Separate data from evidence
Health records, registries and other care sources generate real-world data. Evidence arises from analysing a defined question. Before comparing treatments, describe inclusion in the analysis, when follow-up starts and how exposure and outcomes are identified.
Investigate why groups differ
Treatment choice in practice may depend on severity, access and comorbidities. These factors also influence outcomes. Examine confounding, missing data, changes over time and sensitivity analyses; adjusting for some variables does not eliminate all bias.
Assess whether the source is fit for purpose
Check registry coverage, traceable definitions and date quality. Ask whether information was recorded for the purpose required by the analysis. Observational studies can complement trials, particularly on context and underrepresented groups, provided their limitations are explicit.
Interpretation limits
Large data volumes do not correct unsuitable selection or missing essential variables.
An observational association does not automatically establish causation.
Original references
Continue exploring
Research questions
Select tools by method, population and purpose. A thematic connection does not validate use in every case.
Educational content for framing research. Consult the original document, study population and methodology before interpreting a conclusion. Methodology