Conference Paper/Proceeding/Abstract 135 views
Explainable Machine Learning: Predicting Clinical Outcomes in Welsh Emergency Departments
,
Gareth Jenkins,
Jiaxiang Zhang
Lecture Notes in Computer Science, Volume: 14976, Pages: 290 - 301
Swansea University Authors:
Meg Morgan, Alma Rahat , Jiaxiang Zhang
Full text not available from this repository: check for access using links below.
DOI (Published version): 10.1007/978-3-031-67285-9_21
Abstract
The UK NHS emergency departments (EDs) are the front-line for patient care, with a wide range of patient presentations but limited resources. Using over 1.5 million ED data entries collected during 2015–2023 from a health board in Wales, we explored the application of machine learning models in pred...
| Published in: | Lecture Notes in Computer Science |
|---|---|
| ISBN: | 9783031672842 9783031672859 |
| ISSN: | 0302-9743 1611-3349 |
| Published: |
Cham
Springer Nature Switzerland
2024
|
| Online Access: |
Check full text
|
| URI: | https://cronfa.swan.ac.uk/Record/cronfa67432 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Abstract: |
The UK NHS emergency departments (EDs) are the front-line for patient care, with a wide range of patient presentations but limited resources. Using over 1.5 million ED data entries collected during 2015–2023 from a health board in Wales, we explored the application of machine learning models in predicting clinical outcomes. The features in the models included age, sex, incident type, the reason for attendance, method of arrival, and the County of the ED. First, five supervised learning models were trained and evaluated on data collected before the 3rd quarter of 2019, and the Random Forest classifier outperformed other models with a weighted F1 score of 0.78. The same classifier yielded similar performance on subsequent data collected during 2019–2023. We then evaluated the stability of the classification performance by conducting cross-classification, using quarterly data before, during and after the COVID-19 pandemic (2019–2023). Predictive classification from models trained with historical data is stable during this period. Furthermore, we examined the feature importance (quantified by the mean decrease impurity score) of the fitted model. As expected, incident type (major vs minor) is the most important feature. Interestingly, the importance of age in predicting clinical outcomes varies substantially before and after the 4th quarter of 2019. Together, these results demonstrate the capacity of machine learning and its explainability in predicting ED clinical outcomes from simple features. |
|---|---|
| Keywords: |
Machine learning; Explainable data mining; Emergency departments; Clinical outcomes |
| College: |
Faculty of Science and Engineering |
| Start Page: |
290 |
| End Page: |
301 |