentoft et al^Citation57	1993	Case-control (S)	974	Older age, living alone, history of SA, no respirator treatment
Nordstrom et al^Citation58	1995	5-year follow-up, prospective cohort	1,573	Male sex, older age female
Ojehagen et al^Citation59	1991	Cross-sectional	46 repeaters, 33 first attempters	Unemployed status, disability pension, psychiatric disorder, psychiatric treatment, psychiatric inpatient
Ostamo and Lönnqvist^Citation60	2001	Case-control (S)	2,782	Male sex, married, widowed or divorced status, older age class
Osváth et al^Citation61	2003	Cross-sectional	549 first attempters, 609 repeaters	Psychiatric disorders (personality disorders, mood disorders, and alcohol abuse), divorced status, unemployed status, low education status, age-groups 20–35 and 35–44 years
Owens et al^Citation62	1991	1-year follow-up, prospective cohort (admission)	687 admitted to a ward, 305 discharged from ER	No differences
Owens et al^Citation63	1994	Case-control	992	Ingestion of more than one drug, history of SA, age class 25–54 years, psychiatric disorder, unemployed status, psychiatric admission, expression of a threat to another person or written a note
Owens et al^Citation64	2005	Case-control (S)	1,091	Older age, male sex, impairment of consciousness, psychiatric disorders, admission during the daytime, discharge from accident and emergency after psychiatric assessment, no history of SA
Pettit et al^Citation65	2004	Case-control	123	Suicidal ideation, presuicidal crisis
Platt et al^Citation66	1992	Case-control	Not specified	Age class 15–34, female sex
Schmidtke et al^Citation8	1996	Cross-sectional	16,394 (repeaters 42% of M and 45% of F)	Divorced
Scoliers et al^Citation67	2009	Case-control	361	History of SA, female sex, age classes 20–29, 30–39, and 40–49 years, education, suicide ideation, medium Buglass and Horton risk, anxiety, depression, psychiatric disorder
Sheikholeslami et al^Citation68	2008	Cross-sectional	35 repeaters, 49 first attempters	Not married status, psychiatric disorder, personality disorder, depression, hopeless, suicide ideation, impulsiveness, less satisfaction, negative events
Sidley et al^Citation69	1999	Case-control	66	History of SA, hopelessness
Sinclair et al^Citation70	2007	Case-control	68	Sexual abuse, mood disorder
Stenager et al^Citation71	1994	Case-control (S and SA)	139	S: older age SA: no somatic diseases
Suokas et al^Citation72	2001	5-year follow-up, prospective cohort (history of SA) (S)	1,018 P	SA
Suominen et al^Citation73	2000	Retrospective cohort (personality disorders)	114 with personality disorders, 65 without	Personality disorder
Suominen et al^Citation74	2004	Case-control	224	Suicide ideation
Taylor et al^Citation75	1994	Cross-sectional	53 repeaters, 47 first attempters	Panic disorder, psychiatric disorder, history of SA, history of sexual abuse, PTSD
Tejedor et al^Citation76	1999	Case-control (S and SA)	150	S: low global functioning SA: history of SA, low global functioning
Tidemalm et al^Citation77	2008	Case-control (S and SA)	27,004	S: psychosis, mood disorder, depressive disorder, anxiety disorder, alcohol abuse/dependence, personality disorder SA: psychosis, mood disorder
Townsend et al^Citation78	2001	Case-control	1,719	Single drug used
Trémeau et al^Citation79	2005	Retrospective cohort (family history of S)	480	Family history of S
Verkes et al^Citation80	1997	1-year follow-up, prospective cohort (serotonergic parameters)	220	Serotonergic parameters
Wang and Mortensen^Citation81	2006	Case-control	48,000	Age class <40 years, being a newcomer, alcohol influence, suicide letter
Westling et al^Citation82	2004	Case-control (CSF leptin)	78	No differences
Ystgaard et al^Citation83	2004	Cross-sectional	58 repeaters, 77 first attempters	History of sexual abuse, history of physical abuse
Zahl and Hawton^Citation84	2004	Case-control (S and SA)	8,879 first attempters, 2,704 repeaters	S: history of SA SA: younger age, female sex

Abbreviations: M, male; F, female; PTSD, posttraumatic stress disorder; CSF, cerebrospinal fluid; S, completed suicide; SA, suicide attempt(s); P, patients; C, controls; ER, emergency room.

Discussion

At present, there is no psychological test, clinical technique, or biological marker sensitive and specific enough to predict either short-term suicide or repetition. In line with Appleby et al,³ there is a north–south gradient in the repetition rate of suicide. A study by Pokorny⁵ illustrates how a method to predict suicide based on recognized risk factors will not only lead to a better identification of individuals at risk but also to a higher number of lost-to-follow-up or undetected cases. In this study, the authors attempted to identify which of 4,800 consecutive patients would commit suicide. On the basis of 21 risk factors, they identified 803 patients having increased risk of suicide. Thirty of 803 (3.7%) committed suicide in a 5-year follow-up. None of these risk factors was detected in 37/67 suicides. These results are confirmed by a review of twelve studies conducted by Diekstra in 1985.⁸⁵ About 50% of suicidal people had committed at least one previous attempt. Also in this review, it is shown that it is easier to detect a nonfatal SA than a fatal one. This means that S is multifactorial, and involves not only medical but also philosophical aspects, eg, life is or is not worth living, and it is often a difficult but aware choice. The goal of a suicide assessment is not to predict suicide, but to place a person along a putative risk continuum to evaluate suicidality, especially in the period immediately following the attempt, and allow for a more informed intervention. In fact, according to Reulbach and Bleich,⁸⁶ up to 45% of people who deliberately harm themselves leave accident and emergency departments without receiving an adequate psychiatric assessment; after the discharge, the patients should not be lost in aftercare, especially if they suffer from depression, bipolar disorder, or schizophrenia.⁸⁶

In fact, after adjustment for baseline demographic and clinical characteristics and hospital differences, being referred for specialist follow-up was associated with reduction in repetition rate.⁸⁷

Synthesis of the available results

It is difficult to identify risk factors for repetition of nonfatal SA, and even more for repetition ending in S. The studies evaluated in this review had different designs and follow-up, so they are not comparable for a systematic review with meta-analysis of the available data. However, some intriguing results are available. Alcohol/substance abuse or dependence and suicide ideation, which are, along with depressed mood, the most consistent predictors for self-harm and suicide,⁴ do not seem as strong for nonfatal repetition. The presence of a previous SA is a more consistent finding for nonfatal repetition than for S, but it is the best risk factor for both and persists for many decades. The presence of a personality disorder, depression, sexual abuse in childhood, alcohol dependence, or unemployed or unmarried status are more consistently significant in nonfatal than in fatal SAs, while in nonrepeated SA, having a personality disorder increased rates among both fatal and nonfatal attempts.⁴ Impulsivity seems to be correlated with SA and inversely correlated with suicide completion. On the other hand, having a suicide ideation, (older) age, and (male) sex are thought to be more consistently found in fatal repetition, although the role of sex is not very clear. Female sex and younger age, in contrast with data on nonfatal SA, are not likely to predict repetition. This means that once a first SA has been made (an event more frequent in females and younger people), the risk for a second attempt does not appear increased in these two categories.

Other variables, such as family environment, problem-solving, and global functioning, have a positive correlation with fatal and/or nonfatal SA repetition, but data available are not sufficient to identify them as “predictors” of repetition. Further studies are needed to confirm this correlation.

Methodological pitfalls

Many other variables have been studied, eg, Caucasians commit suicide twice as frequently as other races, and Protestants are more likely to commit suicide than Catholics or Jews.⁴ A nonheterosexual orientation carries an increased risk for attempted but not for completed SAs.⁴ In all these cases, data on SA repetition are inconsistently reported.

Moreover, since a previous SA is the best risk factor for both fatal and nonfatal repetition, most findings presented here might not be specific to repetition. Only three studies in our group investigated the risk factors in first attempters for future attempts,^38,45,78 and only one⁴⁵ studied it prospectively.

According to Owens et al,⁷ the median proportion of patients repeating nonfatal SA is 16% at 1 year and 23% in studies lasting longer than 4 years. For a subsequent suicide, after a longer follow-up, the suicide rate increases from less than 2% at 1 year to more than 5% in studies lasting over 9 years. However, as most prospective studies lasted 1 year, the risk factors for S in subsequent years may differ from those detected at early follow-up.

Future perspectives

Further studies would ideally examine a well-defined inception cohort (ie, patients at time of first SA) identified and followed prospectively. A long-term follow-up (at least 4 years) is recommended. Standard definitions of risk and prognostic factors should be determined when planning the study. Interacting factors such as previous attempts or selected samples should be controlled for at the planning or the analysis stage. Some variables, like sexual child abuse, family environment, problem-solving, and global functioning, should be included, to evaluate their role for a repeated episode. Ideally, a study would compare different ethnicities and religions and investigate the differences in suicide repetition between immigrants and nonimmigrants. Sexual orientation should be investigated as well.

Conclusion

SA repetition (whether fatal or nonfatal) is a common event in developed countries. Prediction of recurrent SA in a patient who committed a first SA is an important task for the psychiatrist. However, it is hard to find independent predictors out of all the many variables associated with repeated and especially with S. Based on the available evidence, only a previous SA, depression, sexual abuse in childhood, and personality disorders have been found to predict nonfatal SA, while previous SA and older age were found to predict fatal SA. Suicidal ideation, which is one of the most consistent predictors for SA and S, does not seem as strong for repeated SA, while it remains consistent for S. In several cases, no apparent risk factor was detected, and it makes it difficult to prevent fatal and nonfatal attempts.

A large multicenter prospective investigation of first SAs should be undertaken comparing different countries and differing social and cultural backgrounds and settings within each country.

Acknowledgments

Jennifer Covino, PhD and Countney Clabby, PhD, for their help on the data collection, and Ettore Beghi, MD, for critically reviewing the study proposal.

Disclosure

The authors report no conflicts of interest in this work.