Latest articles
- Review · Systematic Review & Meta-analysisClinically Significant Errors in Large Language Model-Generated Clinical Summaries: A Systematic Review and Meta-analysis of Omissions, Inaccuracies and Fabrications per Summary
Raphael Malikian
First Contact 2026;1(2):78–117 · November 2026
How often do LLM-generated clinical summaries omit, distort or fabricate — counted per summary, across the published evaluation literature.
- Review · Systematic Review & Meta-analysisUnsafe or potentially unsafe advice from patient-facing large language model chatbots in bench evaluations of primary-care question sets: a systematic review and meta-analysis
Raphael Malikian
First Contact 2026;1(2):118–147 · November 2026
A pooled look at bench evaluations of patient-facing chatbots answering primary-care questions — and how often the advice was judged unsafe.
- Review · Systematic Review & Meta-analysisHigh-dose vitamin C in people with cancer: a Cochrane-level systematic review and meta-analysis of oral and intravenous ascorbate for treatment outcomes, quality of life and harms
Raphael Malikian
First Contact 2026;1(2):148–166 · November 2026 · DOI: 10.5281/zenodo.24213997
Decades of claims for high-dose ascorbate in cancer, tested against the trial evidence for outcomes, quality of life and harms.
- Research · Retrospective AnalysisSynthetic Patients and What They Can and Cannot Teach: A Retrospective Analysis of Open-Source Contributions to Synthea
Raphael Malikian
First Contact 2026;1(2):266–281 · November 2026 · DOI: 10.5281/zenodo.23964017
What building for the Synthea synthetic-patient ecosystem shows about what synthetic data can — and cannot — teach clinical AI.
- Review · Systematic Review & Meta-analysisAmbient artificial intelligence scribes for clinical documentation: effects on documentation time, after-hours work, clinician burnout, note quality and note errors — A systematic review and meta-analysis
Raphael Malikian
First Contact 2026;1(1):1–43 · October 2026 · DOI: 10.5281/zenodo.23565499
Do ambient AI scribes actually save time and reduce burnout — and what do they do to the quality and accuracy of the note?
- Review · Systematic Review & Meta-analysisFabricated and inaccurate citations generated by large language models for biomedical topics: a systematic review and meta-analysis
Raphael Malikian
First Contact 2026;1(1):44–77 · October 2026 · DOI: 10.5281/zenodo.23585717
How often do LLMs invent or corrupt the biomedical references they cite — a meta-analysis of the verification studies.
- Research · Retrospective AnalysisWhen Agent Plumbing Fails Quietly: A Retrospective Analysis of Open-Source Contributions to LLM Tooling and Agent Infrastructure
Raphael Malikian
First Contact 2026;1(1):253–265 · October 2026 · DOI: 10.5281/zenodo.23741622
The failures that don't crash — silent defects in LLM tooling and agent infrastructure, found by contributing to the code itself.
- CommentEighty-Three Percent Growth, Thirty-Eight Papers: Direct Primary Care and the Outcomes Accounting It Owes
Raphael Malikian
First Contact 2026;1(1) · October 2026 · DOI: 10.5281/zenodo.23562906
Direct primary care is growing fast; its published outcomes evidence is not keeping pace. A commentary on the accounting owed.