Current Projects
Learn more about the CACHE program’s portfolio of projects at https://aihealth.duke.edu/cache-projects/
Events in the Duke AI Health Community of Practice
Learn more about AI Health’s portfolio of summits, symposia, poster showcases, writing workshops, machine learning schools, and data studios at https://aihealth.duke.edu/community-of-practice/Evaluation for the Duke AI Health Community of Practice
Continuous evaluation with the goal to maximize the impact and effectiveness of the AI Health Community of Practice. READ MORE
Predictive Accuracy of Stroke Prediction Models
Duke AI Health and American Heart Association researchers are harmonizing data from diverse patient cohorts and developing new machine learning-based models to improve stroke risk prediction. READ MORE
Cardiovascular Disease Risk Prediction for Women with Adverse Pregnancy Outcomes
Collaborators from Duke AI Health, UNC, and Brigham and Women’s Hospital are applying machine learning methods to predict and prevent cardiovascular disease in women with adverse pregnancy outcomes. READ MORE
Interpretable Long-Horizon Diagnosis Prediction via Learned Retrieval of Predictive Descriptions in Clinical Notes
Clinical notes contain descriptive findings not found in structured EHR fields that are relevant to early prediction of slowly evolving conditions, including neurodevelopmental conditions and chronic diseases.
PCORI HSII: A Multimodal Approach To Improving Antibiotic Prescribing for Pediatric ARTI
The primary goal of our implementation project is to improve guideline concordant-prescribing for acute upper respiratory tract infection (ARTI) in children seen at outpatient clinical sites within our health system. READ MORE
PCORI HSII: Implementation of Electronic Patient-Reported Outcomes Monitoring During Cancer Treatment
In this project, Duke University Health System will put in place an electronic patient-reported outcomes (ePROs) monitoring program to improve symptom management for patients with cancer at sites in one state, drawing on evidence from PCORI-funded research. READ MORE
Debiasing Clinical Care Algorithms: Evaluating PREVENT Equation Across Demographic and Socioeconomic Groups
This project evaluates the American Heart Association’s PREVENT cardiovascular risk equations using large-scale EHR data from Duke Health and national sources. READ MORE Evaluation of AI Tools in Healthcare
This project develops and applies structured frameworks to evaluate generative artificial intelligence (AI), especially large language models (LLMs) in clinical settings. We combine qualitative assessments with automated metrics to assess linguistic quality, completeness, and trustworthiness of AI-generated outputs. READ MORE Predicting Psychosocial Distress in Glaucoma Patients
Living with glaucoma can take a serious emotional toll on patients, with anxiety, depression, and other forms of psychosocial distress often going unrecognized in routine eye care. These emotional challenges can impact how patients manage their condition and may influence long-term outcomes. READ MORE
Investigating Patient-Facing Usage of Large Language Models
People are increasingly receiving healthcare information from large language models (LLMs), such as via chatbots and AI-augmented search engines. Despite its commonality, the nature of these patient-LLM interactions remain largely unexplored. READ MORE
Understanding the Clinical Data Behind Large Language Models
In recent years, there has been significant excitement around the application of large language models (LLMs) to diverse clinical applications. Recent research has found that general domain models can perform just as well as these medically finetuned LLMs on standard benchmarks, despite being trained only on general online corpora. READ MORE A Steamlined Pipeline for Predicting Drug-Targeted Interactions
Drug discovery remains one of the most resource-intensive endeavors in biomedicine, with estimates placing the average cost of bringing a single new therapeutic to market at over $2 billion and spanning more than a decade. READ MORE Leveraging Routinely Collected Health Data to Improve Early Identification of Autism and Co-Occurring Conditions
This project comes out of Duke’s Autism Center of Excellence (ACE). The ACE’s overall goal is to use a translational digital health and computational approach to address the critical need for more effective autism screening tools, objective outcome measures, and brain-based biomarkers that can be used in clinical trials with young autistic children. READ MORE Geospatial Analysis of Readmission Risk for Older Patients with Fractures
Older patients with fractures face a high hospital readmission risk, which can lead to additional health complications and increase costs for both patients and healthcare systems. To inform preventative care, we studied readmission risk using data from 1,202 Duke patients in Durham and surrounding areas who experienced upper extremity fractures at age 50 or above. READ MORE EHR Foundation Models for Smarter, More Reliable, and Dynamically Deployed Clinical Decision Support
Traditional frameworks for developing clinical decision support (CDS) tools from electronic health record (EHR) data often rely on extensive data preprocessing. This typically involves tedious and inefficient steps such as cleaning raw data, aggregating clinical event sequences into tabular count-based features, and deriving longitudinal measures for labs and vitals. READ MORE Early Prediction of Mental Health Risk in Adolescents
Despite decades of mental health research, accurately predicting psychiatric outcomes for individuals remains elusive. Recent advances in artificial intelligence (AI) are making this goal more attainable, and the potential public health significance of these efforts cannot be overstated. READ MORE Project Archive
PCORI HSII Capacity-Building for Duke University Health System (2023-2024)
As part of PCORI’s Health Systems Implementation Initiative (HSII), the Capacity Building projects directly prepared health systems to undertake Implementation Projects to actively advance the adoption of evidence as part of care delivery within their systems and to undertake required program evaluation as part of these projects. READ MORE
Strategies and Practical Approaches in Research and Collaboration (SPARC) Initiative for Algorithmic Bias Detection and Mitigation (2023-2024)
This project brought together Duke faculty colleagues and other experts from different disciplines to discuss harmful bias and created innovative training to eliminate bias in algorithms that are used to make decisions and target resources that affect patient care. READ MORE
A Novel Platform to Support Contact Tracing during the COVID-19 Pandemic (2020-2024)
Early in the pandemic as infections were spreading, Duke collaborators developed a cloud platform to support snowball sampling for more efficient contact tracing.
Delphi Study during the Duke Summit on AI for Health Innovation (2024)
To probe for consensus on the barriers and facilitators of innovation in Health AI, we conducted a Delphi study during the Duke Summit on AI for Health Innovation. READ MORE
