Refining Substance Use Classification: An Ontological Framework for Enhancing Large-Scale Data Collection

Presentation Time: 08:00 AM - 08:12 AM

Abstract Keywords: Controlled Terminologies, Ontologies, and Vocabularies, Knowledge Representation and Information Modeling, Administrative Systems
Primary Track: Applications
Programmatic Theme: Translational Bioinformatics

Substance use disorders (SUD) remain prevalent in the United States. The Office of Addiction Services and Support plays a critical role in tracking SUD trends in New York State and reports data to the federal system. However, ambiguities in substance classification pose challenges to data accuracy and consistency. To address these issues, we developed the foundations for the Addiction Substance Ontology (ASO) using Basic Formal Ontology principles. Definitions in the ASO are expressed in terms of genus and differentiae which form the backbone for a taxonomy of substances in function of their chemical composition and certain other characteristics essential for tracking their acquisition and use. While 143 classes have been developed thus far based on a specific program admission use case, pilot testing and stakeholder collaboration are necessary to refine the ASO and validate its application in real-world settings. These efforts aim to improve data reliability, enhance tracking of SUD patterns, and support effective public health interventions.

Speaker:
Chi-Hua Lu, PharmD/MS
University at Buffalo

Authors:
Kenneth Leonard, PhD - University at Buffalo; Werner Ceusters, MD - Jacobs School of Medicine and Biomedical Science;

---

HIBERT: A Hybrid Clustering BERT for Interpretable Opioid Overdose Risk Prediction

Presentation Time: 08:12 AM - 08:24 AM

Abstract Keywords: Clinical Decision Support, Deep Learning, Machine Learning, Public Health
Primary Track: Foundations
Programmatic Theme: Clinical Research Informatics

Drug overdose, mostly from opioids, is a continuing crisis in the US. Highly accurate models for early detection of opioid overdose (OD) risk are crucial for early intervention and prevention. While deep learning has shown promise in using electronic health records (EHRs) for OD risk prediction, its clinical utility is often limited by challenges posed by data sparsity, heterogeneity, and label imbalance, and lack of interpretability. We present HIBERT, a hybrid BERT model that combines the transformer model with deep clustering. HIBERT uses a multiple BERT architecture integrating specialized BERT modules for distinct EHR feature categories, and incorporates deep significance clustering to generate clinically meaningful risk stratification. HIBERT outperforms conventional and state-of-the-art models based on evaluation with the Health Facts database and identifies four distinct risk clusters, in addition to ranked critical features. It provides actionable, personalized OD risk assessment with improved interpretability.

Speaker:
Zihan Ding, Master's of Science
Stony Brook University

Authors:
Zihan Ding, Master's of Science - Stony Brook University; Xinyu Dong, MS - Stony Brook University; Yinan Liu, Ms. - Stony Brook University; Tengfei Ma, PhD; Xia Zhao, MS - Stony Brook University Hospital; Rachel Wong, MD, MPH, MBA, MS - Stony Brook University, School of Medicine; Richard Rosenthal, MD; Fusheng Wang, Ph.D. - Stony Brook University;

---

Implementation and Assessment of Machine Learning Models for Forecasting Suspected Opioid Overdoses in Emergency Medical Services Data

Presentation Time: 08:24 AM - 08:36 AM

Abstract Keywords: Machine Learning, Evaluation, Data Sharing
Primary Track: Applications
Programmatic Theme: Public Health Informatics

We present efforts in the fields of machine learning and time series forecasting to accurately predict counts of future suspected opioid overdoses recorded by Emergency Medical Services (EMS) in the state of Kentucky. Forecasts help government agencies properly prepare and distribute resources related to opioid overdoses. Our approach uses county and district level aggregations of suspected opioid overdose encounters and forecasts future counts for different time intervals. Models with different levels of complexity were evaluated to minimize forecasting error. A variety of additional covariates relevant to opioid overdoses and public health were tested to determine their impact on model performance. Our evaluation shows that useful predictions can be generated with limited error for different types of regions, and high performance can be achieved using commonly available covariates and relatively simple forecasting models.

Speaker:
Aaron Mullen, B.S.
University of Kentucky

Authors:
Daniel Harris, PhD - University of Kentucky; Peter Rock, MPH - University of Kentucky; Katherine Thompson, PhD - University of Kentucky; Svetla Slavova, PhD - University of Kentucky; Jeffery Talbert, PhD - University of Kentucky; Cody Bumgardner, PhD - University of Kentucky;

---

A Treatment Selection Model for Opioid Use Disorder Using Electronic Health Record and ZIP-Level Data

Presentation Time: 08:36 AM - 08:48 AM

Abstract Keywords: Machine Learning, Clinical Decision Support, Population Health
Primary Track: Applications
Programmatic Theme: Clinical Informatics

Background: Buprenorphine and methadone are effective medications for opioid use disorder (OUD) but remain underused, particularly when specialists are not leading care decisions. Objective: We developed a predictive model to guide treatment selection for OUD. Methods: Models predicted the probability of treatment response for each medication, which we defined as the absence of an adverse outcome during hospitalization and within 90 days of hospital discharge. Models considered electronic health record (EHR) and ZIP-level data. We constructed generalized linear regression, random forest, gradient boosted machines, and deep learning models and tested different combinations of EHR and ZIP-level data using early and late fusion methods. Results: EHR-only models performed better than ZIP-only models did. ZIP-level data did not significantly improve the performance of EHR-only models. Models consistently recommended buprenorphine over methadone. Conclusion: Future work should explore different approaches to modeling OUD treatment response and capturing relevant social and external factors.

Speaker:
Leigh Anne Tang, PhD
Indiana University/Regenstrief Institute

Authors:
Leigh Anne Tang, PhD - Indiana University/Regenstrief Institute; Kristopher Kast, MD - Vanderbilt University Medical Center; Colin Walsh, MD MA - Department of Biomedical Informatics, Vanderbilt University;

---

Prospective Validation of Ensemble Learning Models for Predicting Opioid-Related Overdose Risk Using Tennessee Statewide Data

Presentation Time: 08:48 AM - 09:00 AM

Abstract Keywords: Public Health, Data Modernization, Machine Learning
Primary Track: Applications
Programmatic Theme: Public Health Informatics

Overdose risk prediction models using prescription drug monitoring program data have not been prospectively validated to our knowledge. We prospectively validated fatal and nonfatal opioid-related overdose risk prediction models originally developed using Tennessee statewide data (2012-2017). We used Tennessee statewide data (2022) to assess model discrimination, calibration, and risk concentration. Models achieved similar performance without updating or retraining but were miscalibrated. Risk quantiles captured fentanyl-involved fatal overdoses at a comparable rate to opioid-related fatal overdoses.

Speaker:
Leigh Anne Tang, PhD
Indiana University/Regenstrief Institute

Authors:
Leigh Anne Tang, PhD - Indiana University/Regenstrief Institute; Jessica Korona-Bailey, MPH - Tennessee Department of Health; Michael Ripperger - Vanderbilt University Medical Center; Sutapa Mukhopadhyay, PhD - Tennessee Department of Health; Colin Walsh, MD MA - Department of Biomedical Informatics, Vanderbilt University;

---

Enabling Next Generation Illicit Drug Surveillance Systems using Generative Artificial Intelligence

Presentation Time: 09:00 AM - 09:12 AM

Abstract Keywords: Artificial Intelligence, Public Health, Population Health, Large Language Models (LLMs), Informatics Implementation
Primary Track: Applications
Programmatic Theme: Public Health Informatics

Xylazine, a non-opioid analgesic and sedative approved for acute care veterinary use, is infiltrating unregulated drug markets in the US. In 2023, the White House declared fentanyl adulterated with xylazine as an emerging drug threat and issued a response plan leveraging the Emerging Threats Committee and other vital stakeholders. This declaration was due to evidence of xylazine’s impact on the opioid crisis, including its growing role in overdose deaths in every region of the United States. This declaration was due to evidence of xylazine’s impact on the opioid crisis, including its growing role in overdose deaths in every region of the United States. We harness the power of high-performance transformer models, specifically open foundation large language models (LLMs) to to unlock rich information about emerging drug threats such as xylazine, and other illicit drugs, that is now “locked up” in the clinical progress notes of patient EHRs across the United States. Our framework is now deployed prospectively in the largest integrated healthcare delivery system in the United states, the Department of Veterans Affairs. The results of our system can inform the federal government’s focus, and support clinical decision making at the patient level. Our lessons learned and open-source tools, which are designed for the open foundation model, Llama3.1, can allow other health-care systems, policy makers, and scientists, to accelerate the development of faster and more accurate methods of illicit drug surveillance.

Speaker:
Suzanne Tamang, PhD
Stanford University

Authors:
Ana Maldonado, PhD - Department of Veterans Affairs; David Zhu, BS - Virgina Commonwealth University; Michael Stringer, BS - University of Utah; William Kazanis, Ph.D. - Stanford University; Hannah Eyre - US Department of Veterans Affairs; Michael Slentz, BS - Microsoft; Vilija Joyce, MA - VA; Carla Garcia, MPH - VA; Robin Kinard, MPH - VA; Ashanti Corey, MPH - VA; Elizabeth Olivia, PhD - VA; Elliot Fielstein, PhD - US Dept of Veterans Affairs; Joseph Liberto, PhD - VA; Dominick DePhilppis, PhD - VA; Kamonica Craig, PhD - VA; Joseph Erdos, MD, PhD - Department of Veterans Affairs; Jodie Trafton, PhD - VA; Ruth Reeves - Tennessee Valley Health Care System, US Veterans' Affairs;

---