Table 3.

Estimates for intervention effects on condom use as a function of sample and study features.

Dimension and Level*	d+ (95% CI)	OR (95% CI)	β	P
Condom skills training per session			0.18	0.0558
60 minutes/session	0.34 (0.11, 0.56)	1.75 (1.20, 2.52)
0 minutes/session	0.092 (0.022, 0.16)	1.16 (1.037, 1.30)
Motivation training per session			0.22	0.0194
46 minutes/session	0.45 (0.18, 0.73)	2.10 (1.35, 3.34)
0 minutes/session	0.11 (0.046, 0.17)	1.20 (1.08, 1.32)
Intervention vs. control reduction in sex frequencies			0.12	0.2050
Intervention group members have much less (d = 1.55)	0.36 (0.0096, 0.70)	1.81 (1.02, 3.17)
Intervention group has less (d = 0.35)	0.19 (0.089, 0.28)	1.37 (1.16, 1.59)
The same amount (d = 0.00)	0.14 (0.077, 0.19)	1.26 (1.14, 1.37)
Control group members have less than intervention (d = −0.35)	0.085 (−0.011, 0.18)	1.15 (0.98, 1.35)

Note. Estimates of effect size values are greater than zero (d₊) or than 1 (OR) for differences that favor increased condom use in the treatment relative to control group and are adjusted for the presence of the other study dimensions. Models used the inverse of the random-effects variance for each effect size as weights (under fixed-effects, all terms were significant, Ps<.001). Terms were zero-centered or contrast-coded prior to estimating values for each extreme and missing values were imputed. These study or sample features reduced the relations of several other carriers to non-significance, including racial composition of the sample, date of study (linear function), dosage interpersonal skills training, and irrelevant content control group. Studies without observations on sex frequencies were imputed at zero. R = 0.32.

^*Levels are extremes observed for the study dimension in question.

CI, Confidence interval. k, number of studies. OR, Odds ratio (transformed from d₊). β, standardized regression coefficient.