There is growing interest in generating evidence from routinely collected real-world data to support medical and regulatory decision-making. However, longitudinal study designs using real-world data are often complex, and text-only descriptions make it difficult for most readers to understand their designs. To address this issue, in 2019, experts from industry, government, and academia developed the “design diagram,” a framework for visualizing longitudinal study designs. The design diagram uses standardized terminology and a graphical structure to communicate study design details to readers, thereby improving reproducibility. Based on previous work by a joint task force between the International Society for Pharmacoepidemiology (ISPE) and the Professional Society for Health Economics and Outcomes Research (ISPOR), the diagram includes a comprehensive set of key study parameters related to reproducibility. It successfully presents study designs in an unambiguous and intuitive manner. Diagrams have been proposed for various study designs, including cohort, nested case-control, and self-controlled designs. Recently, a new diagram was developed that adds at-a-glance elements to show the observability of the source data used in the study. The use of design diagrams is recommended in both the ISPE/ISPOR-endorsed harmonized protocol template (HARPER) and in reporting guidelines for pharmacoepidemiological research, and its widespread use is expected. This paper describes the structure of the design diagram and provides examples of its use. Effective use of design diagrams is expected to improve the reproducibility and reliability of database studies.

View full abstract

A case-control design is one of common study designs in epidemiology. A case with an outcome event of interest is identified and a corresponding control without an outcome event is sampled from a study base, which give rise to the cases. Distribution of a past exposure to an agent of interest before the timing of sampling is compared between a case group and a control group, to yield an odds ratio of exposure as a risk index. A cohort design is usually costly because it requires a large sample size and a long-term follow-up period to power a study especially to detect rare outcome events. In contrast, a traditional case-control design brings efficiency in resource and time to study the association of an exposure and an outcome event by reducing a sample size to study an exposure and covariates after sampling compared with a traditional cohort design. This review article discusses whether a case-control sampling strategy in healthcare database studies, where all the data for the study variables necessary for analysis already exist and are readily available, remains advantageous over a cohort design from the viewpoints of study cost and utility.

View full abstract