Suicide and Self-Harm Risk Assessment: A Systematic Review of Prospective Research

Figures & data

TABLE 2. Quality appraisal and level of evidence assessment (n = 21).

Reference	Similar groups from same population	Similar exposure for study groups	Valid and reliable exposure measure	Confounders identified	Confounders addressed	Groups free of outcome at start	Valid and reliable outcomes measure	Follow-up time reported	Follow-up complete	Strategies for incomplete follow-up	Suitable statistical analysis	Level of evidence
Brucker et al. (2019)	NA	NA	Y	U	U	Y	Y	Y	Y	NA	Y	2+
Campbell and Beech (2018)	NA	NA	Y	Y	Y	Y	Y	Y	Y	NA	Y	2+
Chang and Tan (2015)	NA	NA	Y	U	U	Y	Y	Y	Y	NA	Y	2+
de Beurs et al. (2015)	NA	NA	Y	Y	Y	Y	Y	Y	Y	NA	Y	2+
Dickens and O’Shea (2015)	NA	NA	Y	Y	Y	Y	Y	Y	Y	NA	Y	2+
Glassmire et al. (2016)	NA	NA	Y	Y	Y	Y	Y	Y	Y	NA	Y	2+
Green et al. (2015)	NA	NA	Y	Y	Y	Y	Y	Y	Y	NA	Y	2+
Harrison et al. (2018)	NA	NA	Y	U	U	Y	Y	Y	Y	NA	Y	2+
Hawes et al. (2017)	NA	NA	Y	U	U	Y	Y	Y	Y	U	Y	2+
Jordan and McNiel (2018)	NA	NA	Y	Y	Y	Y	Y	Y	Y	NA	Y	2+
Katz et al. (2017)	NA	NA	Y	U	U	N	Y	Y	Y	Y	Y	2+
Madan et al. (2016)	NA	NA	Y	U	U	U	Y	Y	N	U	Y	2-
O’Shea et al. (2014)	NA	NA	Y	Y	Y	Y	Y	Y	Y	NA	Y	2+
Quinlivan et al. (2017)	NA	NA	Y	U	U	U	Y	Y	Y	U	Y	2+
Randall et al. (2019)	NA	NA	Y	Y	Y	Y	Y	Y	Y	U	Y	2+
Reilly-Harrington et al. (2016)	NA	NA	Y	U	U	U	Y	Y	U	U	Y	2+
Saunders et al. (2014)	NA	NA	Y	U	U	U	Y	Y	Y	U	Y	2+
Stefansson et al. (2015)	NA	NA	Y	Y	Y	Y	Y	Y	Y	NA	Y	2+
Villegas et al. (2018)	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	2+
Wang et al. (2016)	NA	NA	Y	U	U	U	Y	Y	Y	Y	Y	2+
Wu et al. (2014)	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	2+

N: no; NA: not applicable; U: unclear; Y: yes.

FIGURE 1. Study identification, screening, and selection process.

TABLE 1. Study characteristics (n = 21).

Country	USA (n = 9)
	UK (n = 6)
	Canada (n = 3)
	Australia (n = 1)
	Sweden (n = 1)
	Taiwan (n = 1)
Study design	Prospective (cohort) study (n = 17)
	Pseudo-prospective cohort study (n = 2)
	Longitudinal prospective study (n = 1)
	Psychometric evaluation with follow-up (n = 1)
Sample size (min-max)	50–5,462
Settings	Emergency department (n = 7)
	Acute care/hospital/speciality not specified (n = 4)
	Forensic inpatient (n = 2)
	Inpatient mental health care (n = 2)
	Emergency department and general medicine (n = 1)
	Inpatient and community (n = 1)
	Liaison psychiatry service (n = 1)
	Maximum security facility (n = 1)
	Psychiatric outpatient (n = 1)
	Psychiatry services (n = 1)
Instruments^a	SAD PERSONS (and modified/Chinese) Scale (n = 6)
	Beck Scale for Suicide Ideation (n = 4)
	Beck Hopelessness Scale (n = 3)
	Columbia Suicide Severity Rating Scale (n = 2)
	Concise Health Risk Tracking (n = 2)
	Historical, Clinical and Risk Management Scales (n = 2)
	Patient Health Questionnaire 9 (n = 2)
	Others (n = 21)
Follow-up (min-max)	2 weeks–20 years

^aSeveral studies used more than one instrument; n corresponds to the number of times an instrument was used.

TABLE 3. Summary of findings from individual studies (n = 21).

Reference	Country	Design	Sample	Setting	Instrument	Follow-up	Outcome	Findings
Brucker et al. (2019)	USA	Prospective	338	ED	SCS, Physician gestalt VAS, CFI-S	6 mon	Suicidality spectrum	SCS: modest diagnostic accuracy sensitivity 14/32 = 44%, (95%CI 26–62%) and specificity 275/306 = 90%, (95%CI 86–93%); VAS: moderate diagnostic accuracy (AUC = 0.75, 95%CI 0.66–0.85); CFI-S: Best for diagnostic accuracy (AUC = 0.81, 95%CI 0.76–0.87).
Campbell and Beech (2018)	UK	Prospective	89	Forensic inpatient	HCR-20, FAM	1 yr	Self-harm	Positive association between mean scores on HCR-20 and frequency of self-harm, with (p < 0.001) and without FAM (p < 0.001).
Chang and Tan (2015)	USA	Prospective	50	ED	PHQ-9, BSSI, C-SSRS, SPS	2 wk	Admission Length of stay Adverse events	19 (38%) ≥ 1 adverse event. BSSI, PHQ-9, C-SSRS poor at predicting outcomes (AUC ≤ 0.5). SPS better at predicting admission (p = 0.009) and stay (p = 0.006) but not ED adverse events. Positive SPS score: 8.18 times the odds of admission.
de Beurs et al. (2015)	UK	Prospective	366	Liaison psychiatry	BSSI	15 mon	Repeat suicide	94 (26%) treated after repeat suicide. Depression (OR = 1.04, 95%CI 1.02–1.06, p < 0.001), hopelessness (OR = 1.07, 95%CI 1.02–1.13, p < 0.006), suicidal ideations higher among attempters vs non-attempters (M = 23 ± 8.4 vs M = 18 ± 10.6, p < 0.001).
Dickens and O’Shea (2015)	UK	Pseudo-prospective	217	Secure inpatient	START	3 mon	Self-harm Suicide	Survival rates differed between groups rated as low and moderate risk (χ^2 = 24.94, df = 1, p < 0.001), and between low- and high-risk groups (χ^2 = 33.29, df = 1, p < 0.001), but did not between moderate- and high-risk groups (χ^2 = 1.591, df = 1, p = 0.207).
Glassmire et al. (2016)	USA	Prospective	229	Maximum security facility	MMPI-2-RF SUI items	1 yr	Suicidal ideation and behavior	80 (34.9%) endorsed ≥1 item. Endorsing 2,3,4–5 items:10,12,14 times greater risk. Past attempt and ideation, and current ideation: 5,6,10 times greater risk. Endorsing ≥2 items: 3–9 times greater risk. Having suicidal ideation and endorsing ≥1 item: 15–22 times greater risk.
Green et al. (2015)	USA	Prospective	S1: 5,200 S2: 119	Psychiatric outpatient	Diagnostic interviews, BDI, BSSI	S1: 20 yr S2: 18 mon	Repeat suicide Death by suicide	BDI predicted suicide in S1 (Wald X^21 = 35.67, p < 0.001, HR = 2.79, 95%CI 2–3.91) and S2 (Wald X^21 = 8.82, p < 0.01, HR = 1.69, 95%CI 1.19–2.38) BDI predicted death by suicide (p < 0.01) in both samples. AUC for BDI suicide = 0.63 and for BSSI = 0.67.
Harrison et al. (2018)	Australia	Prospective	128	ED	d/s-IAT	3 mon 6 mon	Suicidal ideations and attempts	Clinician and patient prediction accounted for 4.9% and 7.2% of variance in ideation (p = 0.049 and p = 0.011) at 3 and 6 mon, respectively. Clinician prediction as the only significant predictor of ideation, unaffected by adding d/s-IAT. Clinician and patient prediction and d/s-IAR did not significantly predict attempts.
Hawes et al. (2017)	USA	Prospective	59	Inpatient	MARIS	4–8 wk	Suicidal ideations and attempts	Correlation between MARIS score, lifetime attempts (rho = 0.30, p = 0.005), depression (rho = 0.46, p < 0.001), lifetime ideations (rho = 0.25, p = 0.023), and ideations in past month (rho = 0.35, p = 0.001). Those with behaviors had higher MARIS scores than those who did not (MD = 15.69,2.96; Cohen's d = 1.54,0.77; U = 33,119.5; p = 0.001,0.036, respectively).
Jordan and McNiel (2018)	USA	Prospective	218	Inpatient	SIS	1 yr	Repeat suicide attempt	69 (31.7%) attempted suicide. Cross-validated iterative receiver operator curve analysis of SIS data predicted subsequent attempts (AUC = 0.43, 95%CI 0.34–0.58, sensitivity = 61.54%, specificity = 56.91%, PPV = 37.65%, NPV = 77.78%).
Katz et al. (2017)	Canada	Longitudinal prospective	5,462	ED	SPS, MSPS	6 mon 1 yr 2 yr	Death by suicide	77 (1.4%) died by suicide. Of those, half were assessed as low risk at 6 and 12 mon. AUC for SPS and MSPS was 0.56 (95%CI 0.45–0.67) and 0.59 (95%CI 0.49–0.69). Scaled Brier score for SPS and MSPS indicated poor performance.
Madan et al. (2016)	USA	Prospective	1,055	Specialised hospital	C-SSRS, PHQ-9, BHS, SCogS	6 mon	Suicide-related behavior	C-SSRS correctly classified suicide-related behavior within 6 mon of discharge (AUC = 0.757, p < 0.001) with total and summary score from the ideation/behavior factor providing balance between sensitivity (0.694) and specificity (0.652–0.674).
O’Shea et al. (2014)	UK	Pseudo-prospective	504	Forensic inpatient	HCR-20	3 mon	Violence risk self-harm	HCR-20 scores higher in moderate- (M = 28.90 ± 4.78, p < 0.001) and high-risk groups (M = 29.52 ± 5.45, p < 0.001) compared to low-risk group (M = 24.66 ± 6.30). HCR-20 predicted self-harm but effect sizes not large (0.345–0.749).
Quinlivan et al. (2017)	UK	Prospective	464	Psychiatry services	MSHR, ReACT SHR, SPS, MSPS, BIS	6 mon	Repeat self-harm	145 (30%) repeat self-harm. Sensitivity 1% (95%CI 0–5) for SPS and 97% (95%CI 93–99) for MSHR. PPV 13% (95% CI 2–47) for MSPS and 47% (95%CI 41–53) for clinician assessment. Remaining scales performed worse than clinician and patient estimates.
Randall et al. (2019)	Canada	Prospective	2,643	ED	SLS, SALS, C-CASA	6 mon	Suicide attempt Death by suicide	56 (2.1%) attempts, 13 (0.5%) deaths at 6 mon and 85 (3.2%) attempts, 20 (0.8%) deaths at 12 mon. Clinician assessment not effective at predicting deaths (AUC = 0.546, 95%CI 0.36–0.73) and moderately accurate for attempts (AUC = 0.728, 95%CI 0.66–0.79). C-CASA moderately accurate at predicting attempts and deaths (AUC = 0.666 and 0.678).
Reilly-Harrington et al. (2016)	USA	Prospective	482	Acute care	CHRT	6 mon	Suicide-related serious adverse event	26 (5%) experiencing ≥ 1 adverse events had a sig higher CHRT score than those without adverse events (M = 29.9 ± 11.7 vs M = 23.7 ± 9.6, p = 0.003) this trend remained at follow-up (HR = 1.76, 95%CI 1.01–3.05, p = 0.04). CHRT highly correlated with clinician ratings of depression, anxiety, and overall functioning.
Saunders et al. (2014)	UK	Prospective	126	Hospital	SPS	6 mon	Repeat self-harm Admission Aftercare	5 (4%) psychiatric hospital admissions, 70 (55.5%) community aftercare, and 31 (24.6%) repeated self-harm. Specificity of SPS > 90% for all outcomes, yet only 2% sensitivity for admission, 5.8% for community aftercare, and 6.6% for repetition of self-harm.
Stefansson et al. (2015)	Sweden	Prospective	81	University hospital	SIS, KIVS	15–20 yr	Suicide	SIS alone: 52% specificity, 17% PPV; SIS and KIVS combined assessment: 63% specificity, 19% PPV; SIS and KIVS together: 83% sensitivity, 80% specificity, 26% PPV
Villegas et al. (2018)	USA	Prospective	109	Inpatient psychiatry	CHRT-SR12, CHRT-SR7, BHS, QIDS-SR, LOT-R, PANAS	6 wk 12 wk	Self-injury Suicide attempt	Decreased suicide risk using CHRT-SR12 (b = 0.24, p = 0.009) and CHRT-SR7 (b = 0.16, p = 0.004). Decreased hopelessness and depression using BHS (b = 0.21, p ≤ 0.001) and QIDS (b = 0.33, p ≤ 0.001)
Wang et al. (2016)	Canada	Prospective	2,792	ED	Clinician prediction scale, SPS	6 mon	Suicide attempts	136 (4.9%) attempted suicide within 6 mon. Clinicians were able to predict future attempts with sig greater accuracy (AUC = 0.73, 95%CI 0.68–0.77, p < 0.001) in comparison to SPS (z = 3.79, p < 0.001). Both showed PPV <7%
Wu et al. (2014)	Taiwan	Psychometric evaluation	147 cases 284 controls	ED Medicine	Chinese SPS, PSIS, HAMD, BSSI, BHS	6 mon	Repeat self-harm	SPS: AUC = 0.66, p = 0.02, 95%CI 0.53–0.79; PSIS: AUC = 0.49, p = 0.88, 95%CI 0.36–0.61; HAMD: AUC = 0.56, p = 0.40, 95%CI 0.42–0.70; BSSI: AUC = 0.59, p = 0.18, 95%CI 0.45–0.74; BHS: AUC = 0.55, p = 0.47, 95%CI 0.42–0.70. SPS cutoff point of 4/5 determined to predict repeat self-harm with 65.4% sensitivity and 58.1% specificity

AUC: Area under the curve; b: beta; BDI: Beck Depression Inventory; BHS: Beck Hopelessness Scale; BIS: Barratt Impulsiveness Scale; BSSI: Beck Scale for Suicide Ideation; C-CASA: Columbia Classification Algorithm of Suicide Assessment; CFI-S: Convergent Functional Information for Suicidality; CHRT: Concise Health Risk Tracking; CI: confidence interval; C-SSRS: Columbia Suicide Severity Rating Scale; d/s-IAT: Death/Suicide Implicit Association Test; ED: emergency department; FAM: Female Additional Manual; HAMD: Hamilton Rating Scale for Depression; HCR-20: Historical: Clinical and Risk Management Scales; HR: hazard ratio; IQR: interquartile range; KIVS: Karolinska Interpersonal Violence Scale; LOT-R: Life Orientation Test-Revised; M: mean; MARIS: Modular Assessment of Risk for Imminent Suicide; MMPI-2-RF: Minnesota Multiphasic Personality Inventory–2 Restructured Form; MSHR: Manchester Self-Harm Rule; MSPS: Modified SAD PERSONS Scale; NPV: negative predictive value; OR: odds ratio; PANAS: Positive and Negative Affect Schedule; PHQ-9: Patient Health Questionnaire 9; PPV: positive predictive value PSIS: Pierce Suicide Intent Scale; QIDS-SR: Depressive Symptomatology Self-Report; S1: sample 1; S2: sample 2; SALS: Suicide Attempt Likelihood Scale; SCogS: Suicide Cognitions Scale; SCS: Suicide Clinical Screening; SHR: Self-Harm Rule; SIS: Suicide Intent Scale; SLS: Suicide Likelihood Scale; SPS: SAD PERSONS Scale; START: Short Term Assessment of Risk and Treatability; SUI: Suicidal/Death Ideation; UK: United Kingdom; USA: United States of America; VAS: Physician gestalt Visual Analog Scale; χ²: chi square; z: standard score.

Brucker, K., Duggan, C., Niezer, J., Roseberry, K., Le‐Niculescu, H., Niculescu, A. B., & Kline, J. A. (2019). Assessing risk of future suicidality in emergency department patients. Academic Emergency Medicine, 26(4), 376–383. doi:10.1111/acem.13562

 PubMed Web of Science ®Google Scholar

Campbell, L., & Beech, A. (2018). Do scores on the HCR-20 and FAM predict frequency of self-harm in females within a secure psychiatric hospital? The Journal of Forensic Psychiatry & Psychology, 29(6), 914–933. doi:10.1080/14789949.2018.1477975

 Web of Science ®Google Scholar

Chang, B. P., & Tan, T. M. (2015). Suicide screening tools and their association with near-term adverse events in the ED. The American Journal of Emergency Medicine, 33(11), 1680–1683. doi:10.1016/j.ajem.2015.08.013

 PubMed Web of Science ®Google Scholar

de Beurs, D. P., de Groot, M. H., de Keijser, J., Mokkenstorm, J., van Duijn, E., de Winter, R. F., & Kerkhof, A. J. (2015). The effect of an e-learning supported Train-the-Trainer programme on implementation of suicide guidelines in mental health care. Journal of Affective Disorders, 175, 446–453. doi:10.1016/j.jad.2015.01.046

 PubMed Web of Science ®Google Scholar

Dickens, G. L., & O’Shea, L. E. (2015). How short should short‐term risk assessment be? Determining the optimum interval for START reassessment in a secure mental health service. Journal of Psychiatric and Mental Health Nursing, 22(6), 397–406. doi:10.1111/jpm.12232

 PubMed Web of Science ®Google Scholar

Glassmire, D. M., Tarescavage, A. M., Burchett, D., Martinez, J., & Gomez, A. (2016). Clinical utility of the MMPI-2-RF SUI items and scale in a forensic inpatient setting: Association with interview self-report and future suicidal behaviors. Psychological Assessment, 28(11), 1502. doi:10.1037/pas0000220

 PubMed Web of Science ®Google Scholar

Green, K. L., Brown, G. K., Jager-Hyman, S., Cha, J., Steer, R. A., & Beck, A. T. (2015). The predictive validity of the beck depression inventory suicide item. The Journal of Clinical Psychiatry, 76(12), 1683–1686. doi:10.4088/JCP.14m09391

 PubMed Web of Science ®Google Scholar

Harrison, D. P., Stritzke, W. G., Fay, N., & Hudaib, A. R. (2018). Suicide risk assessment: Trust an implicit probe or listen to the patient? Psychological Assessment, 30(10), 1317–1329. doi:10.1037/pas0000577

 PubMed Web of Science ®Google Scholar

Hawes, M., Yaseen, Z., Briggs, J., & Galynker, I. (2017). The Modular Assessment of Risk for Imminent Suicide (MARIS): A proof of concept for a multi-informant tool for evaluation of short-term suicide risk. Comprehensive Psychiatry, 72, 88–96. doi:10.1016/j.comppsych.2016.10.002

 PubMed Web of Science ®Google Scholar

Jordan, J. T., & McNiel, D. E. (2018). Characteristics of a suicide attempt predict who makes another attempt after hospital discharge: A decision-tree investigation. Psychiatry Research, 268, 317–322. doi:10.1016/j.psychres.2018.07.040

 PubMed Web of Science ®Google Scholar

Katz, C., Randall, J. R., Sareen, J., Chateau, D., Walld, R., Leslie, W. D., … Bolton, J. M. (2017). Predicting suicide with the SAD PERSONS scale. Depression and Anxiety, 34(9), 809–816. doi:10.1002/da.22632

 PubMed Web of Science ®Google Scholar

Madan, A., Frueh, B. C., Allen, J. G., Ellis, T. E., Rufino, K. A., Oldham, J. M., & Fowler, J. C. (2016). Psychometric reevaluation of the Columbia-Suicide Severity Rating Scale: Findings from a prospective, inpatient cohort of severely mentally ill adults. The Journal of Clinical Psychiatry, 77(07), e867-73–e873. doi:10.4088/JCP.15m10069

 PubMedGoogle Scholar

O’Shea, L. E., Picchioni, M. M., Mason, F. L., Sugarman, P. A., & Dickens, G. L. (2014). Predictive validity of the HCR-20 for inpatient self-harm. Comprehensive Psychiatry, 55(8), 1937–1949. doi:10.1016/j.comppsych.2014.07.010

 PubMed Web of Science ®Google Scholar

Quinlivan, L., Cooper, J., Meehan, D., Longson, D., Potokar, J., Hulme, T., … Kapur, N. (2017). Predictive accuracy of risk scales following self-harm: Multicentre, prospective cohort study. British Journal of Psychiatry, 210(6), 429–436. doi:10.1192/bjp.bp.116.189993

 PubMed Web of Science ®Google Scholar

Randall, J. R., Sareen, J., Chateau, D., & Bolton, J. M. (2019). Predicting future suicide: Clinician opinion versus a standardized assessment tool. Suicide and Life-Threatening Behavior, 49(4), 941–951. doi:10.1111/sltb.12481

 PubMed Web of Science ®Google Scholar

Reilly-Harrington, N. A., Shelton, R. C., Kamali, M., Rabideau, D. J., Shesler, L. W., Trivedi, M. H., … Nierenberg, A. A. (2016). A tool to predict suicidal ideation and behavior in bipolar disorder: The Concise Health Risk Tracking Self-Report. Journal of Affective Disorders, 192, 212–218. doi:10.1016/j.jad.2015.12.036

 PubMed Web of Science ®Google Scholar

Saunders, K., Brand, F., Lascelles, K., & Hawton, K. (2014). The sad truth about the SADPERSONS Scale: An evaluation of its clinical utility in self-harm patients. Emergency Medicine Journal, 31(10), 796–798. doi:10.1136/emermed-2013-202781

 PubMed Web of Science ®Google Scholar

Stefansson, J., Nordström, P., Runeson, B., Åsberg, M., & Jokinen, J. (2015). Combining the suicide intent scale and the Karolinska interpersonal violence scale in suicide risk assessments. BMC Psychiatry, 15(1), 226. doi:10.1186/s12888-015-0607-6

 PubMedGoogle Scholar

Villegas, A. C., DuBois, C. M., Celano, C. M., Beale, E. E., Mastromauro, C. A., Stewart, J. G., … Hoeppner, B. B. (2018). A longitudinal investigation of the Concise Health Risk Tracking Self-Report (CHRT-SR) in suicidal patients during and after hospitalization. Psychiatry Research, 262, 558–565. doi:10.1016/j.psychres.2017.09.044

 PubMed Web of Science ®Google Scholar

Wang, Y., Bhaskaran, J., Sareen, J., Bolton, S. L., Chateau, D., & Bolton, J. M. (2016). Clinician prediction of future suicide attempts: A longitudinal study. The Canadian Journal of Psychiatry, 61(7), 428–432. doi:10.1177/0706743716645287

 Web of Science ®Google Scholar

Wu, C. Y., Huang, H. C., Wu, S. I., Sun, F. J., Huang, C. R., & Liu, S. I. (2014). Validation of the Chinese SAD PERSONS Scale to predict repeated self-harm in emergency attendees in Taiwan. BMC Psychiatry, 14(1), 44. doi:10.1186/1471-244X-14-44

 PubMedGoogle Scholar

DATA AVAILABILITY STATEMENT

The authors confirm that the data supporting the findings of this review are available within the article and its supplementary materials.