Effects of the Informed Health Choices secondary school intervention on the ability of lower secondary students in Kenya to think critically about health choices: 1-year follow-up of a cluster-randomized trial

Table 4 Sensitivity analyses for main results

		Adjusted difference (95% CI)	Odds ratio (95% CI)	p	ICC
Students		2446
Primary outcome^a
Students with a passing score (≥ 9/18)^b	IPW^c	20.8% (13.6 to 28.0%)	2.6 (1.9 to 3.7)	< 0.0001	0.12
Secondary outcomes^a
Students with a mastery score (≥ 14/18)^b	IPW^c	9.9% (6.4 to 13.4%)	8.3 (4.5 to 15.3)	< 0.0001	0.20
Mean score for students^d	IPW^c	8.9% (5.9 to 12.0%)	Effect size^f	< 0.0001	0.13
Lee Bounds^e		8.3 to 10.0% (3.5 to 14.4%)	0.55 (0.36 to 0.73)
Teachers^g		67
Mean score^d		21.5% (14.6–28.5)	Effect size^f
Lee bounds^e		19.0 to 21.4% (8.2 to 30.5%)	1.51 (0.96 to 2.06)

Note: Data are % (SD), % (95% CI), or n (%). ICC intraclass correlation coefficient, IPW inverse probability weighting
^aThe cluster design was accounted for using random intercepts at the level of school
^bLogistic regression was used to estimate an adjusted odds ratio, which is re-expressed as an adjusted risk difference
^cStudents in each school were given a weight equal to the inverse of the proportion of students in the school that completed the CTH test
^dLinear regression was used to estimate an adjusted difference in means
^eThese were constructed by trimming the group with less attrition at the upper and lower tails of the outcome (test score) distribution respectively
^fAdjusted Hedges’ g
^gTeachers were treated as equivalent to the units of randomization (schools), so these models did not include random intercepts. The stratification variables were modelled as fixed effects in all analyses. Wald-type confidence intervals and two-sided normal p-values were computed in all analyses

ISSN: 1745-6215