Interactive AI assistance for predictive and flexible control in discharge and transition management: Development phase (KIAFlex-Research)

Interactive AI assistance for predictive and flexible control in discharge and transition management: Development phase (KIAFlex-Research) - KIAFlex-Research

Status

Recruiting

Phases

Unknown

Study type

Observational

Source

DRKS

Registry ID

DRKS00033139

Enrollment

400000

Registered

2024-09-24

Start date

2024-08-31

Completion date

Unknown

Last updated

2025-10-06

For informational purposes only — not medical advice. Sourced from public registries and may not reflect the latest updates. Terms

Conditions

No specific diseases are examined. The project aims to develop an AI-based assistance system to optimize clinical and administrative processes in discharge management. The prediction models for the system will be trained on clinical cases in order to realistically predict follow-up care needs and discharge dates.

Interventions

Group 1: The AI model to be developed here will be trained using retrospektive inpatient cases in the DRG (Diagnosis-Related Group) fee schedule (somatic medicine) from both the University Hospital Ma

Eligibility

Sex/Gender

All

Age

18 Years to No maximum

Inclusion criteria

Inclusion criteria: Included are all inpatient treatment cases in the german diagnosis-related groups (ger. DRG) fee area (somatic medicine) in the period covered (inpatient cases with discharge from 01.01.2019 - 31.12.2023).

Exclusion criteria

Exclusion criteria: Excluded are all inpatient treatment cases in the compensation system for psychiatric and psychosomatic institutions (ger. PEPP) fee area (psychiatry, psychosomatics, geriatric psychiatry), as different framework conditions apply to social services staff in these areas.

Design outcomes

Primary

Measure	Time frame
The primary endpoint of the study is to develop an AI model with the highest possible sensitivity and specificity in determining a) the potential length of stay, b) the possible need for follow-up care after a hospital stay (yes/no), and c) the type of follow-up care that may be required. Common machine learning metrics, such as accuracy, precision, recall, F1-score, specificity, and area under the receiver operating characteristic curve (AUC-ROC), will be used to train the AI model. The accuracy of the conformal predictions, a machine learning framework for quantifying uncertainty (e.g. the 95% probability that a patient will require follow-up care), will be verified in the next phase of the project (KIAFlex evaluation) in collaboration with hospital staff.	—

Measure

Time frame

The primary endpoint of the study is to develop an AI model with the highest possible sensitivity and specificity in determining a) the potential length of stay, b) the possible need for follow-up care after a hospital stay (yes/no), and c) the type of follow-up care that may be required. Common machine learning metrics, such as accuracy, precision, recall, F1-score, specificity, and area under the receiver operating characteristic curve (AUC-ROC), will be used to train the AI model. The accuracy of the conformal predictions, a machine learning framework for quantifying uncertainty (e.g. the 95% probability that a patient will require follow-up care), will be verified in the next phase of the project (KIAFlex evaluation) in collaboration with hospital staff.

—

Secondary

Measure	Time frame
See primary outcome	—

Countries

Germany

Contacts

Public ContactAlfred Dahmen

Universitätsklinikum Bonn, Abteilung Prozessmanagement Geschäftsbereich 5 - Medizinmanagement

alfred.dahmen@ukbonn.de+49 228 287-51136

Outcome results

None listed

Source: DRKS (via WHO ICTRP) · Data processed: Feb 4, 2026