Predictive validity of the Suicide Trigger Scale (STS-3) for post-discharge suicide attempt in high-risk psychiatric inpatients

Abstract

Background: The greatly increased risk of suicide after psychiatric hospitalization is a critical problem, yet we are unable to identify individuals who would attempt suicide upon discharge. The Suicide Trigger Scale v.3 (STS-3), was designed to measure the construct of an affective 'suicide trigger state' hypothesized to precede a suicide attempt (SA). This study aims to test the predictive validity of the STS-3 for post-discharge SA on a high-risk psychiatric-inpatient sample.

Methods: The STS-3, and a psychological test battery measuring suicidality, mood, impulsivity, trauma history, and attachment style were administered to 161 adult psychiatric patients hospitalized following suicidal ideation (SI) or SA. Receiver Operator Characteristic and logistic regression analyses were used to assess prediction of SA in the 6-month period following discharge from hospitalization.

Results: STS-3 scores for the patients who made post-discharge SA followed a bimodal distribution skewed to high and low scores, thus a distance from median transform was applied to the scores. The transformed score was a significant predictor of post-discharge SA (AUC 0.731), and a subset of six STS-3 scale items was identified that produced improved prediction of post-discharge SA (AUC 0.814). Scores on C-SSRS and BSS were not predictive. Patients with ultra-high (90(th) percentile) STS-3 scores differed significantly from ultra-low (10(th) percentile) scorers on measures of affective intensity, depression, impulsiveness, abuse history, and attachment security.

Conclusion: STS-3 transformed scores at admission to the psychiatric hospital predict suicide attempts following discharge among the high-risk group of suicidal inpatients. Patients with high transformed scores appear to comprise two clinically distinct groups; an impulsive, affectively intense, fearfully attached group with high raw STS-3 scores and a low-impulsivity, low affect and low trauma-reporting group with low raw STS-3 scores. These groups may correspond to low-plan and planned suicide attempts, respectively, but this remains to be established by future research.

Figure 1. STS score distributions for a. entire sample, b. subjects lost to follow-up, c. subjects without post discharge suicide attempt, and d. subjects with post-discharge suicide attempt.

Figure 2. ROC curves for prediction of post-discharge SA by median transformed total (AUC = 0.731) and optimal 6-item subscale (AUC = 0.814) scores.