Table 3.

Heterogeneous treatment effects by cluster population size, cluster distance to nearest PHC, and household wealth on the primary outcome, prompt treatment within the health sector, during the three-year trial period overall*

	ARC	ARI	C vs I, AORCS (95% CI)	P-value
Cluster distance to PHC
≤5.0 km	0.54	0.55	1.01 (0.84-1.22)	0.918
>5.0 km	0.50	0.54	1.18 (1.01-1.38)	0.039
LR test				0.2193
Cluster population size
<700 people	0.53	0.57	1.18 (0.99-1.41)	0.072
≥700	0.51	0.53	1.07 (0.91-1.24)	0.419
LR test				0.4132
Household wealth†
Less poor	0.53	0.55	1.08 (0.95-1.22)	0.243
Poorest	0.49	0.54	1.23 (1.03-1.46)	0.022
LR test				0.1000

PHC – public health centre, LR – likelihood ratio, AOR_CS – cluster-specific adjusted odds ratio, ARC – absolute risk of events in the control arm, ARI – absolute risk of events in the intervention arm, C – control clusters, CI – confidence interval, I – intervention clusters

*We ran three separate models, one for each of the predefined effect modifiers that included an interaction term between treatment arm and the modifier. We report the results of the LR tests for interaction between arm and modifier in each model. All models controlled for the same covariates as the main model for overall effects during the three-year trial period; we removed the baseline cluster-level summary of wealth in the models that assessed heterogeneous effects by this variable at the household level.

†For 20% of sick child-year observations included in the analysis, their household wealth was measured during the follow-up period rather than at baseline. The co-intervention in both arms to remove user fees could have influenced household wealth in the follow-up period.