Dr.MaxHealth Science
Research behind patient trajectory intelligence
Dr.MaxHealth researches longitudinal patient modelling, biomarker forecasting, trained medical orchestration, evidence-grounded reasoning, and physician-led evaluation.
Programme
Six lines of work, one system
The research programme is not six separate topics. Foundations make the record projectable, reasoning is built on top of them, and evaluation and physician collaboration hold both to account.
Foundations
01
Longitudinal patient modelling
Representing the available clinical record as one ordered sequence a model can reason over.
02
Biomarker forecasting
Projecting recurring measurements forward against the patient's own history, with quantified uncertainty.
Reasoning
03
Medical orchestration
Deciding which specialised reasoning, predictive models, and retrieval a case requires, and reconciling their output.
04
Evidence-grounded reasoning
Connecting findings to the record observations they derive from and to relevant medical evidence.
Accountability
05
Medical AI evaluation
Measuring the system inside a review workflow, against defined comparators, on held-out longitudinal data.
06
Physician–AI collaboration
Studying how clinicians confirm, correct, and dismiss findings, and feeding that back into evaluation.
Current work
PULSE is currently being evaluated through retrospective longitudinal datasets, controlled model comparisons, and physician-led case review.
Evaluation
How PULSE is evaluated
Clinical AI has to be tested as a system inside a workflow, against comparators, with clinicians in the loop.
- Held-out data
- Models are assessed on retrospective cases that were not used in development.
- Defined comparators
- Output is compared against defined baselines, including general-purpose foundation models.
- Physician adjudication
- Qualified clinicians review output against defined clinical criteria.
- Omission analysis
- Evaluation measures what a system misses, not only what it gets right.
- Uncertainty evaluation
- Projections are assessed on how well their stated uncertainty matches observed outcomes.
- Documented methods
- Each evaluation records design, dataset period, comparator, endpoints, and evaluation date.
Background
Selected scientific work by the founding team
This work predates Dr.MaxHealth and concerns infectious disease and analytical method development. It establishes the founding team's scientific record; it does not evaluate PULSE.
- 01
Use of Hu-PBL Mice to Study Pathogenesis of Human-Restricted Viruses
Brunetti, J. E., Kitsera, M., Muñoz-Fontela, C., & Rodríguez, E.
Viruses 15(1), 228 · 2023
Review of humanised mouse models used to study human-restricted viral pathogens.
Review · Peer-reviewed · DOI 10.3390/v15010228
- 02
Simultaneous quantification of enterotoxins tilimycin and tilivalline in biological matrices using HPLC high resolution ESMS2
Glabonjat, R. A., Kitsera, M., Unterhauser, K., Lembacher-Fadum, C., Högenauer, C., Raber, G., Breinbauer, R., & Zechner, E. L.
Talanta 222, 121677 · 2021
Analytical method for quantifying bacterial enterotoxins in complex biological matrices.
Journal article · Peer-reviewed · DOI 10.1016/j.talanta.2020.121677
Collaboration
Work with us on longitudinal data
Dr.MaxHealth works with clinical and academic partners on longitudinal cohort analysis, joint evaluation design, and co-authored publication. Research collaborations are scoped individually and governed by a written agreement.
Who we collaborate with
- Universities
- Hospitals and clinical networks
- Medical informatics groups
- Clinical investigators
- Evaluation researchers
- Organizations with longitudinal clinical data