Population PK + machine learning + MAP Bayesian AUC-guided dosing for adult inpatients across all specialties. Built on 558 MNGHA Jeddah patients. Externally validated on 210.
Launch GuardDose → How it worksFrom a priori dose estimation to MAP Bayesian individualisation, shared patient database, analytics and batch import — one secure institutional platform.
Population PK and a 19-feature ridge model predict individual clearance from routine covariates before any level is drawn. A concordance check flags patients where the two disagree.
One or more levels individualise CL and V with full dose-history superposition, inter-occasion variability, η estimates, OFV and a 90% confidence interval on AUC₂₄.
Target AUC₂₄ 400–600 mg·h/L per the 2020 ASHP/IDSA/SIDP guideline. Regimen grid with predicted AUC, peak and trough for every option; loading-dose guidance.
Cases saved centrally with ward, user, outcome and MAP results. Searchable, filterable, reopen any case, export to CSV.
Target attainment, AUC distribution, cases by ward, month and user, PopPK–ML concordance rate — live from the shared database.
Admin, user and viewer roles; ward assignment; suspend or remove accounts; account lockout; full audit log of sign-ins and data changes.
Import hundreds of patients from Excel or CSV. Column names are matched automatically; each row is validated, calculated (MAP if a trough is supplied) and reported individually.
HTTPS, role-based sign-in, session expiry, audit trail. Patient data is stored in the platform database, not in individual browsers, so the whole team sees the same cases.
Developed and validated on adult inpatients at King Abdulaziz Medical City – Jeddah, with a dedicated model for haematological malignancy and febrile neutropenia.
Every model was developed on MNGHA data and evaluated on an independent external cohort. The population model provides the prior; ridge provides an independent second opinion; MAP individualises.
From prescription to individualised dose adjustment
Age, sex, weight, height, SCr and regimen. CrCL auto-calculated by Cockcroft–Gault; choose the general or HM population.
PopPK and ridge predict CL and AUC₂₄; concordance is checked; the best regimen for the target AUC is suggested. Save to the database.
Trough, peak + trough or any timed samples — at steady state or with the full dose history. Exact sampling time required.
Individual CL, V and AUC₂₄ with 90% CI and uncertainty reduction. New regimen recommended; residuals flagged if inconsistent.
GuardDose is the applied output of a doctoral research programme in pharmacometrics and model-informed precision dosing conducted at MNGHA Jeddah in collaboration with King Abdulaziz University.
PharmD, MSc, BCPS · Developer and principal author · Senior Pharmacist, Department of Pharmaceutical Care, King Abdulaziz Medical City – Jeddah (MNGHA) · PhD candidate, Department of Pharmacology, Faculty of Medicine, King Abdulaziz University
Main PhD supervisor · Department of Pharmacology, Faculty of Medicine, King Abdulaziz University
Principal investigator and co-advisor · King Abdulaziz Medical City – Jeddah, Ministry of National Guard – Health Affairs
Doctoral research advisor · King Abdulaziz University
Doctoral research advisor · Pharmacometrics and clinical pharmacokinetics
IRB approval NRJ24/006/12. Retrospective–prospective vancomycin TDM data from adult inpatients at KAMC-Jeddah; no patient identifiers leave the institutional server.