Economic evaluations using real-world data: What is missing?

Economic evaluations play a vital role in healthcare reimbursement decisions by providing a systematic and transparent comparison of different healthcare interventions or programs in terms of both their costs and health outcomes. These evaluations enable decision-makers to determine which interventions offer the best "value for money," helping optimize the allocation of limited healthcare resources. Traditionally conducted alongside Randomized Controlled Trials (RCTs), there is a growing interest in incorporating Real-World Data (RWD) into economic evaluations to provide complementary insights. For healthcare decision-makers to make valid reimbursement decisions, it is crucial that economic evaluation results are both reliable and applicable to broader contexts. While RCTs offer high internal validity, their generalizability is often limited, which is why RWD is increasingly recognized as a valuable complement. RWD provides advantages over RCTs, such as better generalizability due to less strict patient selection and a more accurate reflection of actual treatment patterns. However, RWD also comes with its own set of challenges, including confounding and missing data. Confounding, caused by the non-random allocation of treatment, is a major hurdle in RWD analysis and requires careful adjustment to avoid biased results. Additionally, RWD studies tend to have higher rates of missing data compared to RCTs because RWD is typically collected for clinical or administrative purposes, while RCTs are specifically designed for research and include more controls to retain participants. Missing data in RWD can be attributed to various factors, such as time constraints, multiple responsibilities, and no-shows, while RCTs typically have stricter controls over these issues. The goal of this thesis is to explore and evaluate methods for addressing confounding and missing data in economic evaluations using RWD. While various methods have been proposed to handle these challenges, their effectiveness across different scenarios is not fully understood. This thesis aims to clarify terminology, review available methods for adjusting confounding and missing data, and assess the performance of various techniques in the context of economic evaluations using RWD. The first objective was to bridge the gap between epidemiologists and economists by providing a glossary to facilitate interdisciplinary collaboration (Chapter 2). The thesis then presents scoping reviews of methods for addressing time-invariant confounding (Chapter 3) and missing data (Chapter 4). After identifying a range of methods, the focus shifts to evaluating the most promising approaches for handling missing data in different levels of confounding and comparing them to current research practices in economic evaluations (Chapter 5). This exploration extends the research by examining methods for dealing with missing data at both single-level and multilevel data using a real-world dataset (Chapter 6). Finally, the general discussion offers insights and recommendations for improving the validity and reliability of economic evaluations using RWD, based on the scoping reviews, simulations, and real-world data studies included in this thesis (Chapter 7).