. 2020 Nov 18;78(2):1–8. doi: 10.1001/jamapsychiatry.2020.3598

Table 3. Logistic Regression Analyses of BPD Criteria as Prospective Factors Associated With SA+ Over 10 Years of Follow-up, Covarying for Individually Significant Demographic and Clinical Variables^a.

Variable	No. (%)			Models BPD criteria separately		Multivariable model with all BPD criteria
Variable	Total (n = 701)	SA+ (n = 148)	SA− (n = 553)	Wald χ²	OR (95% CI)	Wald χ²	OR (95% CI)
Anger	302 (45.6)	102 (68.9)	218 (39.5)	22.36^b	2.88 (1.86-4.47)	1.80	1.44 (0.85-2.45)
Affective instability	362 (51.6)	116 (78.4)	246 (44.5)	25.48^b	3.33 (2.09-5.32)	3.00	1.64 (0.94-2.87)
Emptiness	292 (41.7)	93 (62.8)	199 (36.1)	20.18^b	2.61 (1.72-3.98)	4.36^c	1.63 (1.03-2.57)
Identity disturbance	212 (30.2)	80 (54.1)	132 (23.9)	31.91^b	3.37 (2.21-5.13)	10.59^d	2.21 (1.37-3.56)
Dissociation/paranoid	218 (31.1)	79 (53.4)	139 (25.1)	18.31^b	2.55 (1.66-3.92)	0.20	1.12 (0.68-1.86)
Abandonment	190 (27.1)	69 (46.6)	121 (21.9)	24.28^b	2.99 (1.93-4.62)	6.71^c	1.93 (1.17-3.16)
Impulsivity	318 (45.4)	100 (67.6)	218 (39.4)	19.50^b	2.69 (1.73-4.17)	3.78	1.67 (1.00-2.79)
Unstable relations	267 (38.1)	84 (56.8)	183 (33.1)	11.39^c	2.10 (1.37-3.24)	0.95	0.76 (0.44-1.32)

Abbreviations: BPD, borderline personality disorder; OR, odds ratio; SA+, suicide attempt over 10 years of follow-up; SA−, no suicide attempts over follow-up.

^{^a}

Individually significant relevant demographic and clinical covariates include sex, education, employment, sexual abuse history, posttraumatic stress disorder, alcohol use disorder, drug use disorder, antisocial personality disorder, and obsessive-compulsive personality disorder. Recruitment site was examined as a post hoc control variable. Although suicide attempts were significantly uneven across sites (χ²₃ = 18.95; P <.001), controlling for site as a structural set of dummy coded variables did not change the patterns of significance in any of the models examined. Where not specified, results presented are based on model excluding self-injurious behavior criterion to minimize potential construct contamination.

^{^b}

P < .001.

^{^c}

P < .05.

^{^d}

P < .01.

Table 3. Logistic Regression Analyses of BPD Criteria as Prospective Factors Associated With SA+ Over 10 Years of Follow-up, Covarying for Individually Significant Demographic and Clinical Variablesa.

Table 3. Logistic Regression Analyses of BPD Criteria as Prospective Factors Associated With SA+ Over 10 Years of Follow-up, Covarying for Individually Significant Demographic and Clinical Variables^a.