Table 2

Moderators of effect size magnitude for condom use with main partner

Moderator/leveld+ (95% CI)†kβ‡
Study characteristics
Year data collected280.58**
 Earliest (1985)1.69 (1.13 to 2.25)
 Mean (1998)0.80 (0.54 to 1.07)
 Latest (2009)0.05 (−0.43 to 0.53)
HDI280.50**
 Least developed (0.232)1.55 (0.97 to 2.12)
 Mean (0.674)0.78 (0.50 to 1.06)
 Most developed (0.899)0.39 (0.01 to 0.77)
Location0.45**
 U.S.0.41 (0.05 to 0.77)15
 Africa0.98 (0.64 to 1.32)10
Participant demographic characteristics
Length of relationship (months)140.60*
 Minimum (30.48)0.28 (−0.22 to 0.78)
 Mean (67.90)0.75 (0.41 to 1.10)
 Maximum (138.31)1.69 (0.90 to 2.48)
Prop. other STI150.57*
 Minimum (0)0.22 (−0.15 to 0.58)
 Mean (0.13)0.57 (0.35 to 0.80)
 Maximum (0.38)1.25 (0.66 to 1.83)
Sample includes serodiscordant couples0.62**
 No0.28 (−0.28 to 0.40)9
 Yes, discordant and concordant0.59 (0.12 to 1.07)10
 Yes, only discordant1.50 (0.80 to 2.21)9
Intervention characteristics
Number of sessions280.31p=0.06
 Minimum (1)0.56 (0.18 to 0.95)
 Mean (3.79)0.78 (0.48 to 1.09)
 Maximum (18)1.91 (0.68 to 3.14)
Mode of delivery0.51*
 Delivered to individuals0.74 (0.06 to 1.43)4
 Delivered to couples0.65 (0.20 to 1.11)9
 Delivered to groups0.29 (−0.23 to 0.81)7
 Delivered in multiple ways1.40 (0.91 to 1.88)14
Provided VCT0.52**
 Yes1.27 (0.88 to 1.67)12
 No0.42 (0.08 to 0.76)16
Condom skills training0.38**
 Passive training0.42 (−0.34 to 0.49)4
 Active training alone0.36 (0.30 to 0.41)7
 Active training with partner0.60 (0.53 to 0.67)5
First available follow-up (months)280.52**
 Minimum (0.50)0.28 (−0.13 to 0.70)
 Mean (6.93)0.78 (0.03 to 0.13)
 Maximum (24)2.12 (1.24 to 2.99)
Used face-to-face interviews0.43*
 Yes1.05 (0.71 to 1.39)18
 No0.31 (−0.15 to 0.77)10
  • Each moderator listed was evaluated individually without controlling for other listed moderators.

  • †Weighted mean effect sizes (d+) and 95% CIs are based on random-effects assumptions; for continuous moderators, they are estimates produced by moving the constant to the specified levels.

  • ‡Standardised regression weight for continuous moderators; multiple R for categorical models.

  • *p<0.05, **p<0.01.

  • HDI, Human Development Index.