Anonymized but Useful Synthetic Tabular Health Data for AI based Fall Risk Assessment

Clinical Snapshot

Key Findings

• P Value: Not reported in the abstract

• Effect Size: Described qualitatively as 'on par' — no quantitative effect size (e.g., AUROC difference, sensitivity/specificity delta) is reported in the abstract

• Primary Outcome: Predictive performance of fall risk classification models trained on SynTabFall (synthetic data) compared to models trained on the original real patient data

• Nnt Or Sensitivity: Not reported in the abstract; for a diagnostic/prognostic tool, sensitivity, specificity, AUROC, and calibration metrics would be the relevant statistics — these are absent from the abstract and require full-paper review

• Confidence Interval: Not reported in the abstract

Clinical Application

As a data resource, SynTabFall is immediately accessible to AI researchers and clinical informaticians without requiring ethics approval or data access agreements. Clinicians wishing to deploy a fall risk model trained on this dataset in practice would still require local validation, institutional governance approval, and integration with existing electronic medical record systems. The accompanying open-source software lowers the barrier for institutions to generate their own synthetic datasets. Australia's National Safety and Quality Health Service (NSQHS) Standards (Standard 8: Preventing Falls and Harm from Falls) mandate systematic fall risk assessment in all acute hospitals. Validated tools such as the Morse Fall Scale and St Thomas's Risk Assessment Tool (STRATIFY) are widely used. An AI-based fall risk model trained on SynTabFall could theoretically complement these tools, but would require validation against Australian inpatient populations before clinical adoption. The Australian Digital Health Agency's Framework for Action and the TGA's emerging Software as a Medical Device (SaMD) regulatory pathway would apply to any clinical deployment. PBS implications are not directly relevant to a data resource paper, though cost-effectiveness of AI-assisted fall prevention is an active area of health technology assessment in Australia. RACGP guidelines on falls prevention in older persons (particularly in primary care and aged care) extend beyond the inpatient focus of this dataset. Hospitalised adult patients at risk of falls, particularly in acute and subacute inpatient settings. The dataset's 44 attributes (demographics, diagnoses, mobility, cognition) align with standard fall risk factor profiles used in clinical practice. Direct applicability to community-dwelling elderly or residential aged care populations is uncertain.

Abstract

Artificial Intelligence (AI) bears potential for improving health care, but this depends on the availability of open-access, realistic, and useful data. To facilitate AI model development in health care we release SynTabFall, a novel synthetic dataset for fall risk assessment. With a total of 745,380 samples and 44 attributes such as demographics, diseases, mobility and cognition related risk factors, this tabular dataset allows for training fall risk prediction models without access to the original patient data. Models trained on our synthetic dataset can reach predictive performance scores in fall risk assessment which are on par with models trained on real data. To support others in sharing data we also describe a process that was developed over multiple years in one of Germany's largest hospitals in close collaboration between data protection officers, health care staff, informaticians and AI engineers. The proposed data sharing approach combines established methods for anonymization and modern generative AI (genAI) methods for synthesizing tabular data and allows for sharing health care data responsibly without sacrificing its utility. We release the synthetic fall risk dataset along with the software developed for synthetic data generation and evaluation.

References

1. 1.Nanevski, I., Jäger, S., Mohebi, M., Schulte-Althoff, M., Pohle, J., Chandler, N., Gubser, R., Nowak, A., Prasser, F., Fürstenau, D., Balzer, F., & Biessmann, F. (2026). Anonymized but useful synthetic tabular health data for AI based fall risk assessment. Scientific Data. https://doi.org/10.14778/3231751.3231757