Missing Data Sensitivity (bioSkills)

A Claude Code skill for handling missing endpoint data in confirmatory clinical trials the way regulators expect: an estimand fixed first, MMRM or multiple imputation as the primary analysis, and reference-based and tipping-point sensitivity analyses around it.

   
Type Claude Skill
Supplier GPTomics bioSkills (community OSS, MIT)
Availability GA — part of the bioSkills collection
Pricing Free / OSS (MIT) — the R packages it drives (mmrm and rbmi, Apache-2.0; mice and mitools, GPL-2) and the Python ones (scikit-learn, statsmodels, numpy, pandas, BSD-3) are separately installed OSS
Capabilities Read/Write — Claude runs the skill’s workflow locally (R and Python), not as an MCP tool
Verified works · 2026-08-03
Security cleared · 2026-08-03 — GPTomics/bioSkills MIT confirmed, provenance matches, no advisories

How to install

bioSkills is not an npm package — skills are plain markdown/code read directly by the agent. Clone the repo, then either run the installer for the whole category or copy the single skill directory.

  • Claude Code — clone and install via the bundled script: 
    git clone https://github.com/GPTomics/bioSkills
cd bioSkills
./install-claude.sh --categories "clinical-biostatistics"

    The installer copies matching skills into ~/.claude/skills/ (default target). Use ./install-claude.sh --list to preview the skills first.

  • Claude Code / other agents — copy just this one skill: 
    cp -r bioSkills/clinical-biostatistics/missing-data-sensitivity ~/.claude/skills/

    (run from inside your clone — the previous step left you in bioSkills/; otherwise replace bioSkills/ with the absolute path of your clone). Install the analysis packages on first use, e.g. install.packages(c("mmrm", "rbmi", "mice", "mitools")) in R.

What it does

Six stages, from estimand to submission-ready sensitivity table:

  • Estimand pre-specification — fix the intercurrent-event strategy (treatment policy, hypothetical, composite, while-on-treatment, principal stratum) per ICH E9(R1) before choosing any analysis method.
  • Missingness-mechanism assessment — inspect the CDISC DS domain for differential dropout, run Little’s MCAR test, and document the MAR-versus-MNAR argument in the SAP rather than assuming one.
  • Primary analysis — MMRM under MAR with unstructured covariance and the Kenward–Roger degrees-of-freedom correction for continuous longitudinal endpoints, or multiple imputation under MAR for less regular patterns.
  • Sensitivity analyses — reference-based multiple imputation in the Carpenter–Roger 2013 family (jump-to-reference, copy-reference, copy-increments-in-reference, last-mean-carried-forward); Permutt delta-adjustment and tipping-point analysis; pattern-mixture identifying restrictions (CCMV, NCMV, ACMV).
  • Variance reconciliation — report both Rubin’s (information-anchored) and frequentist (conditional MI plus jackknife) variances for reference-based MI, which is the substance of the Cro-versus-Bartlett debate.
  • Regulatory reporting — express the tipping delta in residual-SD units, articulate what the MNAR assumption means clinically, and keep the SAP’s pre-specified fallback hierarchy.

Software: R mmrm ≥0.3 (Roche/openpharma), rbmi ≥1.5, mice ≥3.16, mitools ≥2.4; Python scikit-learn ≥1.4 (IterativeImputer with sample_posterior=True, BayesianRidge only), statsmodels ≥0.14, numpy, pandas. Imputation count follows von Hippel’s m ≥ 100 × FMI rule for a stable pooled standard error.

Primary use cases: primary-endpoint analysis with dropout, a submission’s missing-data sensitivity package, defending an MNAR assumption to a regulator.

Notes

The skill is explicit about one trap: statsmodels.mixedlm in Python has no Kenward–Roger correction, so it is exploratory only — confirmatory MMRM must go through R mmrm. The upstream framing follows NRC 2010 and ICH E9(R1).

Distributed as a SKILL.md (plus reference material) in the bioSkills collection — Claude executes the workflow locally rather than as an MCP server. The upstream skill front-matter name is bio-clinical-biostatistics-missing-data; if invoked as a namespaced plugin command it resolves under the bioSkills plugin, not as a bare /missing-data-sensitivity. Upstream directory: clinical-biostatistics/missing-data-sensitivity.

Overlaps deliberately with Trial Reporting, which covers the same MMRM/rbmi machinery inside a wider CONSORT reporting workflow; use this skill when the missing-data package is itself the deliverable. Design-stage interim and re-estimation questions belong to Adaptive Designs.

Sources

Installed this tool?

Share feedback — install path, OS, errors, workarounds. The form opens with this tool pre-selected and a link back to this page.