Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Early predicting of hospital admission using machine learning algorithms: Priority queues approach

Jakub Antczak, James Montgomery, Małgorzata O'Reilly, Zbigniew Palmowski, Richard Turner

TL;DR

This study addresses emergency department overcrowding by forecasting ward-specific demand stratified by clinical complexity over a 7-day horizon. It compares three predictive paradigms—SARIMAX, XGBoost, and LSTM—using 2017–2021 Australian hospital data, with COVID-period distortions replaced by Prophet-based counterfactuals. Ward-level decomposition and complexity stratification reveal that XGBoost excels for total admissions while SARIMAX slightly outperforms for major complexity, though all models outperform a seasonal naive baseline and share difficulty predicting rare demand surges. The work highlights the practical value for hospital resource planning and suggests ensemble approaches and anomaly-aware loss functions to further improve robustness in face of spikes.

Abstract

Emergency Department overcrowding is a critical issue that compromises patient safety and operational efficiency, necessitating accurate demand forecasting for effective resource allocation. This study evaluates and compares three distinct predictive models: Seasonal AutoRegressive Integrated Moving Average with eXogenous regressors (SARIMAX), EXtreme Gradient Boosting (XGBoost) and Long Short-Term Memory (LSTM) networks for forecasting daily ED arrivals over a seven-day horizon. Utilizing data from an Australian tertiary referral hospital spanning January 2017 to December 2021, this research distinguishes itself by decomposing demand into eight specific ward categories and stratifying patients by clinical complexity. To address data distortions caused by the COVID-19 pandemic, the study employs the Prophet model to generate synthetic counterfactual values for the anomalous period. Experimental results demonstrate that all three proposed models consistently outperform a seasonal naive baseline. XGBoost demonstrated the highest accuracy for predicting total daily admissions with a Mean Absolute Error of 6.63, while the statistical SARIMAX model proved marginally superior for forecasting major complexity cases with an MAE of 3.77. The study concludes that while these techniques successfully reproduce regular day-to-day patterns, they share a common limitation in underestimating sudden, infrequent surges in patient volume.

Early predicting of hospital admission using machine learning algorithms: Priority queues approach

TL;DR

This study addresses emergency department overcrowding by forecasting ward-specific demand stratified by clinical complexity over a 7-day horizon. It compares three predictive paradigms—SARIMAX, XGBoost, and LSTM—using 2017–2021 Australian hospital data, with COVID-period distortions replaced by Prophet-based counterfactuals. Ward-level decomposition and complexity stratification reveal that XGBoost excels for total admissions while SARIMAX slightly outperforms for major complexity, though all models outperform a seasonal naive baseline and share difficulty predicting rare demand surges. The work highlights the practical value for hospital resource planning and suggests ensemble approaches and anomaly-aware loss functions to further improve robustness in face of spikes.

Abstract

Emergency Department overcrowding is a critical issue that compromises patient safety and operational efficiency, necessitating accurate demand forecasting for effective resource allocation. This study evaluates and compares three distinct predictive models: Seasonal AutoRegressive Integrated Moving Average with eXogenous regressors (SARIMAX), EXtreme Gradient Boosting (XGBoost) and Long Short-Term Memory (LSTM) networks for forecasting daily ED arrivals over a seven-day horizon. Utilizing data from an Australian tertiary referral hospital spanning January 2017 to December 2021, this research distinguishes itself by decomposing demand into eight specific ward categories and stratifying patients by clinical complexity. To address data distortions caused by the COVID-19 pandemic, the study employs the Prophet model to generate synthetic counterfactual values for the anomalous period. Experimental results demonstrate that all three proposed models consistently outperform a seasonal naive baseline. XGBoost demonstrated the highest accuracy for predicting total daily admissions with a Mean Absolute Error of 6.63, while the statistical SARIMAX model proved marginally superior for forecasting major complexity cases with an MAE of 3.77. The study concludes that while these techniques successfully reproduce regular day-to-day patterns, they share a common limitation in underestimating sudden, infrequent surges in patient volume.
Paper Structure (14 sections, 7 equations, 7 figures, 3 tables)

This paper contains 14 sections, 7 equations, 7 figures, 3 tables.

Table of Contents

  1. Highlights
  2. Introduction
  3. Literature review
  4. Dataset
  5. Methods
  6. SARIMAX
  7. XGBoost
  8. LSTM
  9. Experiments and results
  10. Admissions with major complexity
  11. Overall admissions
  12. Conclusion
  13. Statements and declarations
  14. Feature importance and SHAP

Figures (7)

  • Figure 1: Daily number of admissions from 2017-01-01 to 2021-12-31, with the COVID-19 period indicated by vertical lines.
  • Figure 2: Daily number of admissions from 2017-01-01 to 2021-12-31, with the COVID-19 period replaced with synthetic data.
  • Figure 3: Architecture of the proposed LSTM model for multistep forecasting.
  • Figure 4: Actual and predicted total arrivals on best week and worst week for major complexity admissions -- comparison between XGBoost, LSTM and SARIMAX.
  • Figure 5: Actual and predicted total arrivals on best week and worst week -- comparison between XGBoost, LSTM and SARIMAX.
  • ...and 2 more figures