Table 3.

Bradford Hill causality criteria, as applied to plausibility assessment

Criterion	Description	Assumptions
Strength of association	Strong associations are more likely to have causal components than weaker associations.	Associations can be measured
Consistency	Observing similar evaluation results across evaluation methods, over time, and across countries from meta-analyses increases the likelihood of causal relationships.	Results have been measure consistently over time and space
Specificity	Observing an association specific to outcomes of interest among specific groups increases the argument for causal effect.	Malaria interventions are highly likely to reduce all-cause under-five mortality, particularly among vulnerable groups
Temporality	Changes in program must precede changes in disease or coverage outcomes.	Scale-up of interventions has been measured
Gradient	Changes in disease or coverage outcomes increase the same amount for increases to program exposure or intensity.	Coverage has been measured in different geographic areas
Plausibility	Biological plausibility links exposure to intervention with health outcome.	Malaria contributes to all-cause child mortality
Coherence	Causal inference is possible only if the literature or substantive knowledge supports this conclusion	There are documented studies showing that malaria interventions affect mortality
Experiment	Causation is a valid conclusion if researchers have seen observed associations in prior experimental studies.	There are documented studies showing that malaria interventions affect mortality
Analogy	For similar programs operating, similar results can be expected to bolster the causal inference concluded.	Program context has been similar in the past