A systematic review of suicidal ideation and behaviors among adults with spinal cord injury

Abstract

Context

Suicide is a public health crisis within the United States. Individuals with spinal cord injury (SCI) are vulnerable to negative outcomes such as suicide.

Objective

This systematic review describes frequency of suicidal ideation (SI) and behaviors (suicide attempt [SA] and deaths) among samples of adults living with SCI. Associated risk and protective factors of SI and suicidal behaviors were also explored.

Methods

On July 7, 2021, OVID Medline, EMBASE, OVID PsycINFO, Web of Science Core Collection, CINAHL, Cochrane Library, and Google Scholar databases were searched for SI, SA, and deaths by suicide among adults with a history of SCI. Risk of bias (RoB) was assessed using the Effective Public Health Practice Project (EPHPP) quality assessment tool. Results were synthesized descriptively considering the likely impact of RoB. The updated Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 reporting method was used.

Results

Sixty-nine articles were included. Frequency of SI and SAs within study samples ranged from 4% to 67%, and 0% to approximately 66%, respectively. While assessment methods for SI and SA widely varied, suicide deaths data sources were more consistent. Studies assessing SI and SA generally had high RoB and the overall strength of evidence was low. Stronger observational study designs assessing death by suicide had low RoB and the overall strength of evidence was high. Very few studies examined protective factors.

Discussion

Findings support previous work. Quality of evidence, and therefore confidence in the frequency of these outcomes, was impacted by the use of assessment measures without sufficient psychometric properties. The field will benefit from use of psychometrically sound measures to assess for SI, SA and suicide deaths. Furthermore, additional work is needed to elucidate risk and protective factors and to clearly characterize samples to increase generalizability of findings to the larger SCI community, including standardized reporting of SCI characteristics (e.g. use of American Spinal Injury Association classification system). Ultimately, improved suicide assessment and risk management is expected to prevent suicide among those living with SCI.

Registration

CRD42020164686

Introduction

Suicide is a public health crisis in the United States (U.S.) ranking as the 10th leading cause of death and is a rising concern worldwide.^1,2 In the U.S. (2019), 1.4 million people reported making a suicide attempt (SA).³ In 2016, the World Health Organization (WHO) reported an annual global age-standardized suicide rate of 10.5 per 100,000.⁴ An estimated suicide death in the U.S. occurs once every 11 min.¹ While SAs are a strong predictor of eventual suicide, rates of suicidal ideation (SI) are also concerning. In fact, in 2018, 10.7 million individuals living in the U.S. endorsed serious SI, with 3.3 million Americans making a suicide plan. During the same time period, 1.4 million in the U.S. made a SA.⁵ Nock and colleagues interviewed adults (n = 84,850) across 17 countries and found that 9.2% endorsed lifetime SI, 3.1% reported a lifetime suicide plan, and 2.7% reported a prior SA.⁶ Across all countries, 60% of the transitions from SI to first SA occurred within the first year of SI onset.⁶

Individuals who have sustained a SCI are frequent consumers of healthcare, both with routine medical visits and re-hospitalizations.^7–10 Ongoing medical complications and hospital admissions can also take a toll on the quality of life and daily functioning of those living with a SCI.¹¹ Further, individuals with SCI have higher rates of physical concerns (e.g. disability, chronic medical conditions),¹² psychiatric disorders (e.g. depression, anxiety, posttraumatic stress disorder),¹³ and psychosocial challenges (e.g. social stress, isolation, quality of life),¹⁴ which are also factors associated with increased suicide risk.^15,16 The intersection of these negative outcomes may contribute to risk for suicide among individuals living with a SCI.

To our knowledge, the earliest review on suicide and SCI was embedded in a larger review encompassing suicide as an outcome for medical disorders.¹⁷ Harris and Barraclough conducted a MEDLINE search of nine medical disorders that were associated with possible suicide risk.¹⁷ The search strategy did not include the term “suicide” instead the term “mortality” was employed. Data regarding suicide after SCI was examined in 17 references published between 1963 and 1991. The authors noted an increased risk of death by suicide among those with SCI, with substantial variations in suicide risk across studies.¹⁷ A more recent review published by Kennedy and Garmon-Jones (2017) examined the proportion of SCI caused by suicidal behavior, the proportion of suicide deaths in the SCI population, and risk factors associated with suicidal behavior.¹⁸ Using a limited search strategy, this review included 22 English articles published between 1990 and 2016. The authors concluded that the percent of suicide deaths ranged between 4 and 11% following SCI.¹⁸

Although extant literature has reported on the significance of suicide risk among adults with SCI, the objective of this updated systematic review is to describe the frequency of suicidal thoughts and behaviors (including death) within study samples, and associated risk and protective factors. Another aim is to extend the two prior reviews with adherence to Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 updated guideline for reporting systematic reviews.¹⁹ To conduct this review, the team deployed a rigorous search strategy and risk of bias appraisal informing the synthesis of evidence on self-directed violence after SCI across nearly 70 years of research.

Methods

Study design

The present review was conducted using the PRISMA 2020 reporting method.¹⁹ This included a rigorous assessment of bias,²⁰ with a thorough search strategy informed by subject matter expert guidance on key words. Search strategy terms for SI and suicidal behaviors were adopted from the Self-Directed Violence Classification System and through expert consensus from team members who have successfully conducted multiple suicide-specific systematic reviews.²¹ A completed PRISMA 2020 Checklist is available (See Supplement 1). Key questions (KQ) investigate the frequency of SI and behaviors (i.e. SA, death by suicide) (KQ1) following SCI. Associated risk factors (KQ2) and protective factors (KQ3) for reduced SI and suicidal behaviors identified within the included studies were also explored.

Participants, interventions, comparators

For the PRISMA screening stage, at least two reviewers (LMB, KSY, MH) independently screened each title and abstract for retrieval in Covidence.²² No automation tools were used. When not in agreement, a third reviewer (ASH, LAB) evaluated the record for the final retrieval decision. Covidence was also used to evaluate records selected for the PRISMA eligibility stage of the review. Each full-text record was assessed by at least two reviewers (LMB, KSY). The full text decision process was stepwise and based on the PI(E)COTS eligibility criteria model such that each article was assigned one reason for exclusion even if the record may have been excluded for multiple reasons. The reference list of all records excluded at full text review along with each associated reason for exclusion is available upon request. Any disagreements at eligibility stage were similarly resolved by a third blind reviewer (ASH). Discussion by consensus was achieved regarding any final inclusion decisions.

Eligibility Criteria were defined according to the PI(E)CO(TS) framework²⁰:

• Population – Adults (18 + years old) with reported suicide-related thoughts and/or behaviors following diagnosis of SCI

• Intervention/Exposure – An intervention was not be required for inclusion. Suicide-related outcome measurement must have occurred after exposure to SCI.

• Comparator – No comparison group required for inclusion.

• Outcome – Reported rates of suicide-related thoughts and behaviors (preparatory behaviors, SAs, and suicide) among adults living with SCI.²¹ Reports may include associated risk/protective factors with suicide-related thoughts and behaviors among adults with SCI.

• Timing/Setting – Restrictions were not based on timing, setting, or study design.

• Only studies including original data and published in a peer-reviewed journal were included.

Systematic review protocol

A protocol for this review is registered in PROSPERO International prospective register of systematic reviews under registration number CRD42020164686.²³

Search strategy

Databases were searched since inception using controlled subject heading vocabulary and key words for suicidal thoughts and behaviors, combined with controlled subject heading vocabulary and key words for spinal cord injury (see Supplement 2). Subject matter experts (author initials) were involved in reviewing and finalizing the SCI-specific key words and subject headings prior to implementation in the search. Searches were limited to English only. Citation searching of reference lists were mined for additional studies. Complete references were exported into EndNote X9,²⁴ duplicates were removed, and then imported into Covidence review software.²⁵

Data sources, studies sections and data extraction

The final literature searches of OVID Medline, EMBASE, OVID PsycINFO, Web of Science Core Collection, CINAHL, Cochrane Library, and Google Scholar were conducted on July 7, 2021.

Data from included articles were independently abstracted into evidence tables by two authors and checked for accuracy by another author (LMB, MH; conflicts resolved via discussion with LMB, ASH, KSY). Extracted data from each article included a description of the study setting and time period, data source, categorization of SCI, outcome collection measures, and outcome results, as well as risk and protective factors (see Supplement 3). Studies included in this review were independently evaluated for study design and RoB by two reviewers (ASH, KSY) in a custom Research Electronic Data Capture (REDCap) database,²⁶ with RoB and study design disagreements resolved by consensus discussion.

Strength of evidence and risk of bias

Included studies were assessed for RoB using the Effective Public Health Practice Project (EPHPP) quality assessment tool for quantitative studies.²⁷ The EPHPP assessment was conducted in a custom REDCap database. The EPHPP tool groups items for reporting bias by domain: selection bias; study design; confounders; blinding; data collection; withdrawals/dropouts; and, other sources (e.g. no disclosure of conflicts of interest).²⁷ To inform the study design appraisal, studies were classified using the Taxonomy of Study Design Tool.²⁸ Each of the RoB domains, if applicable, was rated as low, moderate, or high, and an overall RoB rating was generated.^27,29 RoB ratings were descriptively synthesized to examine the overall strength of evidence and inform confidence in the results for SI, SA, and death by suicide. Ratings were based only on information available in the published studies.

Data analysis

Variability of study designs and outcome measurement precluded a meta-analytic approach to synthesis. Results were not quantitatively synthesized because included studies were mostly a mix of observational designs that utilized different follow-up periods and diverse measurement approaches to assess a variety of outcomes relevant to the review KQs. Therefore, a descriptive synthesis approach was used. Studies were grouped for synthesis by reported outcomes (SI, SA, and death by suicide). Associated risk and protective factors reported in study results and discussion (i.e. associated with risk for suicide, factors predictive of suicide within the sample, higher frequencies noted within the studies) were abstracted by two authors (LMB, MH) for KQ2 and 3. Certainty assessment was guided by RoB synthesis of studies contributing to each KQ and confidence in the body of evidence was described for each outcome of interest.

Results

Study selection and characteristics

Of the 2,118 unique records screened, 69 reports met eligibility criteria and were included in the review (See PRISMA 2020 flow diagram; Fig. 1). Due to the large number of included results, KQ1 was divided to evaluate: suicidal ideation only (KQ1a); suicidal ideation and behaviors (e.g. attempts/deaths) (KQ1b); and, suicidal behaviors only, including SA and death by suicide (KQ1c). A total of 8 articles addressed KQ1a. A total of 15 articles reported information regarding KQ1b. Suicidal behaviors only (KQ1c) was assessed by 46 articles. KQ2 and KQ3 were addressed by 45 articles.

Figure 1.

PRISMA 2020 flow diagram for new systematic reviews which included searches of databases, registers and other sources.

Study design

About half (n = 36) of included studies were cohort designs (n = 22 retrospective only, n = 4 mixed retrospective/prospective, n = 8 prospective only, n = 2 non-concurrent).^30–65 The next most common were cross-sectional quantitative study designs (n = 17),^66–82 followed by before-after quantitative studies (n = 3),^83–85 as well as a handful of qualitative studies (n = 5) consisting of both cross-sectional (n = 4) and before-after (n = 1) designs,^86–90 n = 1 case–control study,⁹¹ and n = 7 non-comparative case reports/case series.^92–98 (See Table 1).

Table 1.

Risk of bias and study design of included evidence (n = 69).

Study	Study Design	Key Question(s)	Source of Bias						Overall Bias
			Selection Bias	Study Design	Confounders	Blinding	Data collection	Withdrawals/dropouts
Ahoniemi et al.30	Retrospective cohort	Behaviors
Anderson et al.66	Cross-sectional	SI
Badiye et al.67	Cross-sectional (repeated)	Behaviors
Bahrudin et al.86	Cross-sectional (qualitative)	SI
Biering-Sørensen et al.31	Retrospective & prospective cohort	SI & behaviors
Cao et al.32	Retrospective cohort	Behaviors
Chamberlain et al.33	Prospective cohort	Behaviors
Charlifue et al.91	Case control	Behaviors
Colleran et al.68	Cross-sectional	SI & behaviors
Craig et al.34	Prospective cohort	SI
Crump et al.35	Retrospective cohort	Behaviors
DeVivo et al.36	Retrospective cohort	Behaviors
DeVivo et al.37	Retrospective cohort	Behaviors
DeVivo et al.38	Prospective cohort	Behaviors
DeVivo et al.39	Non-concurrent cohort	Behaviors
DeVivo et al.40	Retrospective cohort	Behaviors
Dorsett et al.41	Prospective cohort	Behaviors
Fann et al.69	Cross-sectional	SI & behaviors
Frankel et al.42	Retrospective cohort	Behaviors
Garshick et al.43	Prospective cohort	Behaviors
Geisler et al.70	Cross-sectional	Behaviors
Geisler et al.71	Cross-sectional	Behaviors
Gerner et al.72	Cross-sectional	SI
Ghajarzadeh et al.44	Prospective cohort	Behaviors
Ghajarzadeh et al.73	Cross-sectional	Behaviors
Hackler45	Non-concurrent cohort	Behaviors
Hagen et al.46	Retrospective cohort	Behaviors
Hartkopp et al.47	Retrospective cohort	Behaviors
Hartkopp et al.48	Retrospective cohort	Behaviors
Imai et al.49	Retrospective cohort	Behaviors
Judd & Brown92	Non-comparative (case series)	Behaviors
Jurisic & Marusic74	Cross-sectional	SI & behaviors
Kewman & Tate93	Non-comparative study (case report)	Behaviors
Khazaeipour et al.75	Cross-sectional	SI
Khazaeipour et al.76	Cross-sectional	SI & behaviors
Kishi & Robinson83	Before-after	SI & behaviors
Kishi et al.84	Before-after	SI & behaviors
Kishi et al.85	Before-after	SI & behaviors
Kraus et al.50	Retrospective cohort	Behaviors
Krause et al.51	Prospective cohort	SI
Krause et al.77	Cross-sectional	SI & behaviors
Kriz et al.65	Retrospective cohort	Behaviors
Le & Price52	Retrospective cohort	Behaviors
Lidal et al.53	Retrospective cohort	Behaviors
Lohne & Severinsson87	Before-after (qualitative)	SI & behaviors
McCullumsmith et al.78	Cross-sectional	SI & behaviors
Minaire et al.79	Cross-sectional	Behaviors
Nam et al.80	Cross-sectional	SI & behaviors
Niemi-Nikkola et al.54	Retrospective cohort	Behaviors
Nyquist55	Retrospective cohort	Behaviors
Nyquist & Bors56	Retrospective cohort	Behaviors
Petrus & Balaban94	Non-comparative (case report)	Behaviors
Pinkerton & Griffin81	Cross-sectional	SI & behaviors
Price95	Non-comparative (case series)	Behaviors
Rish et al.57	Retrospective cohort	Behaviors
Sabre et al.58	Retrospective cohort	Behaviors
Sauri et al.82	Cross-sectional	SI
Savic et al.59	Retrospective & prospective cohort	Behaviors
Savic et al.60	Retrospective & prospective cohort	Behaviors
Silver & Gibbon96	Non-comparative study (case series)	Behaviors
Smith & Caddick88	Cross-sectional (qualitative)	SI
Soden et al.61	Retrospective cohort	Behaviors
Stanford et al.62	Retrospective & prospective cohort	Behaviors
Tchajkova et al.90	Cross-sectional (qualitative)	SI & behaviors
Thietje et al.63	Retrospective cohort	Behaviors
Tribe97	Non-comparative study (case series)	Behaviors
Waals et al.98	Non-comparative study (case series)	Behaviors
Wang et al.89	Cross-sectional (qualitative)	SI & behaviors
Wilcox & Stauffer64	Prospective cohort	Behaviors

Notes: Risk of bias:  = low;  = moderate;  = high;  = not applicable. Full reference list of included studies can be found in Supplement 3.

Synthesized findings and risk of bias

Studies assessing SI and SA had high overall RoB and the strength of evidence was low. Studies assessing death by suicide had low RoB and the overall strength of evidence was high. About 30% of included studies (n = 21) were assessed as low RoB, about 22% (n = 15) were rated moderate RoB, and the remaining almost half (n = 33) were high RoB. The low RoB studies were exclusively cohort designs informing the death by suicide evidence, with notable strengths (e.g. representative selection of participants for generalizability, reliable and valid data collection approaches, and adequate control for confounding). However, there were many common sources of bias found in the moderate and high RoB studies that largely informed the evidence for SI and SA. The cross-sectional and non-comparative studies introduced high RoB in the study design domain (n = 28). Similarly, lack of blinding was a high RoB issue in n = 26 studies, including among the generally more rigorous prospective cohort designs. The majority of studies earned low or moderate RoB scores in the remaining RoB domains (e.g. selection bias, confounding, data collection, and withdrawal/dropouts).

Synthesis of evidence

Tables synthesizing evidence are presented in Supplement 3. A summary of the synthesis of evidence is described below.

KQ1a suicidal ideation only:

A total of 8 studies reported on rates of SI only.^{34,51,66,72,75,82,86,88} The overall strength of evidence, and therefore confidence in supporting the results for SI was determined to be low, as all studies were assessed at high RoB (6 cross-sectional designs and 1 prospective cohort) except for one with moderate RoB (prospective cohort).³⁴ Suicidal ideation was assessed using a variety of measures. The Patient Health Questionnaire-9 item #9 was used to assess SI within the last two weeks for three studies.^51,66,82 Two studies evaluated suicidal thoughts via qualitative interview,^86,88 and another two studies used author generated questionnaires.^72,75 One study used the Mini International Neuropsychiatric Interview (MINI).³⁴ Report of SI was predominantly expressed in percent or number of individuals. Presence of SI in the last two weeks ranged from a low of 1% (SI more than half the days, and SI every day, respectively),⁶⁶ to 4.9% upon hospital discharge,³⁴ to a high of 22.5%.⁸² Higher percentages of SI were reported in two studies using author-created questionnaires, with 40.16% of 122 participants in a study in Germany reporting having SI “occasionally”,⁷² and 30.3% of 119 participants in a study conducted in Iran reported having SI.⁷⁵ Four studies reported on sample characteristics of injury classification type (i.e. paraplegia, tetraplegia) and severity of injury (i.e. complete, incomplete).^34,66,75,82 Only one study used the American Spinal Injury Association (ASIA) classification.⁵¹ None of these studies reported on rates of SI per injury classification type, severity, or ASIA classification.

KQ1b suicidal ideation & suicidal behaviors:

A total of 15 studies assessed for both SI and SA or suicide deaths.^{31,68,69,74,76–78,80,81,83–85,87,89,90} The strength of evidence for these studies was determined to be low. The lack of confidence in this evidence was demonstrated by high RoB findings. Ten studies were cross-sectional, 4 studies were before/after, and 1 study analyzed retrospective records and conducted prospective follow-up.³¹ Only 2 studies used the validated PHQ-9 to assess SI.^69,78 One study used the Suicide Behavior Questionnaire to assess for suicidal behaviors.⁶⁸ Three studies assessed SI and suicidal behaviors by qualitative interview,^87,89,90 1 study used the Present State Examination structured mental status exam to gather data on suicidal plans and suicidal acts,⁸³ with the suicidal plans considered as SI, and 1 study used medical records reporting SI, SAs, and suicide deaths.³¹ The remaining studies used author-created questionnaires or interviews with the individual and/or collateral informants (e.g. general practitioners) to assess SI and SA or suicide deaths.

Only two studies reported on ASIA classification and either type of SCI (i.e. paraplegia, tetraplegia) or neurological level and severity of SCI,^80,90 though only one study reported SI and SA by level and severity.⁹⁰ Ten studies reported on type, severity or neurological level, with only one of these reporting on SI per type of SCI.^{31,69,74,76–78,81,83,87,89}

Suicidal ideation ranged from lows of about 4% to 10%^31,83,87 of a sample, to a high of 67%.⁶⁸ However, the period of time associated with reported SI greatly varied between the last two weeks and ever since the SCI. Eleven of the 15 studies reported SAs, two studies reported suicide deaths. Four studies reported SAs occurring prior to the SCI in addition to suicide behavior after SCI.^68,78,81,87 Biering-Sorensen and colleagues enrolled 45 individuals whom had obtained their SCI via SA between 1965-1987, and during the study follow-up period in 1988–89 they found three individuals ultimately died by suicide after SCI and seven individuals made a SA, with two having made a SA more than once, and two reported ongoing SI.³¹ McCullumsmith et al. found that 4.4% of their sample attempted suicide both prior to, and post-, SCI.⁷⁸ Two studies each found that 10%, or one individual, from each respective sample attempted suicide that resulted in the SCI.^68,87 SA frequency ranged from none to 65.8% across the studies.^68,77 Suicide death rate was reported as 17.3% (22 out of 127 individuals) in one study,⁸⁰ while a study with a smaller sample (N = 45) reported 6.67% suicide deaths.³¹

KQ1c: suicide attempts and deaths post-SCI

Whereas three included studies reported on SAs and deaths, and one reported on SA only, 42 (61%) of studies addressed suicide deaths only. Although there was a range of overall RoB among articles reporting on SA and suicide deaths, all of the low RoB studies included in this review (N = 21) contributed to the evidence informing death by suicide. Of those with a low RoB, there is high certainty of confidence in the evidence comprised of 15 retrospective cohorts, 3 prospective cohorts, 2 retrospective and prospective, and 1 non-concurrent cohort. Predominantly death registers, autopsy reports, and health insurance or hospital records were used to document number of deaths. Interviews with the individual and/or collateral informants (e.g. family, general practitioners) was the other mode of data collection for attempts. Among the non-case report studies, percentage of suicide deaths ranged from less than 1%,^39,67 to approximately 10%.^{30,47,48,63,65,71} Cao and colleagues reported both suicide mortality rates and standardized mortality ratios among those with a SCI in the U.S. across three decades; 1970, 1980, and 1990.³² Crude annual suicide mortality rates during the first 12 years post-SCI for those in the 1973–1979 cohort was 91 per 100,000 person years, 69 for the 1980–1989 cohort, and 46 for the 1990–1999 cohort. Standardized mortality ratios (SMR) were 5.2, 3.7, and 3.0, respectively. In a slightly older study, a suicide rate of 59.2 per 100,000 was reported.⁹¹ Hagen and colleagues in Norway found a male cause-specific suicide SMR of 3.70 and a female SMR of 37.59.⁴⁶ Hartkopp and colleagues reported a suicide SMR of 4.6 among a Denmark population.^{47, 48} In Estonia, the suicide SMR was found to be 8.3 (among all males),⁵⁸ and in Australia the suicide SMR has been reported at 4.4.⁶² More recently, Kriz and colleagues reported a cause-specific suicide SMR of 12.9, as well as suicide SMR based on age, with a suicide SMR of 10.5 for patients less than 60 years of age and 21.6 for patients 60 years or older.⁶⁵

Approximately half (N = 23) of the studies reporting on SA and suicide death provided ASIA or Frankel classification,^{32,40,42–44,46–49,52–54,56–60,63,73,92,93,95,98} or level of injury, and 17 studies reported on type and/or severity of SCI.^{30,33,36–39,41,43,44,61,64,65,70,71,73,79,91,94,96,97} Of these 40 studies, one study reported on SA and type of SCI (e.g. paraplegia), 5 studies were case studies that reported on suicide death and type/severity or ASIA/level classification,^93,95–98 and about half of the studies reported suicide deaths by either type and/or severity (N = 11),^{30,33,36–39,41,61,64,65,70,71,79,91} or ASIA/level classification (N = 9).^{46–48,52,53,57,58,60,63} Four studies also reported on attempts causing the SCI,^46,60,62,91 and two of these reported on attempts prior to, but not causing, the SCI.^62,91 Method of SA or suicide death was reported in 11 studies, most frequently reporting firearms, drug/substance overdose, carbon monoxide poisoning, cutting, motor vehicle accidents, suffocation/asphyxiation, and drowning.

KQ2 & 3: associated risk and protective factors

A total of 45 papers included some discussion of risk or protective factors examined in their respective studies. Identification of possible risk factors was more prevalent than identification of protective factors (n = 4 studies).^32,34,56,59 The most common risk factors mentioned in reviewed studies included younger age, history of a mental disorder and/or personality features, and/or presence of mental health symptoms. Risk among level of injury/classification of SCI varied. That is, some studies found that paraplegia was associated with higher risk for SI and suicide deaths, while other studies reported that tetraplegia/quadriplegia was associated with higher risk. Older age and longer time since SCI injury were reported as protective factors.^56,70 Another study found that reduction of major depressive disorder symptoms reduced suicide risk at hospital discharge and at 6-months post-discharge.³⁴

Discussion

Notable rates of suicidal ideation and behaviors, including deaths, were reported among individuals with SCI. Results from this review confirm findings from past reviews on the topic.^17,18 Most of the studies in this review reported on SA and suicide deaths, with relatively fewer studies reporting on SI. About 20% of the studies reported on SI and SA. RoB was highest among studies in which data regarding SI was reported. SA evidence quality from RoB analyses was better than SI evidence, but still lacking. Weakness in the quality of SI and SA evidence was driven by threats to selection bias, reduced representativeness of study samples, as well as confounding in both design and analyses, and approaches to data collection using assessment methods that were not shown to be valid and reliable. These risks lower confidence in the evidence and lead to instability in frequency estimates for SI and SA derived from the included studies. Furthermore, measurement of key injury level and severity characteristics and outcomes were not consistent across studies. All these limitations present opportunities for improved research method and sample designs. Assessment of SI varied by type of measure (i.e. validated self-report measures, clinical interviews, author-created questionnaires, qualitative interviews) and frequency of SI (i.e. daily, weekly, ever). Furthermore, rates of SI and the timeframe in which SI was measured widely varied (e.g. within the last two weeks to lifetime), leading to further uncertainty in conclusions that can be made from this body of research. Increased utilization of psychometrically sound measurements to ascertain frequency, duration, and intensity of suicidal thoughts and enhancement of suicide risk identification and follow-up among individuals with SCI is needed.

The highest confidence among results was ascertained from reports on suicide deaths after SCI. The evidence consistently showed that individuals with a history of SCI are at increased risk of death by suicide. While a small number of studies reported on both SAs and suicide deaths, the majority reported on suicide deaths only (65%). While most studies reported N or percent frequency of SA or deaths, relatively few compared the rate of SA or suicide deaths to the general population. Twelve studies reported on suicide standardized mortality (SMR) ratios, with all SMRs reported occurring at greater than 1.0 (range 2.2–12.9 SMR),^39,65 such that the observed deaths greatly exceed the normally expected deaths in that population. Approximately 22% of the articles that reported SA or suicide deaths (n = 60) provided information regarding method of attempt. Methods varied and included self-inflicted gunshot wound, asphyxia, drowning, hanging, jumping, motor vehicle accidents, and drug overdose.

A majority of papers (64%) included some discussion of risk factors. Although common risk factors frequently observed among members of the general population (e.g. male, white race) were not commonly noted, there was considerable overlap (e.g. age, history of mental disorder and/or personality features, presence of mental health symptoms). Younger age (less than 50 years) among individuals with SCI was another commonly mentioned risk factor. Only four studies reported on time since sustaining SCI as a risk factor for suicidal behaviors. Charlifue and colleagues reported that 50% of suicides occurred within three years post-injury, and 76% of suicides occurred within four years.⁹¹ McCullumsmith and colleagues reported that the mean number of years since injury to attempt was 6.1 (SD = 6.7).⁷⁸ Similarly, Savic et al. reported that the mean number of years since injury to suicide death was 6.6 (SD = 4.2) and that 91% of suicide deaths in the sample occurred within the first ten years post-injury.⁶⁰ Another study found that all but one suicide death (out of 13) occurred within one year after a traumatic SCI.⁵⁸ Three studies reported on suicide attempts prior to, or causing the SCI, that were also noted as risk factors for subsequent SAs or suicide deaths. Although an additional eight studies did report on SA prior to subsequent SA or death following SCI, these were not identified by the authors as risk factors for suicide in their sample. Further research specifically exploring timing of injury to death by suicide is needed.

Data regarding suicide risk pertaining to SCI level of injury/classification severity widely varied, from studies not reporting any characteristics of the SCI, to studies reporting multiple characteristics (i.e. type, severity, level, ASIA or Frankel classification). Only 2 studies pertaining to K1b reported on SI and SCI characteristics, and 21 studies reported on SA or suicide deaths and SCI characteristics, limiting the ability to draw conclusions regarding the association between SCI characteristics and suicidal thoughts and/or behaviors.

Only four studies specifically reported on protective factors. One study found that a longer time since SCI was associated with less suicide deaths.⁵⁹ In another, no suicide deaths were identified among those over the age of 53, prompting the authors to suggest that age is a protective factor.⁵⁵ In one study cohorts of individuals with SCI sustained in more recent decades had lower odds of dying by suicide when compared to cohorts whose injuries were acquired earlier in time.³² Although the authors did not make empirical conclusions regarding these findings, they suggested that use of newer antidepressants and increases in social and physical accessibility may contribute to protective factors.³² Finally, reduction of major depressive disorder symptoms was found to be associated with reduced suicide risk at 6-months post-discharge from acute rehabilitation.³⁴

Gaps in reviewed studies & future investigation

Our results highlight the lack of consistency in methods for acquiring data regarding SI and SA among members of this high-risk population. Only eight studies used psychometrically sound measures, such as the PHQ-9, or clinical interviews such as the MINI. Use of measures that have been validated among individuals with disabilities and across medical settings are necessary to ensure identification of outcomes of interest.⁹⁹

This review also confirmed previously identified inconsistencies in reporting of SCI-specific characteristics, such as injury classification (paraplegia and quadriplegia/tetraplegia) and severity of injury (complete/incomplete). Less than 20% of studies included in this review (n = 13) used the ASIA or older Frankel classification system or partial classification to describe injuries among members of the sample.^{32,39,42,44,53,58–60,63,73,80,90,93}

Limitations of the current review

Limitations of this review included lack of examining the presence and rate of SI and suicidal behaviors in pediatric populations. Another limitation was the exclusion of articles published solely on SAs resulting in SCI. Eight papers cited in this review reported SAs pre- and post-SCI. Understanding how suicide risk increases or decreases pre- and post-injury may also provide the field with greater knowledge of how to assess risk and crucial timepoints during which assessment and risk mitigation strategies (e.g. safety planning, lethal means counseling, evidence-based psychotherapy) should be implemented.

Clinical implications

Assessing for SI among those with SCI may help to identify those at greater risk for attempting suicide prior to escalation of such behaviors. Use of measures that provide both high sensitivity and specificity to improve screening and assessment are encouraged. Ease of access, administration, and interpretation is expected to facilitate feasibility of implementation. Whereas a limited number of studies (N = 4) reported on time since injury as a risk factor, it is not well known if this is a period of increased suicide risk. In the general population two recent large scale health systems studies have demonstrated that almost half of suicide deaths often occur within a month of a health care visit,^100,101 and nearly all individuals who died by suicide had some form of health care treatment within the past year. Considering the fact that individuals with SCI are often frequent consumers of healthcare,^{7, 8} this increased contact should provide the opportunity for frequent and ongoing suicide risk screening, assessment, and management. In alignment with current best practices,^102,103 screening tools such as the Columbia-Suicide Severity Rating Scale screener measures should be used to determine current suicide risk.¹⁰⁴ If higher risk is indicated, comprehensive suicide risk evaluation and implementation of Safety Plans or acute risk mitigation strategies may follow.

Conclusions

The frequency of suicidal thoughts and behaviors, including death, among individuals living with SCI warrants increased research and clinical focus. Results from this review highlight variability in methodological approaches (e.g. self-report, clinical or qualitative interviews, medical record reviews) used to identify outcomes of interest. The timeframes in which thoughts and behaviors occurred in relation to the SCIs also varied. Diverse study designs, and in some cases limited data reported, impacted comparison of the results found in studies reviewed. Although several suicide risk factors were identified, limited data exists regarding the strengths of the association between these risk factors and SI, SA, or deaths. Moreover, there is a dearth of knowledge regarding protective factors. The need for rigorous study design and consistent sample characteristic reporting to both enhance the quality of evidence of the literature and allow researchers to draw informed conclusions regarding suicide risk in this population is warranted. This review provides strong evidence for the need for enhanced early detection (i.e. screening assessment) and intervention strategies to reduce suicide risk among those living with SCI.

Supplementary Information

Supplemental data for this article can be accessed https://doi.org/10.1080/10790268.2022.2029282.

Disclaimer statements

Contributors None

Conflict of interest Authors have no conflict of interests to declare

Funding Statement

Support for this effort was provided by the VA Rocky Mountain Mental Illness Research Education and Clinical Center and the Health Sciences Library of the University of Colorado, Anschutz Medical Campus. This manuscript is based on work supported, in part, by the United States Department of Veterans Affairs. The contents do not represent the views of the U.S. Department of Veterans Affairs or the United States Government.