Table 4.

Effect of adverse psychosocial exposures on HIV transmission risk behaviour (inconsistent condom use with male and hijra partners): Multiplicative two-/three-way interactions between violence victimisation, drug use, and frequent alcohol use (N = 22,297).

Adverse psychosocial exposures and product terms	Model 1: Two-way product term (V × D)	Model 2: Two-way product term (V × A)	Model 3: Two-way product term (D × A)	Model 4: All two-way product terms	Model 5: All two- and three-way product terms
	Estimated semi-elasticity (95% CI), p value	Estimated semi-elasticity (95% CI), p value	Estimated semi-elasticity (95% CI), p value	Estimated semi-elasticity (95% CI), p value	Estimated semi-elasticity (95% CI), p value
V × D	0.07 (−0.05, 0.20), p = .26			0.05 (−0.07, 0.18), p = .42	−0.12 (−0.33, 0.09), p = 26
V × A		0.10 (0.008, 0.20), p = .03		0.08 (−0.01, 0.18), p = .10	0.03 (−0.08, 0.15), p = .61
D × A			0.13 (0.02, 0.24), p = .01	0.11 (−0.004, 0.23), p = .05	−0.04 (−0.22, 0.14), p = .64
V × D × A					0.28 (0.10, 0.47), p = .003

Note. 1) The models were adjusted for covariates such as age, education, marital status, sexual identity, forced sex experience during adolescence, HIV risk perception, HIV knowledge, social support and HIV programme exposure. 2) The estimates of the main effects are not shown. The semi-elasticity here is to be interpreted as the percent relative change in the expected value of the outcome (inconsistent condom use) that is associated with the interaction - i.e., the percent relative change in the outcome that can be attributed to the joint effect of two or more exposures, above and beyond their independent associations with the outcome. For example, a semi-elasticity of .10 is interpreted as a 10 percent relative increase in the expected outcome that is associated with the interaction.