[Paper Review] Identifying Risk of Opioid Use Disorder for Patients Taking Opioid Medications with Deep Learning

Xinyu Dong, Jianyuan Deng+7 more|arXiv (Cornell University)|Oct 9, 2020

Machine Learning in Healthcare29 references5 citations

TL;DR

This study proposes a deep learning approach using Long Short-Term Memory (LSTM) networks to predict Opioid Use Disorder (OUD) risk in patients prescribed opioid medications by analyzing their electronic health record (EHR) history. The model achieved an F1 score of 0.8023 and AUROC of 0.9369, outperforming traditional machine learning methods and identifying key clinical features such as OUD-related medications and vital signs.

ABSTRACT

The United States is experiencing an opioid epidemic, and there were more than 10 million opioid misusers aged 12 or older each year. Identifying patients at high risk of Opioid Use Disorder (OUD) can help to make early clinical interventions to reduce the risk of OUD. Our goal is to predict OUD patients among opioid prescription users through analyzing electronic health records with machine learning and deep learning methods. This will help us to better understand the diagnoses of OUD, providing new insights on opioid epidemic. Electronic health records of patients who have been prescribed with medications containing active opioid ingredients were extracted from Cerner Health Facts database between January 1, 2008 and December 31, 2017. Long Short-Term Memory (LSTM) models were applied to predict opioid use disorder risk in the future based on recent five encounters, and compared to Logistic Regression, Random Forest, Decision Tree and Dense Neural Network. Prediction performance was assessed using F-1 score, precision, recall, and AUROC. Our temporal deep learning model provided promising prediction results which outperformed other methods, with a F1 score of 0.8023 and AUCROC of 0.9369. The model can identify OUD related medications and vital signs as important features for the prediction. LSTM based temporal deep learning model is effective on predicting opioid use disorder using a patient past history of electronic health records, with minimal domain knowledge. It has potential to improve clinical decision support for early intervention and prevention to combat the opioid epidemic.

Motivation & Objective

• To develop a predictive model for identifying patients at high risk of developing Opioid Use Disorder (OUD) among those prescribed opioid medications.
• To leverage electronic health record (EHR) data spanning multiple clinical encounters to improve early detection of OUD risk.
• To compare the performance of deep learning models, particularly LSTMs, against traditional machine learning methods such as logistic regression and random forests.
• To identify clinically relevant features from EHR data that contribute most to OUD risk prediction.
• To support clinical decision-making by enabling early interventions to mitigate the opioid epidemic.

Proposed method

• Extracted longitudinal EHR data from the Cerner Health Facts database for patients prescribed opioid-containing medications between January 2008 and December 2017.
• Constructed patient sequences based on the five most recent clinical encounters to capture temporal patterns in health records.
• Applied Long Short-Term Memory (LSTM) networks to model temporal dependencies in EHR data for OUD risk prediction.
• Trained and evaluated multiple models including Logistic Regression, Random Forest, Decision Tree, Dense Neural Network, and LSTM for comparative performance analysis.
• Used feature importance analysis to identify key predictors such as OUD-related medications and vital signs from the trained LSTM model.
• Evaluated model performance using standard metrics: F1 score, precision, recall, and AUROC (Area Under the Receiver Operating Characteristic Curve).

Experimental results

Research questions

• RQ1 Can a deep learning model effectively predict future Opioid Use Disorder (OUD) risk using only a patient’s recent EHR history?
• RQ2 How does the performance of an LSTM-based temporal model compare to traditional machine learning models in predicting OUD risk?
• RQ3 Which clinical features in EHR data are most predictive of OUD development according to the model?
• RQ4 To what extent can minimal domain knowledge be used in a deep learning model to achieve high predictive performance for OUD?
• RQ5 Can the model support early clinical intervention by identifying high-risk patients before OUD onset?

Key findings

• The LSTM-based deep learning model achieved an F1 score of 0.8023 and AUROC of 0.9369, significantly outperforming Logistic Regression, Random Forest, Decision Tree, and Dense Neural Network models.
• The model demonstrated strong generalization and temporal modeling capability, effectively capturing longitudinal patterns in patient EHR data.
• OUD-related medications and vital signs were identified as the most important predictive features in the model’s decision-making process.
• The model required minimal domain knowledge while still achieving high predictive performance, indicating its potential for real-world clinical deployment.
• The results suggest that deep learning models can be effectively used to identify high-risk patients early, enabling timely clinical interventions.
• The study confirms that temporal modeling of EHR data using LSTMs is a promising approach for predicting OUD risk with high accuracy and interpretability.