In making causal inferences about relationships between health outcomes and the use of medicines we need evidence that they covary, and if they do, we need to exclude alternative explanations of the covariation, other than that medicine use is a contributory cause of the improved health outcome. In assessing covariation we are asking whether there has there been a change for the better in a health outcome that is related to medicine use. If health outcomes have improved as medicine use has increased then we need to assess (a) whether there is a statistically reliable relationship between the two and if so, we need to (b) quantify the magnitude of the relationship.
If there is an association between medicine use and improved health outcomes, we then have to evaluate alternative explanations of any relationship that we have observed. If we fail to find a relationship between a population health outcome and medicine use, we need to evaluate alternative explanations of why a relationship may not have been observed before accepting the hypothesis that medicine use failed to improve population health outcomes.
Our capacity to make these inferences depends upon the type of data that are available to assess covariation and the plausibility of alternative explanations. A major distinction can be made between covariation observed in two types of data: linked data on medicine use and health outcomes in individuals; and aggregate data on medicine use in a population and population health outcomes.
Answers to all your Biology Questions
