Causal Inference with Real World Data in Healthcare Research: A Scoping Review and associated Study Within a Review–Protocol

Objective: This study aims to improve the understanding and application of causal inference methods in healthcare research using real-world data (RWD). It seeks to systematically synthesize existing methods across different data sources, and assess whether large language models (LLMs) can support key steps in systematic review processes. Introduction: Randomized controlled trials (RCTs) are considered the gold standard for evaluating treatment effectiveness, as randomization balances confounders and enables causal inference. However, RCTs are often costly, time-consuming, or ethically infeasible. As a result, RWD—such as electronic health records (EHR), medical registries, and health insurance claims—are increasingly used in medical research, supported by, for example, regulatory initiatives in the United States and Europe, provided that sufficient data quality is ensured. Complex methods are necessary to derive causal inference from RWD. Inclusion criteria: Studies that use RWD (e.g., claims, registries, EHR), evaluate any type of health outcome as primary endpoint in order to investigate patient health status, address a causal question related to a treatment or intervention, and estimate a causal effect size related to the said causal question. Methods: PubMed, Embase (Elsevier) and Web of Science will be searched for articles published in peer-reviewed journals between 1st January 2010 and 29th January 2026. A scoping review will be conducted to identify and categorize causal inference methods applied to RWD, focusing on their theoretical frameworks, data requirements, and inferential assumptions in order to develop a taxonomy of methods. The frequency, context of use, and quality of reporting will be systematically assessed. In parallel, a study within a review (SWAR) will evaluate the performance of LLMs for title and abstract screening and full-text data extraction compared against human reviewers.