Interventions to improve treatment adherence among adolescents: A meta-analysis

Table 1 Summary of overall meta-analysis results

Effect of adolescent adherence interventions	K4	Total n5	Unweighted median r (range)	Weighted mean r6 (95%CI)	Unweighted mean r7 (95%CI)	Fail safe n	Standardized odds ratio8 (95%CI)	Standardized relative risk9 (95%CI)
Interventions1	45	3890	0.23 (-0.24-0.71)	0.18 (0.15-0.21)b	0.27 (0.21-0.33)b	2189 (tolerance level 235)	3.03 (2.35, 3.94)b	1.74 (1.53-1.99)b
Adolescent and Children Interventions2	24	1476	0.32 (-0.24-0.71)	0.27 (0.22-0.32)b	0.32 (0.22-0.42)b	800 (tolerance level 130)	3.77 (2.44, 5.99)b	1.94 (1.56, 2.45)b
Adolescent Only Interventions3	21	2414	0.18 (-0.05-0.51)	0.12 (0.08-0.16)b	0.20 (0.13-0.27)b	324 (tolerance level 115)	2.25 (1.69-3.03)b	1.5 (1.30-1.74)b

^bP < 0.01.

¹Analyses for the entire sample of intervention studies;

²Analyses for the subgroup of studies that combined adolescent with pediatric samples;

³Analyses for the subgroup of studies that included adolescent-only samples;

⁴Number of samples;

⁵Total n across all samples;

⁶Effect size obtained from the fixed effects model, or weighted by the total number of participants across studies.

⁷Effect size obtained from the random effects model, or based on the total number of studies included.

⁸The standardized odds ratio depicts the odds of being adherent in the intervention group relative to the control group. Across all analyses, the intervention group had a higher likelihood of improved adherence than the control group (receiving no interventions or standard care).

⁹The standardized relative risk can be interpreted as the control group’s risk for nonadherence as compared to the intervention group’s risk. In each set of analyses, the control group was at significantly greater risk for nonadherence. The Fail Safe n exceeds the level of tolerance for future null results making it unlikely that the “file drawer problem” is a source of bias. The binomial effect size display from the unweighted mean effects (random effects model) was used to obtain the standardized odds ratio and relative risk. The heterogeneity test (k = 45) for the overall adolescent adherence interventions was significant (χ² = 122.82, P < 0.001). The heterogeneity test (k = 24; adolescent and children samples) for adolescent adherence interventions was significant (χ² = 53.36, P < 0.001). The heterogeneity test (k = 21; adolescent only samples) for the second group of adolescent adherence interventions was also significant (χ² = 44.53, P < 0.001).