![IconGems-Print](images/IconGems-Print.png)
425 – Causal Inference
Development of a Web-based Tool for Comparative Effectiveness Research Using Observational Data
Yi Zhang
MTPPI
Mae Thamer
MTPPI
Onkar Kshirsagar
MTPPI
Core patient-centered outcomes research (PCOR) questions are causal: should I do A or B? To conduct comparative effectiveness (CER) research for causal questions using observational data, two key steps are required: 1. Formulating a well-defined causal question that is relevant to patients and useful for decision making; 2. Providing a valid answer using the best available data and CI analysis methods. Getting both steps right is crucial: a poorly formulated causal question may not lead to an actionable answer even when using high-quality data and cutting edge methodology; an incomplete data or inappropriate analytical approach may lead to a biased answer even for a well defined causal question. Step 1 (formulating the causal question) requires input from PCOR stakeholders including clinicians and patients. Step 2 (providing a valid answer) requires input from researchers with expertise in CER and statistical methods. In practice, there will typically be an iterative process to complete these steps and thus sound PCOR necessitates the close collaboration of stakeholders and researchers throughout the entire research process. In this article, we describe an effort to develop a web-based tool CERBOT (Comparative Effectiveness Research Based on Observational Data to Emulate a Target Trial) that aims to 1) provide guidance and support for CER based on 'real-world' observational data, and 2) facilitates close collaboration and communication between researchers and stakeholders.