BMFTR / Zukunft eHealth research consortium

From routine ICU data to explainable decision support.

Charité – Universitätsmedizin Berlin and University Medicine Essen are developing AI-supported dashboards for responsible antibiotic use and early recognition of circulatory instability in intensive care.

Explore the programme Meet the consortium

ICU trajectorytime + context

observationsinterventions

FHIR resource coverageFHIR R4

86%illustrative coverage

research layer

Observation92%

MedicationRequest84%

DiagnosticReport76%

Procedure68%

ICU stays

reference

~200k

development and retrospective evaluation

programme horizon

active

36 mo

coordinated research

workstreams

linked

SP1 through SP4

Governance check

REACT / reviewable

Context travels with the signal

Provenance, quality and clinical interpretation remain linked.

Research pathway / reviewable output

Traceable

FHIR / interoperability

FHIR resource coverage

research layer

Source silos

ICU systems

Signals

Medication

FHIR resources

Observation

Patient / Encounter

Provenance

Routine data / FHIR / governance / decision supportFHIR / reviewable output

Scroll to explore

Mission

A research layer between routine data and decisions at the bedside.

Routine Data Enhanced Analytics for Clinical Trajectories via FHIR-native interoperable Decision Support Systems in the ICU

REACT brings measurements and clinical notes together to help intensive-care teams understand how a patient's condition changes over time. The goal is explainable support within existing workflows, ultimately in real time. Clinical judgement and responsibility remain with the treating clinicians.

FHIR-native by design

Profiles, terminology mappings and reusable resources make ICU analytics transferable between systems.

Governed from source to output

Lineage, access control and quality checks keep every transformation inspectable and accountable.

Built around clinical work

The research outputs are shaped with intensive-care teams, not detached from the decisions they support.

ICU / CONTINUOUS CONTEXT

continuous

A living clinical trajectory.

Critical illness unfolds over time. REACT keeps the changing relationship between observations, interventions and clinical decisions visible.

Research foundation

A common language for intensive-care trajectories.

Measurements and unstructured clinical notes are distributed across many ICU systems. REACT connects them using FHIR (Fast Healthcare Interoperability Resources), an international standard for exchanging health data, while preserving their source, timing and clinical meaning.

Illustration of longitudinal patient trajectories and multimodal ICU data converging into actionable insights — REACT / research view01—02

Monitoring, ventilation, laboratory, medication and clinical notes
FHIR resources with shared profiles and terminology
Traceable transformations that remain open to review

Research programme

A staged route from data foundation to clinical translation.

REACT connects a shared data infrastructure with dashboards for antibiotic therapy and circulatory instability. Around 200,000 ICU stays provide the basis for development and retrospective evaluation before planned clinical testing.

1REACT / RESEARCH

Harmonise

Map heterogeneous ICU routine data into validated, provenance-rich FHIR resources and shared clinical concepts.

Research outputinteroperable data layer

2REACT / RESEARCH

Evaluate

Assess data quality, temporal validity and model performance against retrospective ICU trajectories and clinical reference points.

Research outputtrustworthy evidence

3REACT / RESEARCH

Translate with Impact

Develop explainable dashboards with clinicians and prepare their evaluation in clinical workflows.

Research outputusable decision support

Workstreams / SP1–SP4

Four coordinated research contributions.

Data engineering, trustworthy infrastructure and clinically anchored decision support are developed as one connected research programme.

SP1

Routine-data integration & clinical dashboards

Event-driven ingestion of structured and unstructured ICU data, extraction into provenance-rich FHIR resources, and the development of interoperable web dashboards with clinician-facing CDSS components for reviewing trajectories and supporting decisions.

Principal investigatorDr. rer. medic. René Hosch

InstitutionUniversity Medicine Essen / Institute for Artificial Intelligence in Medicine (IKIM)

SP2

Trustworthy infrastructure

Scalable system architecture for reliable and auditable real-time data flows built on FHIR-native data management, automated conformance testing, policy-as-code data contracts, end-to-end transformation lineage, and fine-grained access control. Provides the technical backbone for all downstream CDSS components.

Principal investigatorDr.-Ing. Elias Grünewald

InstitutionCharité / Institute of Medical Informatics

SP3

Antimicrobial stewardship

Developing an explainable dashboard to assess antibiotic guideline adherence and support de-escalation and therapy adjustment, using computable ICU concepts and patient context.

Principal investigatorDr. med. Mirja Mittermaier

InstitutionCharité / Infectious Diseases, Respiratory Medicine and Critical Care

SP4

Hemodynamic trajectories

Developing temporal models and an explainable dashboard to highlight early warning signs of circulatory instability and show the signals behind each risk assessment.

Principal investigatorDr. med. Nils Daum

InstitutionCharité / Anesthesiology and Intensive Care Medicine

REACT / clinician-facing evidence

Two decision contexts. One governed evidence pathway.

Development and retrospective evaluation draw on around 200,000 ICU stays. A non-interventional pilot at Charité and University Medicine Essen is planned after the project to explore use in clinical workflows and inform possible later regulatory approval and sustained clinical use.

Clinical demonstrator

Antimicrobial stewardship

The dashboard is designed to assess guideline adherence and help teams consider antibiotic de-escalation and therapy adjustment, with explanations grounded in patient context.

guideline adherence
de-escalation
therapy duration
microbiology context

Clinical demonstrator

Hemodynamic trajectories

Learning models are being developed to highlight early warning signs of circulatory instability. The dashboard will show the signals behind each assessment to support clinical judgement.

blood pressure trajectory
vasopressor trend
lactate kinetics
perfusion context

REACT / clinical translation

Research leadership / PIs

Principal investigators connect infrastructure research to clinical validation.

Four investigators from medical informatics and clinical medicine lead the work at Charité and University Medicine Essen. Together, they connect data infrastructure, clinical questions and evaluation with treating teams.