Table 6.

Sensitivity analyses of the syndemic model: Multiplicative interactions between violence victimisation, drug use, and frequent alcohol use in predicting inconsistent condom use with male and hijra partners (N = 22,297), based on alternative categorisation of alcohol use or with exposures specified as continuous variables (multivariable logistic regression).

With alcohol use categorised as any vs none		All exposures specified as continuous variables (scores)
Adverse psychosocial exposures and product terms	Adjusted ORa(95% CI), p value	Adverse psychosocial exposures and product terms	Mean (SD)	Adjusted ORa(95% CI), p value
Violence victimisation (V)	1.46 (1.20, 1.78), p <.001	Violence victimisation score (0–3)a	0.43 (.84)	1.21 (1.11, 1.33), p <0.001
Drug use (D)	1.74 (1.22, 2.48), p = .002	Drug use score	0.16 (.55)	1.13 (0.93, 1.36), p = .20
		(0–3)b
Frequent alcohol use (A)	1.07 (.94, 1.22), p = .28	Alcohol use score	1.56 (2.02)	.98 (0.95, 1.01), p = .27
		(0– 7)c,d
V × D	.63 (.35, 1.13), p = .12	V × D		.94 (0.83, 1.05), p = .30
V × A	1.00 (.77, 1.30), p = .96	V × A		1.00 (0.52, .98), p = .52
D × A	.71 (.47, 1.09), p = .12	D × A		1.00 (0.81, .95), p = .24
V × D × A	2.23 (1.13, 4.40), p = .02	V × D × A		1.03 (1.00, 1.07), p = .03

^aBased on the scoring system: 0 = No experience of violence; 1 = experience of physical violence alone; 2 = experience of sexual violence alone; 3 = experience of physical and sexual violence

^bBased on the scoring system: 0 = No drug use; 1 = Non-injection drug use alone; 2 = Injection drug use alone; 3 = Both non-injection and injection drug use

^cAlcohol consumption in number of days/week

^dWhen a quadratic term for alcohol use was added to this model, the main effect of alcohol became statistically significant (aOR = 1.10; 95% CI, 1.02–1.18) in addition to its significant quadratic term (aOR = .97; 95% CI, .96 to .99).