Deep Learning on Hester Davis Scores for Inpatient Fall Prediction

TL;DR

This paper investigates fall risk prediction in hospitalized patients by contrasting traditional threshold-based Hester Davis Score (HDS) methods with two data-driven approaches: one-step ahead prediction using the current HDS value and sequence-to-point prediction leveraging the full history. It formalizes a threshold baseline and several machine learning models for the one-step ahead task, and evaluates recurrent architectures—RNN, LSTM, and GRU—for sequence-to-point fall prediction. Across a 46,695-patient dataset with 4,245 falls, threshold methods yield mixed results, while ML classifiers (SVM, RF, XGB) achieve around 0.63 accuracy and AUC 0.66–0.70 in the one-step setting; in the sequence-to-point setting, GRU attains the best performance with approximately 0.74 accuracy, 0.67 F1, 0.94 specificity, and 0.53 sensitivity. The findings demonstrate that deep learning models, particularly GRUs, better capture temporal patterns in HDS sequences, offering more reliable and timely fall risk predictions that can enhance patient safety and inform prevention strategies.

Abstract

Fall risk prediction among hospitalized patients is a critical aspect of patient safety in clinical settings, and accurate models can help prevent adverse events. The Hester Davis Score (HDS) is commonly used to assess fall risk, with current clinical practice relying on a threshold-based approach. In this method, a patient is classified as high-risk when their HDS exceeds a predefined threshold. However, this approach may fail to capture dynamic patterns in fall risk over time. In this study, we model the threshold-based approach and propose two machine learning approaches for enhanced fall prediction: One-step ahead fall prediction and sequence-to-point fall prediction. The one-step ahead model uses the HDS at the current timestamp to predict the risk at the next timestamp, while the sequence-to-point model leverages all preceding HDS values to predict fall risk using deep learning. We compare these approaches to assess their accuracy in fall risk prediction, demonstrating that deep learning can outperform the traditional threshold-based method by capturing temporal patterns and improving prediction reliability. These findings highlight the potential for data-driven approaches to enhance patient safety through more reliable fall prevention strategies.

Figures (2)

• Figure 1: Proposed schemes for learning from Hester Davis scores (HDSs) in one-step ahead and sequence-to-point modes. The objective is to predict fall event at the prediction timestamp $x_{n,t_{n+1}}$ using prior HDSs.
• Figure 2: Receiver operating characteristic curve (ROC) of models.