Optimizing Clinical Fall Risk Prediction: A Data-Driven Integration of EHR Variables with the Johns Hopkins Fall Risk Assessment Tool
Summary
This study developed a data-driven constrained score optimization (CSO) model to enhance the Johns Hopkins Fall Risk Assessment Tool (JHFRAT) for inpatient fall risk prediction. The CSO model, integrating EHR variables while maintaining interpretability, significantly improved predictive performance (AUC-ROC 0.91) compared to the original JHFRAT (AUC-ROC 0.86). This approach offers a robust, evidence-based method for health systems to optimize fall prevention protocols and improve patient safety.
Medical Relevance
Inpatient falls are a leading cause of harm, increased healthcare costs, and extended hospital stays. This research offers a data-driven method to significantly improve the accuracy of fall risk prediction, enabling more precise and proactive interventions to prevent falls, thereby enhancing patient safety and optimizing resource utilization in clinical settings.
AI Health Application
The paper applies machine learning models (constrained score optimization and XGBoost) to patient EHR data and existing clinical assessment scores to optimize and enhance the prediction of inpatient fall risk. This is a direct application of AI/ML for improving clinical decision support, patient safety, and resource allocation within healthcare systems.
Key Points
- Retrospective analysis was conducted on 54,209 inpatient admissions from three Johns Hopkins Health System hospitals between March 2022 and October 2023.
- The study utilized Constrained Score Optimization (CSO) models to integrate JHFRAT assessment data with additional Electronic Health Record (EHR) variables, prioritizing interpretability and clinical knowledge.
- The CSO model demonstrated a significant improvement in predictive performance with an AUC-ROC of 0.91, compared to the current JHFRAT's AUC-ROC of 0.86.
- CSO models performed comparably both with and without the explicit inclusion of additional EHR variables, suggesting a robust foundation within the JHFRAT structure.
- A benchmark black-box XGBoost model achieved a higher AUC-ROC of 0.94 but was less robust to variations in risk labeling compared to the knowledge-based CSO model.
- The CSO model's superior robustness to risk labeling variations makes it potentially more reliable for clinical application despite a slightly lower AUC-ROC than the black-box model.
- This evidence-based approach provides a systematic method for health systems to enhance inpatient fall prevention, improve patient safety, and optimize resource allocation.
Methodology
A retrospective analysis of inpatient admission data was performed. Constrained Score Optimization (CSO) models were applied to integrate JHFRAT assessment scores with supplementary EHR variables. This approach aimed to enhance predictive accuracy while maintaining clinical interpretability. The model's performance was evaluated using AUC-ROC and compared against the existing JHFRAT and a benchmark XGBoost model.
Key Findings
The CSO model significantly improved fall risk prediction (AUC-ROC=0.91) over the current JHFRAT (AUC-ROC=0.86). While a black-box XGBoost model achieved the highest AUC-ROC (0.94), the CSO model demonstrated greater robustness to variations in risk labeling, making it a more reliable and clinically interpretable solution for fall risk assessment.
Clinical Impact
The improved predictive accuracy of the CSO model can lead to more effective and individualized fall prevention strategies for inpatients. This has the potential to reduce the incidence of hospital-acquired falls, minimize patient injuries, lower associated healthcare costs, and optimize the allocation of nursing and preventative resources, ultimately enhancing overall patient safety and quality of care.
Limitations
The abstract does not explicitly state limitations. Potential limitations not detailed here could include generalizability to other health systems, the specific definition and labeling of fall risk encounters, and the retrospective nature of the study.
Future Directions
The abstract does not explicitly state future research directions. Potential future directions might include prospective validation of the CSO model in diverse clinical settings, integration into real-time clinical decision support systems, and evaluating its impact on actual fall rates and patient outcomes.
Medical Domains
Keywords
Abstract
In this study we aim to better align fall risk prediction from the Johns Hopkins Fall Risk Assessment Tool (JHFRAT) with additional clinically meaningful measures via a data-driven modelling approach. We conducted a retrospective analysis of 54,209 inpatient admissions from three Johns Hopkins Health System hospitals between March 2022 and October 2023. A total of 20,208 admissions were included as high fall risk encounters, and 13,941 were included as low fall risk encounters. To incorporate clinical knowledge and maintain interpretability, we employed constrained score optimization (CSO) models on JHFRAT assessment data and additional electronic health record (EHR) variables. The model demonstrated significant improvements in predictive performance over the current JHFRAT (CSO AUC-ROC=0.91, JHFRAT AUC-ROC=0.86). The constrained score optimization models performed similarly with and without the EHR variables. Although the benchmark black-box model (XGBoost), improves upon the performance metrics of the knowledge-based constrained logistic regression (AUC-ROC=0.94), the CSO demonstrates more robustness to variations in risk labelling. This evidence-based approach provides a robust foundation for health systems to systematically enhance inpatient fall prevention protocols and patient safety using data-driven optimization techniques, contributing to improved risk assessment and resource allocation in healthcare settings.
Comments
19 pages, 7 figures, 4 tables