Clinical questions are causal: Does this therapy improve outcomes? For which patients? Observational EHR and claims data are abundant; RCTs remain the benchmark but are slow and narrow.

Confounding by indication is central — sicker patients may receive more aggressive treatment. Adjust for baseline severity, comorbidities, and labs; report balance tables; acknowledge unmeasured clinical judgment.

Responsible use

Observational causal estimates inform hypotheses and prioritization; they rarely replace regulatory evidence alone. Pair quantitative results with clinician review and sensitivity analysis.