Table 3.

Multivariate logistic regression analysis of the association between multiple sexual partners and condom use at the last sexual intercourse among 143 sexually active adolescents in four districts in Tanzania, 2008.

Variable	n	Condom use at last sex
		OR	(95% CI)	P-value
Multiple sexual partners in the last 12 months
No	83	Ref.
Yes	60	0.77	(0.35 - 1.67)	0.504
Age (years)
10 - 14	28	Ref.
15 - 19	115	3.69	(1.21 - 11.25)	0.022
District of residence
Rufiji	43	Ref.
Kigoma	15	7.45	(1.79 - 31.06)	0.006
Kilombero	39	8.89	(2.91 - 27.21)	< 0.001
Ulanga	46	5.88	(2.00 - 17.31)	0.001
If you wish, could you get a condom?
No	14	Ref.
Yes	129	4.00	(0.79 - 20.29)	0.095

n = Number of respondents for variable category; OR = Odds ratio; CI = Confidence interval; Ref. = Reference category.

The influence of multiple sexual partners on condom use is adjusted for age, district of residence and whether or not one can obtain a condom if he/she wants. Factors that do not appear in this table (regardless of their significance level in the test of associations in table 2) presented immaterial effect on condom use and this remained impervious when the log likelihood ratio test was used. Since this was somewhat a small sample, the log likelihood ratio test was used to select variables for the multivariate model (rather than relying on the P < 0.2 criterion), except multiple sexual partnerships which was the exposure variable of interest.