Large electronic databases containing anonymous health data are often used to study the effects of drugs on human health. In Canada, these databases are available through provincial governments, which are able to link data from publicly administered drug plans, claims data from physicians and hospitals, as well as vital statistics using a unique patient identifier (e.g. personal health number). These governmental administrative databases are very useful to study the relationship between drugs and rare side effects or other unexpected beneficial effects because of their large size, minimal cost compared to collecting data, quick access as data has already been collected, among other benefits. In order to accurately answer questions regarding the relationship between drugs and diseases or side effects, it is essential that individuals be correctly identified according to their drug therapy. Various drug policies exist that limit coverage to certain individuals unless they meet specified criteria (e.g. special authorization). Drugs that are under these policies are not paid for by provincial governments and are excluded from most provincial administrative databases. These policies also change over time potentially leading to further problems in categorizing a person’s drug therapy. In addition to incorrectly classifying people’s drug therapy, these policies may in tern affect the health of the population (e.g. number of heart attacks). We have designed and are working on several projects assessing the effects of drug policies on human health and how these policies will affect research methods using administrative databases. The four primary studies are as follows:
Study One: To measure the amount of misclassification on a population level of numerous drugs and drug classes subject to a policy (or policy change) within various provincial drug databases.
Study Two: To describe and compare the characteristics and health outcomes of individuals who have decided to pay for a drug subject to a cost-containment drug policy to those who do not elect to pay for the drug themselves.
Study Three: To evaluate the effect of a drug policy for a specific class of diabetes medications on heart disease.
Study Four: A simulation study measuring the amount of bias generated from incorrectly identifying individuals’ drug therapy.
JM Gamble, Dean Eurich, Jeffrey Johnson, Finlay McAlister