Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Developing the Temporal Graph Convolutional Neural Network Model to Predict Hip Replacement using Electronic Health Records

Zoe Hancox, Sarah R. Kingsbury, Andrew Clegg, Philip G. Conaghan, Samuel D. Relton

TL;DR

The paper tackles predicting hip replacement risk one year in advance using electronic health records by representing patient Read Code histories as temporal graphs and applying a Temporal Graph Convolutional Neural Network. The approach integrates a sparse 3D CNN with an LSTM to capture short- and long-term temporal patterns, enhanced by a time-encoding scheme and graph regularization, and is evaluated against multiple baselines with recalibration on unseen data. Key contributions include adapting TG-CNN to clinical EHR data, conducting an ablation study, and validating on two unseen datasets with calibration analyses, achieving AUROC 0.724 and AUPRC 0.185 after recalibration. The findings suggest that leveraging patient trajectories in EHRs can improve prediction of hip replacement risk, offering potential for targeted preventive care and optimized resource allocation in health services.

Abstract

Background: Hip replacement procedures improve patient lives by relieving pain and restoring mobility. Predicting hip replacement in advance could reduce pain by enabling timely interventions, prioritising individuals for surgery or rehabilitation, and utilising physiotherapy to potentially delay the need for joint replacement. This study predicts hip replacement a year in advance to enhance quality of life and health service efficiency. Methods: Adapting previous work using Temporal Graph Convolutional Neural Network (TG-CNN) models, we construct temporal graphs from primary care medical event codes, sourced from ResearchOne EHRs of 40-75-year-old patients, to predict hip replacement risk. We match hip replacement cases to controls by age, sex, and Index of Multiple Deprivation. The model, trained on 9,187 cases and 9,187 controls, predicts hip replacement one year in advance. We validate the model on two unseen datasets, recalibrating for class imbalance. Additionally, we conduct an ablation study and compare against four baseline models. Results: Our best model predicts hip replacement risk one year in advance with an AUROC of 0.724 (95% CI: 0.715-0.733) and an AUPRC of 0.185 (95% CI: 0.160-0.209), achieving a calibration slope of 1.107 (95% CI: 1.074-1.139) after recalibration. Conclusions: The TG-CNN model effectively predicts hip replacement risk by identifying patterns in patient trajectories, potentially improving understanding and management of hip-related conditions.

Developing the Temporal Graph Convolutional Neural Network Model to Predict Hip Replacement using Electronic Health Records

TL;DR

The paper tackles predicting hip replacement risk one year in advance using electronic health records by representing patient Read Code histories as temporal graphs and applying a Temporal Graph Convolutional Neural Network. The approach integrates a sparse 3D CNN with an LSTM to capture short- and long-term temporal patterns, enhanced by a time-encoding scheme and graph regularization, and is evaluated against multiple baselines with recalibration on unseen data. Key contributions include adapting TG-CNN to clinical EHR data, conducting an ablation study, and validating on two unseen datasets with calibration analyses, achieving AUROC 0.724 and AUPRC 0.185 after recalibration. The findings suggest that leveraging patient trajectories in EHRs can improve prediction of hip replacement risk, offering potential for targeted preventive care and optimized resource allocation in health services.

Abstract

Background: Hip replacement procedures improve patient lives by relieving pain and restoring mobility. Predicting hip replacement in advance could reduce pain by enabling timely interventions, prioritising individuals for surgery or rehabilitation, and utilising physiotherapy to potentially delay the need for joint replacement. This study predicts hip replacement a year in advance to enhance quality of life and health service efficiency. Methods: Adapting previous work using Temporal Graph Convolutional Neural Network (TG-CNN) models, we construct temporal graphs from primary care medical event codes, sourced from ResearchOne EHRs of 40-75-year-old patients, to predict hip replacement risk. We match hip replacement cases to controls by age, sex, and Index of Multiple Deprivation. The model, trained on 9,187 cases and 9,187 controls, predicts hip replacement one year in advance. We validate the model on two unseen datasets, recalibrating for class imbalance. Additionally, we conduct an ablation study and compare against four baseline models. Results: Our best model predicts hip replacement risk one year in advance with an AUROC of 0.724 (95% CI: 0.715-0.733) and an AUPRC of 0.185 (95% CI: 0.160-0.209), achieving a calibration slope of 1.107 (95% CI: 1.074-1.139) after recalibration. Conclusions: The TG-CNN model effectively predicts hip replacement risk by identifying patterns in patient trajectories, potentially improving understanding and management of hip-related conditions.
Paper Structure (13 sections, 3 figures, 7 tables)

This paper contains 13 sections, 3 figures, 7 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Methods
  4. Cohort Analysis
  5. Data Extraction
  6. Feature Choices
  7. Temporal Graph Representation of EHRs
  8. Model Architecture
  9. Comparison Models
  10. Evaluation Approach
  11. Results
  12. Discussion
  13. Conclusion

Figures (3)

  • Figure 1: Four patient examples (A-D) and their eligibility for the study cohort. Historical records for these patients were included where available. Patient data were included from the start of the analysis period or from their entry into the database if it occurred after April 1, 1999 (B and C). Patients were followed until the end of the analysis period or until they changed to a practice not using SystmOne. A primary hip replacement event was considered incident if the first hip replacement was recorded within the analysis period (C).
  • Figure 2: From raw data to hip replacement risk prediction. Here we show an example sequence for fictitious 'Patient 1' of four Read Codes being recorded across three time steps (visits). Here there are only 5 Read Codes (nodes), however in reality the nodes span to over 500 codes.
  • Figure 3: Calibration curve before and after recalibration (blue: before recalibration on Test 1 data, purple: after recalibration on the unseen (Test 2) data.