Table 3.

Regression models predicting severity of prior suicidal ideation and history of suicide^a

Variable	Ideation (range, 0–5)^b		Attempt (yes/no)^c

	b	95% CI	Odds ratio	95% CI
Inhalant (solvent) use
None (ref.)	–	–	–	–
Use only	1.36^**	0.47 to 2.24	1.41	0.84–2.37
Abuse/dependence	1.46^**	0.52 to 2.40	5.13^***	2.73–9.65
Male gender	−0.61^***	−1.06 to −0.15	3.79^***	2.24–6.42
Gender × use only	−1.04^*	−1.97 to −0.11	N/A	N/A
Race
White (Ref.)	–	–	–	–
Black	−0.83^***	−1.16 to −0.51	0.41^**	0.23–0.75
Hispanic	−0.60	−1.24 to 0.04	0.24^*	0.08–0.74
Other	0.52^*	0.06 to 0.98	1.51	0.77–2.94
Psychiatric symptoms	0.02^***	0.02 to 0.03	1.02^***	1.01–1.03
Trauma	0.10^*	0.01 to 0.19	1.20^*	1.03–1.39

Only predictors with statistical significance in at least one model are shown in table. Non-significant variables in both models were age, prior residence (urban/rural) and number of drugs used besides volatile solvents. The linear regression model also contained an interaction term for gender and solvent abuse/dependence.

Multiple linear regression; model F(13, 697) = 32.1, p < .001, adjusted R² = .36.

Logistic regression; model χ²(11) = 199.3, p < .001, pseudo-R² = .25.

p < .05.

^**

p < .01.

^***

p < .001.