RIGOROUS
An Eight-Step Framework
RIGOROUS is an eight-step framework designed to help researchers improve the quality, transparency, and interpretability of studies using EHR data.
Research task identification
Identify estimand(s)
Gauge data fitness
Outline and consider key sources of error, bias and threats to validity
Run appropriate analysis
Outline and assess assumptions
Use appropriate language
Satisfy reporting and transparency standards
The framework starts with encouraging researchers to identify which of four (or five) broad tasks their analysis falls into: description, signal discovery, factual prediction, (counterfactual prediction - if agreed), and causal effect estimation. Based on this, they are guided through the stages of formulating a well-defined research question (i.e. estimand), gauging whether their proposed data are appropriate to answer the question, identifying key sources of bias, planning and running the analysis, assessing assumptions, and finally reporting the study details with appropriate language.
Aim
To produce a step-by-step framework for the analysis of electronic health records (EHR) data to encourage a positive cultural change in the way that EHR research is conceived, conducted, reported, and interpreted.
Vision
This ‘RIGOROUS’ framework is intended to serve as a normative framework for the analysis of EHR data. It is not intended to describe all possible analyses in EHR data but instead serve as an improved standards benchmark for the design, conduct, and reporting of ‘higher quality’ research. The hope is that people will aspire towards adopting the RIGOROUS framework to signal that their work is aiming for a higher standard and may even include it in their title (e.g. ‘a prediction modelling study in EHR data using the RIGOROUS framework’).
Relationship to existing frameworks
The RIGOROUS framework has been designed to serve as an end-to-end methodological framework that integrates key stages of the research pipeline into a single structured workflow. It differs from existing reporting guidelines (e.g. RECORD, TRIPOD) by addressing the full design and conduct pipeline rather than post-hoc reporting, from data quality frameworks (e.g. Kahn et al 2016; the OHDSI Data Quality Dashboard) by linking data fitness assessment to the specific research task and estimand, and from task-specific methodological frameworks (e.g. the Causal Roadmap, target trial emulation, the predictimand framework) by applying across the full taxonomy of research tasks an EHR study might address. RIGOROUS therefore complements rather than replaces existing guidance, cross-referencing these works within the relevant steps of the framework.