How does it address unmeasured confounding?

It implements robust sensitivity analysis tools such as E-values, tipping point analysis, and quantitative bias analysis (QBA) to assess how much unmeasured confounding would be needed to void the results.

What R packages does this skill utilize?

This skill leverages leading R packages including TrialEmulation, MatchIt, WeightIt, AIPW, EValue, and tipr for comprehensive biostatistical analysis.

Does it support external control arms?

Yes, the skill provides specific workflows for integrating historical or real-world controls with clinical trial data using propensity score matching and entropy balancing.

Can this skill handle time-varying confounding?

Yes, it includes specialized methods for Marginal Structural Models (MSMs) and Inverse Probability of Censoring Weighting (IPCW) to manage time-dependent variables in longitudinal data.

Real-World Evidence Analysis in R

Name: Real-World Evidence Analysis in R
Author: choxos

bychoxos

0•

Data Science & ML

Conducts advanced biostatistical analysis of real-world data (RWD) using target trial emulation and propensity score methods.

This skill provides a comprehensive framework for generating high-quality Real-World Evidence (RWE) using R. It implements industry-standard causal inference techniques, including Target Trial Emulation (TTE) via the TrialEmulation package, propensity score weighting and matching, and handling of time-varying confounding. It is particularly useful for biostatisticians and epidemiologists performing comparative effectiveness research or building external control arms from observational datasets, ensuring rigorous analysis of non-randomized data.

Key Features

01Advanced Propensity Score methods including IPW, Overlap Weighting, and Matching

02Time-varying confounding management via Marginal Structural Models (MSMs)

03Implements Target Trial Emulation (TTE) frameworks for observational data analysis

04Support for External Control Arm (ECA) integration and historical control matching

05Automated sensitivity analysis for unmeasured confounding using E-values and tipping points

060 GitHub stars

Use Cases

01Emulating a clinical trial protocol using longitudinal electronic health records (EHR)

02Quantifying the impact of unmeasured confounding on observational study results

03Adjusting for selection bias and confounding in non-randomized comparative effectiveness studies

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add choxos/biostatagent real-world-evidence

For use in Claude.ai and ChatGPT

Key Features

01Advanced Propensity Score methods including IPW, Overlap Weighting, and Matching

02Time-varying confounding management via Marginal Structural Models (MSMs)

03Implements Target Trial Emulation (TTE) frameworks for observational data analysis

04Support for External Control Arm (ECA) integration and historical control matching

05Automated sensitivity analysis for unmeasured confounding using E-values and tipping points

060 GitHub stars

Use Cases

01Emulating a clinical trial protocol using longitudinal electronic health records (EHR)

02Quantifying the impact of unmeasured confounding on observational study results

03Adjusting for selection bias and confounding in non-randomized comparative effectiveness studies