At the end of the chapter, the reader will be able to:
Identify the nature of causation and association in pharmacoepidemiology studies
Describe the key criteria to determine causation
Identify the main types and sources of bias in pharmacoepidemiology studies
Explain the concept of confounding
Describe and discuss methods to deal with bias and confounding
Explain the concept of risk adjustment
Describe the methods for adjusting for risk in pharmacoepidemiology studies
One of the main objectives of the discipline of pharmacoepidemiology is to evaluate the use of and the effects of drugs in the postmarketing phase. In other words, pharmacoepidemiology is primarily concerned with the real-world usage and effects of drugs, as opposed to the use and effects in randomized controlled trials. In order to analyze pharmaceuticals in the postmarketing scenario, one must use large databases that are collected for either administrative claims processing or for the purposes of maintaining clinical records. These data are not collected and maintained for research purposes; therefore, they pose special challenges to researchers. Many of these challenges emerge because drugs are used by or prescribed to patients on the basis of a host of patient, provider and societal factors. In addition, drug exposure may be related to factors that may also be associated with its outcome. Furthermore, patient outcomes are generally a result of not only the drug exposure but also a variety of patient and nonpatient characteristics.
More often than not, the treatment and control groups are different from each other in clinical and nonclinical characteristics. These differences can be a range of patient characteristics or risk factors that can influence outcomes. In order to arrive at correct estimates of association between treatment and outcomes, it is important to remove or mitigate the impact of these risk factors. Risk-adjustment measures are used to evaluate the outcomes by statistically controlling for group differences when comparing dissimilar treatment groups.
Analyzing large databases to evaluate associations between treatment and outcomes and interpreting results obtained from such analyses requires understanding of certain key methodological issues in the fields of epidemiology and statistics. In the following sections, several important methodological issues that need to be taken into consideration in investigations of drug treatment and outcomes are discussed.
This chapter begins with a discussion of causation and the criteria that may be used to distinguish causal association from a noncausal association. The next section deals with the issues of bias and confounding commonly encountered in pharmacoepidemiology studies that use large databases. The final section introduces the topic of risk adjustment and outlines some of the key measures and methods of adjusting for risk.
In order to understand the concept of causation, it is important to distinguish it from the related concept of association. A scientific study of evaluation of the relationship between a treatment (i.e., intervention and exposure) and its outcome begins with a selection of participants who ...