Core concepts in pharmacoepidemiology: Key biases arising in pharmacoepidemiologic studies

Abstract

Pharmacoepidemiology has an increasingly important role in informing and improving clinical practice, drug regulation, and health policy. Therefore, unrecognized biases in pharmacoepidemiologic studies can have major implications when study findings are translated to the real world. We propose a simple taxonomy for researchers to use as a starting point when thinking through some of the most pervasive biases in pharmacoepidemiology. We organize this discussion according to biases best assessed with respect to the study population (including confounding by indication, channeling bias, healthy user bias, and protopathic bias), the study design (including prevalent user bias and immortal time bias), and the data source (including misclassification bias and missing data/loss to follow up). This tutorial defines, provides a curated list of recommended references, and illustrates through relevant case examples these key biases to consider when planning, conducting, or evaluating pharmacoepidemiologic studies.

New antipsychotic prescription and recurrent infections among adult sepsis survivors: A population‐based cohort study

“…Second, we did not have information on indications for antipsychotics (e.g., psychiatric diagnoses). Thus, existing differences in antipsychotic dose may reflect distinct indications (e.g., high‐dose haloperidol used for psychosis and low‐dose quetiapine for insomnia) resulting in noncomparable groups 10,16 . Third, our study measured the antipsychotic prescriptions filled following hospitalization, and medication use during ICU admission could not be ascertained.…”

“…First, residual and unmeasured confounding are plausible. 16 Nonetheless, we had extensive data on potential confounders; we used a new-user and active comparator design; and we used a negative control outcome to test the validity of our models. 10,15 Second, we did not have information on indications for antipsychotics (e.g., psychiatric diagnoses).…”

“…Thus, existing differences in antipsychotic dose may reflect distinct indications (e.g., high-dose haloperidol used for psychosis and low-dose quetiapine for insomnia) resulting in noncomparable groups. 10,16 Third, our study measured the antipsychotic prescriptions filled following hospitalization, and medication use during ICU admission could not be ascertained. However, the inclusion of new-users of antipsychotics and the short median time to first antipsychotic filling for high-and low-affinity D 2 antipsychotics after discharge suggests a continuation of hospital indications.…”

“…Several limitations must be acknowledged. First, residual and unmeasured confounding are plausible 16 . Nonetheless, we had extensive data on potential confounders; we used a new‐user and active comparator design; and we used a negative control outcome to test the validity of our models 10,15 .…”

Pharmacoepidemiology: An Overview

“…It is also important to remember competitive outcomes. In addition, with the use of secondary databases, other biases have to be taken into account, apart from the biases mentioned above, such as the immortal-time bias and the immeasurable time bias, which can be dealt with either in the design or analysis step [42]. The immortal time bias occurs when using electronic databases, and the exposure occurs at a different time than the inclusion of the individual in the cohort, so there is a period in which the individual cannot present the outcome by design.…”