A Personalized Dosing Strategy Optimization for Diabetes Management: Applications to Gestational Diabetes Mellitus

Presentation Time: 10:30 AM - 10:45 AM

Abstract Keywords: Disease Models, Clinical Decision Support, Chronic Care Management, Patient / Person Generated Health Data (Patient Reported Outcomes), Population Health, Precision Medicine
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics

Type II diabetes mellitus is a disorder that disrupts the way the body uses glucose. It also causes other problems with the way the body stores and processes other forms of energy, including fat. It is managed by close monitoring of blood glucose level while the clinician experiments with dosing strategy based on some clinical guidelines and his/her own experience. In this study, we propose a treatment planning model that optimizes the dosing strategy for diabetes treatment. The model utilizes patient’s personalized characteristics of disease progression and dose response to optimize the drug dosage. Such personalized evidence is estimated by a drug-dose-drug-effect predictive model using the daily blood glucose data recorded during the titration period. We apply these to a group of patients suffering from gestational diabetes. For each patient, drug-dose-drug-effect prediction was established based on the first four weeks of self-monitored blood glucose. The treatment model then individualizes and optimizes dose regimen based on the patient’s personalized drug-dose-drug-effect characteristics. Consistently, the optimized dose regimens use less amount of drug while achieving better glycemic control than the original regimens used to treat the patients. This results in the first mathematical model that is data-driven evidence-based that quantitatively optimizes dosage for the treatment of diabetes. The model can generate personalized dose regimen that has better treatment outcome and more drug efficient. Clinical trials must be conducted to gauge the overall effectiveness.

Speaker(s):
Eva Lee, PhD
Georgia Institute of Technology

Author(s):
Xin Wei, PhD - Georgia Institute of Technology; Michael Wright, MS, MBA, FACHE - Grady Health System;

Early Disease Prediction Using a Text-Numerical Hybrid Model Using Large-Scale Clinical Real-World Data

Presentation Time: 10:45 AM - 11:00 AM

Abstract Keywords: Disease Models, Machine Learning, Clinical Decision Support
Primary Track: Applications
Programmatic Theme: Clinical Informatics

To assist physicians in predicting diseases, most natural language processing (NLP) models have focused on progress notes in electronic medical records with full descriptions from the initial stage of patient diagnosis to the final stage of discharge. However, accurately predicting diseases in the early stage using initial notes is challenging due to limited information. To address this, a text-numerical hybrid method is developed to improve disease prediction accuracy. The method identifies “Reliably predicted diseases (RPD)” that can be robustly predicted in the NLP and Random Forest models even if there are missing values in the numerical data or the amount of text data is small. Results show that, among the predicted disease groups of the two models, diseases matching the RPD are
preferentially adopted and integrated. Precision@10 reveals that our developed method has a relatively higher
accuracy of 67.0% than the traditional NLP model.

Speaker(s):
Ayaka Oka, Master
Fujitsu Limited

Author(s):
Tatsuya Yamaguchi, PharmD - Fujitsu Limited; Masaki Ishihara, MEng - Fujitsu Limited; Takayuki Baba, PhD - Fujitsu Limited; Tatsuya Sato, BIS - Fujitsu Limited; Kazuki Iwamoto, MEng - Fujitsu Limited; Ryo Iwamura, MMedEng - Fujitsu Limited; Shigetaka Toma, AE - Fujitsu Limited; Kaho Ogura, MIST - Fujitsu Limited; Masahiro Kimura, BSc - Fujitsu Limited; Hokuto Morohoshi, MD, PhD - Department of Hygiene, Public Health and Preventive Medicine, Showa University School of Medicine; Akio Nakamura, MD, PhD - Office of Medical Information Technology, Showa University;

Federated Diabetes Prediction in Canadian Adults Using Real-world Cross-Province Primary Care Data

Presentation Time: 11:00 AM - 11:15 AM

Abstract Keywords: Data Mining, Disease Models, Machine Learning, Privacy and Security
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics

Integrating Electronic Health Records (EHR) and the application of machine learning present opportunities for enhancing healthcare service accuracy and accessibility. In particular, developing data-driven machine learning models can provide early identification of patients with high risk for diabetes, potentially leading to more effective therapeutic strategies and reduced healthcare costs. However, regulation restrictions create barriers to developing centralized predictive models. This paper addresses the challenges by introducing a federated learning approach, which amalgamates predictive models without central data storage, thus avoiding privacy issues. This marks the first application of federated learning to predict diabetes using real clinical datasets in Canada from the Canadian Primary Care Sentinel Surveillance Network (CPCSSN) that predicts diabetes diagnosis based on risk factors without cross-province patient data sharing. We address class-imbalance issues through downsampling techniques and compare federated learning performance against province-based and centralized models. This innovative approach demonstrates federated learning's potential to replace traditional centralized models, offering a privacy-conscious, effective solution for leveraging real-life clinical data in diabetes prediction.

Speaker(s):
Guojun Tang, PhD
University of Calgary

Author(s):
Steve Drew, PhD - University of Calgary; Tyler Williamson, PhD - University of Calgary; Jason Black, MSc - University of Calgary;

Multi-center validation of personalized surgical transfusion risk prediction

Presentation Time: 11:15 AM - 11:30 AM

Abstract Keywords: Machine Learning, Surgery, Clinical Decision Support, Healthcare Economics/Cost of Care, Precision Medicine, Patient Safety
Primary Track: Applications
Programmatic Theme: Clinical Informatics

Accurate estimation of surgical transfusion risk guides perioperative planning and effective resource allocation. Few validated methods incorporating patient factors are available to estimate such risk. Here we externally validate a personalized surgical transfusion risk prediction model versus the standard of care using a national sample of 45 hospitals. We find that the model has generalizable performance and can consistently reduce the number of presurgical blood orders needed by one-third compared to the standard of care.

Speaker(s):
Sunny Lou, MD, PhD
Washington University, St. Louis

Author(s):
Sayantan Kumar, PhD Student in Computer Science - Institute for Informatics at Washington University in St. Louis, School of Medicine; Charles Goss, PhD - Washington University School of Medicine; Michael Avidan, MBBCh - Washington University School of Medicine; Sachin Kheterpal, MD, MBA - University of Michigan; Thomas Kannampallil, PhD - Washington University School of Medicine;

Enhancing Semantic and Structure Modeling of Diseases for Diagnosis Prediction

Presentation Time: 11:30 AM - 11:45 AM

Abstract Keywords: Data Mining, Deep Learning, Disease Models, Large Language Models (LLMs)
Primary Track: Applications
Programmatic Theme: Public Health Informatics

Electronic Health Records (EHRs) are valuable healthcare data, aiding researchers and doctors in improving diagnosis accuracy. Researchers have developed several predictive models by learning disease representations to forecast the potential diagnosis that patients may receive. However, existing studies usually ignore the fine-grained semantic and structure information in EHRs (e.g., the hierarchical relations between diseases and ICD-9 codes), which fails to provide accurate disease representation towards effective diagnosis prediction. To this end, we propose to enhance diagnosis prediction through LabCare, a framework with improved semantic and structure modeling of diseases in EHR data. LabCare can simultaneously capture rich semantic and structural relations among diseases and ICD-9 codes, which is achieved by innovatively integrating language models and box embeddings. Extensive experiments on two EHR datasets show that LabCare surpasses competitors, consistently achieving a 4.29% average improvement in Recall and NDCG metrics.

Speaker(s):
Hang Lv, Master
Fuzhou University

Author(s):
Hang Lv, Master - Fuzhou University; Zehai Chen, Undergraduate Student - College of Computer and Data Science, Fuzhou University, China; Yacong Yang, BS - College of Computer and Data Science, Fuzhou University, China; Shuyao Pan, Master's degree - Fujian Medical University; Bo Xiong, PhD - University of Stuttgart; Yanchao Tan, Doctor - Fuzhou University; Carl Yang, PhD;

Developing and Implementing Prediction for Neonatal Opioid Withdrawal Syndrome

Presentation Time: 11:45 AM - 12:00 PM

Abstract Keywords: Disease Models, Clinical Decision Support, Pediatrics
Primary Track: Applications
Programmatic Theme: Clinical Informatics

Neonatal Opioid Withdrawal Syndrome (NOWS) affects nearly 1 in 100 births nationwide, resulting in longer, more complicated hospital stays. Because tools are not available to accurately assess NOWS risk, the American Academy of Pediatrics recommends 3-7 days of observation for opioid exposed infants. We developed and validated accurate prediction models for NOWS diagnosis and treatment to help determine whether an observation window is needed. We determined that predicting NOWS treatment provides the greatest utility.

Speaker(s):
Thomas Reese, PharmD, PhD
Vanderbilt

Author(s):