Comparative effectiveness research in oncology methodology: observational data

Abstract

The goal of comparative effectiveness research is to inform clinical decisions between alternate treatment strategies using data that reflect real patient populations and real-world clinical scenarios for the purpose of improving patient outcomes. Observational studies using population-based registry data are increasingly relied on to fill the information gaps created by lack of evidence from randomized controlled trials. Administrative data sets have many advantages, including large sample sizes, long-term follow-up, and inclusion of data on physician and systems characteristics as well as cost. In this review, we describe the characteristics of many of the commonly used population-based data sets and discuss the elements included within these data sets. An overview of common research themes that rely on population-based data and illustrative examples are presented. Finally, an overview of the analytic techniques commonly employed by health services researchers to limit the effects of selection bias and confounding is discussed. The analysis of well-designed studies of comparative effectiveness is complex. However, careful framing, appropriate study design, and application of sophisticated analytic techniques can improve the accuracy of nonrandomized studies. There are multiple areas where the unique characteristics of observational studies can inform medical decision making and health policy, and it is critical to appreciate the opportunities, strengths, and limitations of observational research.