Glossary
Definitions of the terms used across the Dr.MaxHealth platform and documentation.
- Biomarker forecasting
- The use of statistical or machine-learning models to project the future trajectory of a clinical biomarker (e.g., LDL-C, HbA1c, ferritin) based on a patient's longitudinal lab history. PULSE reports every projection together with its uncertainty range and the history it was derived from.
- Longitudinal blood panel
- A series of blood test results for the same patient collected over time, enabling trend analysis rather than single-point interpretation. Longitudinal panels are the primary input to Dr.MaxHealth.
- Patient trajectory intelligence
- Software that represents a patient as an ordered longitudinal record, models how that record is changing over time, and prepares evidence-linked findings for physician review. It augments, but does not replace, clinical judgment.
- Clinical decision support (CDS)
- Software that provides clinicians, staff, or patients with knowledge and person-specific information, intelligently filtered or presented to enhance health and care. PULSE prepares findings that require review by a qualified healthcare professional; it does not diagnose, prescribe, or act on patients.
- Uncertainty band
- A visual representation of the prediction interval around a forecasted value, typically a 80% or 95% interval. Showing uncertainty is a hallmark of clinically responsible AI.
- Retrieval-augmented generation (RAG)
- An AI architecture that grounds language-model outputs in retrieved source documents (guidelines, peer-reviewed literature) to reduce hallucination and produce citation-backed evidence snippets.
- Guideline-linked note
- A clinical recommendation generated by Dr.MaxHealth that is hyperlinked to its underlying guideline (e.g., ESC, AHA, NICE, DGK) so the physician can verify the source in one click.
- Calibration curve
- A plot comparing predicted probabilities to observed frequencies. A well-calibrated model's curve sits on the diagonal — essential for trustworthy clinical predictions.
- Mean absolute error (MAE)
- The average magnitude of forecast errors, in the same units as the biomarker. Dr.MaxHealth uses MAE on 7- to 30-day horizons as a primary retrospective validation metric.
- Data drift
- A change in the statistical properties of input data over time (new lab assays, demographic shifts) that can degrade model performance. Continuous drift monitoring is part of our methods-quality stack.
- Bias audit
- A systematic evaluation of model performance across subgroups (sex, age, ethnicity, comorbidities) to detect and mitigate inequitable predictions.
- Clinician-in-the-loop
- A workflow design where AI outputs are always reviewed and acted upon by a licensed clinician, preserving accountability and clinical judgment.
- Decision support, not diagnostic
- A regulatory and product-design distinction: Dr.MaxHealth surfaces patterns and evidence to inform a physician's reasoning but does not output a diagnosis or treatment decision.
- DMH Software UG
- DMH Software UG (haftungsbeschränkt) — the German legal entity (Handelsregister HRB 196754, Hamburg) operating the Dr.MaxHealth platform.
- Hamburg, Germany
- The registered seat of DMH Software UG (Klaus-Groth-Strasse 77, 20535 Hamburg), placing Dr.MaxHealth under EU GDPR and German medical-software regulation.
- GDPR / DSGVO
- The EU General Data Protection Regulation (Datenschutz-Grundverordnung), the regulatory framework governing all processing of patient and user data on the Dr.MaxHealth platform.