Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Public Access Defibrillator Deployment for Cardiac Arrests: A Learn-Then-Optimize Approach with SHAP-based Interpretable Analytics

Chih-Yuan Yang, Keng-Hou Leong, Kexin Cao, Mingchuan Yang, Wai Kin Victor Chan

TL;DR

This work addresses the urgent need to improve AED deployment for OHCA by predicting high-risk areas using only geographic data and then optimizing AED placement with SHAP-informed guidance. It introduces a learn-then-optimize framework comprising a neural predictor trained on POI/building features, SHAP-based interpretability to reveal feature contributions, and a SHAP-guided integer programming model to maximize SHAP-weighted OHCA coverage. The results on Virginia Beach data demonstrate strong predictive signals ($R^2_{train}=0.975$, $R^2_{test}=0.752$) and show that SIP surpasses random deployment in both coverage and survival rate, with robust performance around a 1.2 km minimum spacing and 100 AEDs. The approach offers practical, data-light AED deployment insights with broader applicability to resource distribution in urban emergencies.

Abstract

Out-of-hospital cardiac arrest (OHCA) survival rates remain extremely low due to challenges in the timely accessibility of medical devices. Therefore, effective deployment of automated external defibrillators (AED) can significantly increase survival rates. Precise and interpretable predictions of OHCA occurrences provide a solid foundation for efficient and robust AED deployment optimization. This study develops a novel learn-then-optimize approach, integrating three key components: a machine learning prediction model, SHAP-based interpretable analytics, and a SHAP-guided integer programming (SIP) model. The machine learning model is trained utilizing only geographic data as inputs to overcome data availability obstacles, and its strong predictive performance validates the feasibility of interpretation. Furthermore, the SHAP model elaborates on the contribution of each geographic feature to the OHCA occurrences. Finally, an integer programming model is formulated for optimizing AED deployment, incorporating SHAP-weighted OHCA densities. Various numerical experiments are conducted across different settings. Based on comparative and sensitive analysis, the optimization effect of our approach is verified and valuable insights are derived to provide substantial support for theoretical extension and practical implementation.

Public Access Defibrillator Deployment for Cardiac Arrests: A Learn-Then-Optimize Approach with SHAP-based Interpretable Analytics

TL;DR

This work addresses the urgent need to improve AED deployment for OHCA by predicting high-risk areas using only geographic data and then optimizing AED placement with SHAP-informed guidance. It introduces a learn-then-optimize framework comprising a neural predictor trained on POI/building features, SHAP-based interpretability to reveal feature contributions, and a SHAP-guided integer programming model to maximize SHAP-weighted OHCA coverage. The results on Virginia Beach data demonstrate strong predictive signals (, ) and show that SIP surpasses random deployment in both coverage and survival rate, with robust performance around a 1.2 km minimum spacing and 100 AEDs. The approach offers practical, data-light AED deployment insights with broader applicability to resource distribution in urban emergencies.

Abstract

Out-of-hospital cardiac arrest (OHCA) survival rates remain extremely low due to challenges in the timely accessibility of medical devices. Therefore, effective deployment of automated external defibrillators (AED) can significantly increase survival rates. Precise and interpretable predictions of OHCA occurrences provide a solid foundation for efficient and robust AED deployment optimization. This study develops a novel learn-then-optimize approach, integrating three key components: a machine learning prediction model, SHAP-based interpretable analytics, and a SHAP-guided integer programming (SIP) model. The machine learning model is trained utilizing only geographic data as inputs to overcome data availability obstacles, and its strong predictive performance validates the feasibility of interpretation. Furthermore, the SHAP model elaborates on the contribution of each geographic feature to the OHCA occurrences. Finally, an integer programming model is formulated for optimizing AED deployment, incorporating SHAP-weighted OHCA densities. Various numerical experiments are conducted across different settings. Based on comparative and sensitive analysis, the optimization effect of our approach is verified and valuable insights are derived to provide substantial support for theoretical extension and practical implementation.
Paper Structure (19 sections, 7 equations, 9 figures, 4 tables)

This paper contains 19 sections, 7 equations, 9 figures, 4 tables.

Table of Contents

  1. INTRODUCTION
  2. methodology
  3. Data Preprocessing (Geographic Information System)
  4. Machine Learning Model Construction
  5. SHAP-Based Interpretable Analytics
  6. AED Deployment Optimaization
  7. Experimental Results and Analysis
  8. Data Source
  9. Out-of-Hospital Cardiac Arrest Data
  10. POI and Building Distribution Data
  11. OHCA High-Risk Area Prediction
  12. Model Interpretation
  13. AED Deployment
  14. Visualization of Deployment Decisions
  15. Optimization Effect: SIP v.s. Random Baseline
  16. ...and 4 more sections

Figures (9)

  • Figure 1: Learn-then-optimize framework with SHAP-based interpretable analytics
  • Figure 2: A toy example for the AED deployment optimization problem decisions
  • Figure 3: OHCA occurrences and and prediction results in each H3 Level 7 grid in Virginia Beach.
  • Figure 4: Top 20 geographic features with the highest influence ranked by absolute SHAP values.
  • Figure 5: AED deployment decisions based on the SIP model with $N=100$ and $D_{min}$ = 1.2 km.
  • ...and 4 more figures