Whitepapers - GATMEDI

Preprints

Posted on Research Square (Springer Nature), each with a registered DOI and released under CC BY 4.0

Calibrated within-population patient individualisation of antimicrobial exposure: a pre-registered simulation and external-evaluation study

Youssef Ghaly · Research Square (Springer Nature) · 13 July 2026 · DOI 10.21203/rs.3.rs-10299120/v1

A pre-registered validation of model-informed precision dosing for narrow-therapeutic-index antimicrobials. The protocol, pass/fail criteria, sample sizes and analysis were fixed and cryptographically committed before any result was generated, and the engine was content-hashed to produce deterministic output. Across eight published population priors spanning adult and paediatric critical care, posterior credible intervals were calibrated at the 50, 80, 90 and 95% levels, and individualisation reduced external prediction error by 32 to 52% without increasing bias.

View on Research Square Download PDF

Validated cross-source synthesis of structural links in a compiled medical knowledge base: an independent literature check with a discrimination control

Youssef Ghaly · Research Square (Springer Nature) · 23 July 2026 · DOI 10.21203/rs.3.rs-10427287/v1

An independently validated test of cross-source synthesis over a compiled medical knowledge base: whether the engine composes structural links that no single ingested source states, and whether those links hold against external literature the engine did not use. Nine of nine sampled links were confirmed correct by an architecturally independent check, and a discrimination control — a real cross-organ pattern the engine recognised but deliberately withheld — showed the synthesis is selective rather than indiscriminate. The stated claim is reliable synthesis, not discovery.

View on Research Square Download PDF

Proof points

Pre-registered validation studies of the underlying reasoning engine. Each protocol was cryptographically committed before any result was generated. These manuscripts are dated 1 July 2026 and are not yet posted to a preprint server.

Computational validation of a population-pharmacokinetic exposure engine: a pre-registered reproduction study for three antimicrobials

Youssef Ghaly · GATMEDI · Preprint, version 1 July 2026

A pre-registered check that the deterministic exposure engine computes correctly, before any individualisation claim is built on top of it. The protocol, the four checks, the numeric bars and the case set were cryptographically committed before any result was generated. Across seven quantitative cases spanning three antimicrobials the engine reproduced published exposures within the pre-registered 2% tolerance, with a maximum relative error of 1.77%; numerical integration agreed with the closed-form analytic solution to within 0.077%; and all ten cases presenting missing or unusable inputs were refused rather than answered, with no numeric result emitted. Repeated execution produced an identical content hash. The study establishes model-implementation fidelity — that the engine computes exposure correctly given published parameters — and does not address prospective accuracy against new patient data or any clinical-outcome measure.

Download PDF

Pre-registered derivation of a held-out drug’s dosing behaviour from class-analogy anchors: a dual-arm validation with a split outcome

Youssef Ghaly · GATMEDI · Preprint, version 1 July 2026

Whether a system can derive a held-out drug’s dosing behaviour from related drugs alone, without that drug’s own fitted parameters, and reveal rather than conceal where the analogy fails. Both arms, the pass criteria and a physiology-standard size-scaling correction were fixed and committed before any result was generated. In the adult arm, cefepime derived from penicillin and carbapenem anchors reproduced the reference dosing direction under augmented renal clearance: standard 2 g every 8 h reached a median 42.5% of the dosing interval above the MIC against a 65% target, while more-frequent and continuous regimens met it. In the paediatric arm, an adult-to-child derivation returned an honest negative — the available anchors do not span amoxicillin’s clearance (2.32 L/h derived against a size-scaled reference of approximately 5.4 L/h). The contrast, a directional pass where the analogy is clean and a diagnosed negative where the cross-population gap is widest, is the result. Estimates are declared directional predictions and are not externally validated.

Download PDF

Guidelines for Efficient Patient Triage in a Cruise Ship Medical Infirmary

September 2023

This paper presents a decision tree model tailored for the unique healthcare environment on cruise ships, aiming to streamline patient triage in a ship's medical infirmary. The model supports decision-making for phone operators, assisting them to differentiate between administrative and medical concerns, and further to classify these medical concerns based on the patient's status as an adult, paediatric, or elderly patient. A key feature is its ability to determine the most appropriate mode of consultation, whether teleconsultation or face-to-face with the onboard ship's doctor.

Download PDF

Healthcare Provision Review: Efficiency Savings in the NHS

April 2023

Healthcare systems worldwide, including the National Health Service (NHS), face challenges in delivering efficient and sustainable healthcare due to population growth, ageing demographics, and chronic disease prevalence. This paper explores potential efficiency savings within the NHS and state-funded healthcare systems. Key areas of focus include service integration, workforce optimisation, technology adoption, preventative care, value-based healthcare, centralising IT staffing, streamlining coding of diagnoses, and the concept of a "super trust."

Download PDF

Reducing Litigation Costs in the NHS: A Comprehensive Proposal

April 2023

The NHS has faced increasing financial pressure due to rising litigation costs, which divert resources away from direct patient care. This whitepaper proposes a comprehensive framework for implementing a cap on compensation payouts, including a multi-stage compensation model that takes into account the severity of the injury, age of the person, lifestyle changes, future medical care, and lost income. The paper includes an international comparison of litigation costs and analysis of hypothetical medical malpractice cases.

Download PDF

Non-Invasive Continuous Monitoring at the Bedside and Beyond

December 2024

This paper examines advances in non-invasive patient monitoring technologies and their application in clinical environments. The research explores continuous monitoring solutions that can operate at the bedside and in remote settings, addressing the growing need for real-time patient data without the complications associated with invasive procedures. Key areas include wearable sensors, ambient monitoring systems, and integration with existing hospital infrastructure.

Download PDF

Real-World Data Pipelines in Clinical Settings

October 2025

This whitepaper presents a framework for implementing real-world data pipelines within clinical environments. It addresses the challenges of capturing, processing, and utilising clinical data at scale while maintaining data integrity, patient privacy, and regulatory compliance. The paper provides practical guidance for healthcare organisations seeking to leverage real-world evidence to improve clinical decision-making and operational efficiency.

Download PDF

Upcoming Publications

Research currently in preparation

Autonomous Care Pathways in Extreme Environments: Space and Remote Medicine

Forthcoming

The Next Wave of AI-Native Medical Technology

Forthcoming

Closed-Loop Decision Support in Type 1 Diabetes Mellitus

Forthcoming

Designing Integrated, Data-First Healthcare Systems

Forthcoming

A Streamlined Hospital: Blueprint for an AI-Enabled, Lean Clinical Service

Forthcoming

Interested in our research?

Get in touch to discuss how our insights can help your organisation.

Contact Us