Project Lead:

Prof. Dr. Daniel Fürstenau (Project Lead FUB, ECDF), Matthias Schulte-Althoff (FUB)

Project Team Members:

Project Duration:

08/2022 - 08/2025

Project Description:

AI-based fall prevention in nursing care using privacy-preserving decentralized deep learning approaches.

Falls represent a major problem in the nursing care sector. The application of Artificial Intelligence (AI) in nursing care could help reduce fall incidents through predictive models, as up to 30% of all falls are preventable through appropriate preventive measures (Hshieh et al. 2018).

These systems analyse known risk factors and identify latent risk factors in order to predict fall risks, thereby enabling individual preventive measures. However, existing systems frequently fail to consider relevant risk factors (Seibert et al. 2020). For example, they often rely solely on gait analyses while disregarding changes in medication, even though the risk of falling for many patients is already increased by 56 percent when taking half a daily dose of hypnotics and sedatives. One reason for the incomplete consideration of fall risk factors is the technically and legally complex nature of data access. As a result, the potential of AI systems in nursing care has not yet been fully utilised.

The aim of the KIP-SDM project is to research AI-based fall prevention in nursing care and to develop a decentralized data repository containing nursing treatment data.
To this end, predictive models for fall prediction are being developed using privacy-preserving decentralized deep learning approaches that take all relevant risk factors into account. The data used for this purpose originate from two large care institutions and a startup. To preserve privacy, these data will not leave the respective institutions. Instead, generative deep learning models are intended to learn from the data and generate realistic synthetic patient data. This makes it possible to provide realistic data without having to share actual patient data. Fall prevention should be regarded merely as one example within a broader field of similar nursing care challenges such as pressure ulcers, urinary incontinence, delirium, etc. Using the data integration and data analysis methods developed within the project, as well as the AI application, alternative relevant research questions, data and outcomes could also be evaluated. For the first time, the novel infrastructure enables the development and validation of guideline-compliant AI-based nursing fall prevention across multiple institutions.

For further information, please visit the Charite or the Ai4care website.