A Personal Methods Journal

RWEdnesdays

Real-World Evidence, one Wednesday at a time.
A living repository of methods, operational nuances, and recurring dispatches on evidence that actually moves decisions.

The Repository

Methods Catalog

Causal Inference
Advanced

Target Trial Emulation

Emulating a hypothetical randomized trial in observational data to reduce bias.

Requires explicit protocol (eligibility, treatment strategies, follow-up, outcome, causal contrast) before looking at data.
VIEW FULL ENTRY →
Adherence & Persistence
Intermediate

Proportion of Days Covered (PDC)

Preferred adherence measure for comparative studies; caps at 1.0 and handles overlapping fills via stockpiling rules.

Critical variants: inpatient carry-over, last-day inclusion, death/disenrollment censoring, class vs. molecule level.
VIEW FULL ENTRY →
Bias Control
Advanced

Immortal Time Bias Handling

Bias from misaligned time-zero where "exposed" patients are guaranteed event-free time.

Landmark analysis, cloning/censoring/weighting, or new-user designs with proper index date alignment are required.
VIEW FULL ENTRY →
Quantitative Bias Analysis
Intermediate

E-value Sensitivity Analysis

Minimum strength of unmeasured confounding needed to explain away an observed association.

Report E-value for point estimate and CI limit; pair with negative controls and quantitative bias analysis toolkit.
VIEW FULL ENTRY →
Study Design
Foundational

New-User Design

Restrict to patients initiating treatment after a clean washout period to mimic trial eligibility.

Avoids prevalent-user bias and depletion of susceptibles; requires sufficient lookback for all data sources.
VIEW FULL ENTRY →
Causal Inference
Advanced

Estimands (ATE/ATT + ICE Strategies)

ICH E9(R1) framework: define target population, treatment, summary measure, and intercurrent event handling.

Treatment policy, while-on-treatment, hypothetical, composite, principal stratum — choose before analysis.
VIEW FULL ENTRY →
Full catalog (target trial emulation, PDC variants, immortal time, E-value, estimands, negative controls, and 90+ more) lives here. Powered by the rwe-heor-methods-catalog static export.
Recurring Dispatches

RWEdnesdays

Concise, no-jargon takes on real-world evidence, AI in healthcare, and the messy reality of generating decision-grade evidence. Published (mostly) on Wednesdays.

AI/MLFeb 5, 2026

The Rise of Foundation Models in Healthcare

How large-scale AI models are reshaping clinical decision support, drug discovery, and evidence generation — with real risks around hallucination and bias amplification.

READ FULL DISPATCH →
RegulatoryJan 20, 2026

FDA's Evolving Stance on Real-World Evidence

Recent guidance updates signal growing acceptance of RWE for label expansion and safety monitoring. What sponsors actually need to submit vs. what reviewers require.

READ FULL DISPATCH →
MethodsJan 12, 2026

Best Practices for EHR-Based Studies

Phenotype validation, data quality assessment, confounding control, and sensitivity analyses that strengthen causal claims from electronic health record data.

READ FULL DISPATCH →
Follow the full series on LinkedIn or browse the archive on this site.