Research and Innovation
Confounding Indication Bias Explained
An educational guide to confounding Indication Bias Explained, focused on research literacy rather than hype.
Big idea in one paragraph
Clinicians prescribe based on prognosis-exactly when confounding by indication thrives.
Cardiology registries for PCI, ICDs, and new drugs illustrate treatment selection tied to severity.
Research literacy distinguishes association from causation in real-world data.
Not arguing against registries-arguing for correct inference standards.
Key methods learners should recognize
Cohort studies adjust for measured covariates (EF, creatinine, comorbidity scores).
Instrumental variable analyses occasionally use regional practice variation-assumptions debatable.
Target trial emulation frameworks specify hypothetical RCT protocol applied to observational data.
Negative control outcomes test residual confounding signals.
Landmark findings and honest limitations
SGLT2 initiators after hospitalization may appear sicker at baseline-initiation strategies bias early observational comparisons.
Relative mortality lower in treated arm may reflect survivor selection-inspect new-user designs.
Absolute standardized mortality differences after weighting still uncertain if unmeasured severity differs.
Compare observational HR magnitudes to trial HRs-large divergence hints bias.
Translation barriers
ACC/AHA recommendations prioritize RCTs; registries inform safety and gaps.
Clinic decisions use trials; registries monitor rare harms post-market.
EHR-based studies increasingly clinic-linked-transparency on algorithms needed.
Causal ML methods emerging-not default standard.
Questions that remain open
Sketch directed acyclic graph for ICD indication confounding.
Read examples of target trial emulation in cardiovascular journals.
PubMed ‘confounding by indication heart failure’.
Pair with registry science article in this series.
Putting it together
Indication bias is the central reason observational hero narratives require humility.
Absolute rates and design matter more than fancy statistics labels.
Cardily teaches causal thinking for learners, not registry analytics certification.
Educational disclaimer
This article is part of Cardily’s educational Insights library. Cardily content is for educational and informational purposes only and does not constitute medical advice, diagnosis, or treatment. Always consult a qualified healthcare professional for personal medical concerns. In an emergency, call local emergency services.