Table 3.

Univariate logistic regression analyses based on bootstrap optimism correction.

Variables	AUC	95% CI	Precision	Recall	Brier
ACSS-FAD	0.56	[0.53, 0.60]	0.72	0.58	0.43
Age	0.53	[0.50, 0.56]	0.71	0.44	0.51
AMP—Positive	0.52	[0.48, 0.55]	0.68	0.74	0.47
AMP—Suicide/Self-Injury
Low Intensity	0.52	[0.48, 0.55]	0.68	0.71	0.51
Moderate Intensity	0.51	[0.48, 0.55]	0.68	0.74	0.48
High Intensity	0.53	[0.49, 0.56]	0.69	0.60	0.47
ASQ—Abandonment	0.53	[0.49, 0.56]	0.69	0.58	0.50
ASQ—Anxiety	0.51	[0.49, 0.53]	0.67	0.95	0.37
ASQ—Desperation	0.55	[0.52, 0.58]	0.70	0.72	0.40
ASQ—Guilt	0.51	[0.47, 0.54]	0.67	0.79	0.48
ASQ—Hope	0.50	[0.47, 0.53]	0.66	0.91	0.43
ASQ—Humiliation	0.53	[0.49, 0.56]	0.70	0.52	0.54
ASQ—Loneliness	0.50	[0.47, 0.52]	0.66	0.73	0.61
ASQ—Rage	0.50	[0.47, 0.54]	0.67	0.81	0.47
ASQ—Self-Hate	0.52	[0.49, 0.55]	0.68	0.88	0.39
BAM	0.55	[0.51, 0.58]	0.70	0.60	0.45
BSIb	0.58	[0.55, 0.62]	0.74	0.56	0.42
DWLS—Other	0.55	[0.52, 0.58]	0.71	0.56	0.45
DWLS—Selfa	0.58	[0.55, 0.61]	0.73	0.60	0.41
Employment	0.53	[0.50, 0.57]	0.71	0.59	0.50
Explicit Ratings—Positive	0.52	[0.48, 0.55]	0.68	0.57	0.52
Explicit Ratings—Suicide/Self-Injury	0.53	[0.49, 0.56]	0.69	0.50	0.49
Gender	0.49	[0.48, 0.51]	0.67	0.71	0.67
ISIb	0.59	[0.55, 0.62]	0.74	0.56	0.42
Preparations for Suicidea,b	0.68	[0.65, 0.71]	0.77	0.90	0.25
Confidence in Killing Self during Preparationsa,b	0.73	[0.70, 0.76]	0.83	0.78	0.25
Race	0.52	[0.49, 0.54]	0.70	0.40	0.59
Sexual Orientation	0.51	[0.49, 0.53]	0.74	0.45	0.62
Suicidal desire (BSS)	0.57	[0.54, 0.60]	0.73	0.53	0.45
Suicide Plans	0.53	[0.52, 0.55]	0.68	0.99	0.33
Past Month Frequencyb	0.58	[0.55, 0.61]	0.76	0.50	0.47
Intent on Acting on Plansa,b	0.67	[0.64, 0.70]	0.79	0.69	0.30
Likelihood of Future Plansa	0.57	[0.54, 0.60]	0.72	0.60	0.40

AUC, area under the receiver operating characteristic curve; AUCs of 0.50, chance-level discriminative accuracy; AUCs of 1.0, perfect discriminative accuracy; CI, Confidence Interval; precision, positive predictive value; recall, sensitivity; precision and recall both range from 0 to 1; with higher values indicating better model performance; Brier scores of 0, perfect calibration; with scores closer to 0 indicating better calibration; ACSS, Acquired Capability for Suicide Scale – Fearlessness about Death; AMP, Affect Misattribution Procedure; ASQ, Affective State Questionnaire; BAM, Brief Agitation Measure; BSI, Brief Symptom Inventory; DSWS, Disgust with Self and World Scale; ISI, Insomnia Severity Index; BSS, Beck Scale for Suicidal Ideation; Superscript a indicates the top five variables in the random forest algorithms that yielded the highest mean decrease in accuracy; Superscript b indicates the five most discriminative variables according to the univariate analyses.