ABSTRACT
Introduction
To develop fast, intuitive, and accurate suicide risk screening, this study aimed to optimize 0-to-10 point suicide risk rulers.
Methods
662 adult patients from two emergency departments (EDs) completed five risk rulers, two best-practice screeners, and the Columbia Suicide Severity Rating Scale (CSSRS). Using the CSSRS as the reference, we calculated optimized cutoffs, performance metrics, and AUROCs for differentiating between negligible vs. elevated risk and non-high vs. high risk. Performances of rulers and best-practice screeners were compared.
Results
The rulers demonstrated acceptable performance when identifying elevated risk (AUROCs: 0.71–0.81) and stronger performance when identifying high risk (AUROCs: 0.91–0.95). The sadness ruler best predicted elevated risk (AUROC: 0.81 [95% CI: 0.76–0.86]; optimized score: 5 [3–7]; sensitivity: 0.71 [0.50–0.88]; specificity: 0.74 [0.61–0.89]). The suicidal ideation frequency ruler best predicted high risk (AUROC: 0.95 [0.85–1.00], optimized score: 2 [2–5]; sensitivity: 0.89 [0.60–1.00]; specificity: 0.95 [0.93–0.98]). The rulers outperformed a three-item primary screener, but not a longer 10-item primary and secondary screener, which best predicted both risk thresholds.
Conclusions
Suicide risk rulers offer a fast, intuitive primary screening method for initial risk detection with acceptable operating characteristics. Our findings can inform the screening approach that best fits the needs of a given ED.