TABLE 2.

Health Behavior Change Efficacy by Basic and Sample Features in 62 Meta-Analyses

Features of Meta-Analyses	d+i (95% CI)a	βb
Published in journals, % (n = 62)		0.44
100	0.26 (0.25, 0.27)
45	0.01 (0.01, 0.02)
Studies using self-report outcome measures, % (n = 52)		0.16
0	0.23 (0.22, 0.24)
100	0.17 (0.17, 0.18)
Studies using RCT designs, % (n = 54)		0.15
16.7	0.16 (0.14, 0.18)
100	0.27 (0.26, 0.27)
Age of literature, y (n = 62)		0.60
4	0.01 (−0.00, 0.02)
28	0.45 (0.43, 0.46)
Year of meta-analysis (n = 62)		0.22
1988	0.14 (0.13, 0.15)
2003	0.26 (0.25, 0.27)
Mean age of sample, y (n = 42)c		0.67c
10	0.02 (−0.01, 0.05)
18	0.09 (0.07, 0.10)
36	0.14 (0.13, 0.16)
59	0.27 (0.26, 0.29)
Mean % of sample female (n = 43)		0.15
All female (0% male)	0.16 (0.15, 0.17)
All male (0% female)	0.09 (0.07, 0.12)

Note. CI = confidence interval; RCT = randomized controlled trial; d_+i = weighted mean effect size (positive for improvements in the outcome studied and negative for impairments); n = number of meta-analyses. Each model considers each meta-analysis feature independently rather than simultaneous with the other meta-analysis features.

^aDisplayed values are observations at representative low and high extremes of the study dimension in question, where the dimension is represented in its continuous form.

^bAll values in this column reach statistical significance (P < .001) and are standardized regression coefficients, with the 1 footnoted exception.

^cBased on the cubic regression equation, which also includes the linear and quadratic terms; for this model, β represents the multiple R of the 3 terms.