[Paper Review] Using Deep Learning and Explainable Artificial Intelligence in Patients' Choices of Hospital Levels

Li‐Chin Chen, Yu Tsao+1 more|arXiv (Cornell University)|Jun 24, 2020

Machine Learning in Healthcare35 references4 citations

TL;DR

This study uses deep learning and explainable AI to predict patients' choices among hospital levels using nationwide insurance data, achieving high performance (AUC: 0.90, accuracy: 0.90) even with highly imbalanced labels. It identifies social approval and physician density as top predictive features and demonstrates that data representation changes improve model accuracy.

ABSTRACT

In countries that enabled patients to choose their own providers, a common problem is that the patients did not make rational decisions, and hence, fail to use healthcare resources efficiently. This might cause problems such as overwhelming tertiary facilities with mild condition patients, thus limiting their capacity of treating acute and critical patients. To address such maldistributed patient volume, it is essential to oversee patients choices before further evaluation of a policy or resource allocation. This study used nationwide insurance data, accumulated possible features discussed in existing literature, and used a deep neural network to predict the patients choices of hospital levels. This study also used explainable artificial intelligence methods to interpret the contribution of features for the general public and individuals. In addition, we explored the effectiveness of changing data representations. The results showed that the model was able to predict with high area under the receiver operating characteristics curve (AUC) (0.90), accuracy (0.90), sensitivity (0.94), and specificity (0.97) with highly imbalanced label. Generally, social approval of the provider by the general public (positive or negative) and the number of practicing physicians serving per ten thousand people of the located area are listed as the top effecting features. The changing data representation had a positive effect on the prediction improvement. Deep learning methods can process highly imbalanced data and achieve high accuracy. The effecting features affect the general public and individuals differently. Addressing the sparsity and discrete nature of insurance data leads to better prediction. Applications using deep learning technology are promising in health policy making. More work is required to interpret models and practice implementation.

Motivation & Objective

• To address inefficient healthcare resource use caused by irrational patient choices in systems allowing provider selection.
• To model patients' decisions on hospital level selection using real-world nationwide insurance data.
• To interpret model predictions using explainable AI (XAI) for both general public and individual-level insights.
• To evaluate the impact of data representation changes on model performance in imbalanced healthcare data.
• To support health policy making with data-driven, interpretable insights into patient behavior.

Proposed method

• Trained a deep neural network on nationwide insurance claims data with features derived from existing literature.
• Applied techniques to handle highly imbalanced label distribution in patient hospital-level choices.
• Used explainable AI methods (e.g., SHAP or LIME-like interpretation) to assess feature contributions at both population and individual levels.
• Explored data representation transformations to improve model generalization and predictive power.
• Evaluated model performance using standard metrics: AUC, accuracy, sensitivity, and specificity.
• Identified top contributing features through feature importance analysis from the trained model.

Experimental results

Research questions

• RQ1 Can deep learning models accurately predict patients' choices among hospital levels despite highly imbalanced data?
• RQ2 Which patient and provider-related features most significantly influence hospital level selection decisions?
• RQ3 How do social perception and local healthcare workforce availability affect patient choices?
• RQ4 To what extent does changing data representation improve prediction performance in this context?
• RQ5 Can explainable AI methods effectively interpret model decisions for both public policy and individual-level understanding?

Key findings

• The deep learning model achieved an area under the ROC curve (AUC) of 0.90, indicating strong discriminatory power.
• Model accuracy reached 0.90, with sensitivity of 0.94 and specificity of 0.97, demonstrating robust performance on imbalanced data.
• Social approval of the hospital by the general public was identified as one of the top influencing features.
• The number of practicing physicians per 10,000 people in the local area was another top predictive feature.
• Modifying data representations led to measurable improvements in model prediction performance.
• Feature importance analysis revealed distinct influence patterns on the general public versus individual decision-making.