Data Science in Clinical Care (2026-2027)

Background

Clinical decision support tools draw on electronic health records and other data sources to provide timely information that can help clinicians make better decisions about patient care. In theory, these tools can improve outcomes, standardize care and reduce implicit bias. In practice, however, many decision support tools are not widely adopted because they do not align well with clinical workflows or reflect the complexity of real-world healthcare systems.

At Duke Health, for example, predictive tools designed to identify patients at risk of rapid deterioration or hospital readmission have faced resistance from clinicians because they did not fit easily into daily practice. To address these challenges, Duke Health and the School of Medicine established the Algorithm-Based Clinical Decision Support Oversight Committee, which provides governance and evaluation of clinical decision support tools to ensure patient safety and high-quality care. This project focuses on understanding the barriers and opportunities that shape whether data-driven tools are successfully implemented in clinical settings.

Project Description

This project team will examine how data science tools can be more effectively designed and implemented in clinical care by focusing on the connection between development and real-world use. The team will work with clinicians and patients to better understand the complex systems of care that influence outcomes after surgery, particularly for patients undergoing open surgery for peripheral artery disease.

Using group-based modeling, a collaborative method in which participants build visual models together, team members will help develop causal loop diagrams that show how different clinical, organizational and contextual factors interact over time. Initial models will focus on surgical site infections and will also explore prolonged length of stay, hospital readmission, graft thrombosis and post-operative bleeding.

These system-level models will inform the development of a framework to guide the design and implementation of clinical decision support tools across different clinical contexts. As part of this work, the team will contribute to the creation of an online simulator that models barriers and levers in clinical environments, helping bridge the gap between data science innovation and clinical integration.

Anticipated Outputs

Facilitation manual for group model building sessions
Poster presentation at the Bass Connections Showcase
Manuscripts on system dynamics models of post-operative outcomes in vascular surgery
Simulation models for clinical decision support implementation

Student Opportunities

Ideally, this project team will include 3 graduate or professional students and 6 undergraduate students. Undergraduate students may come from computer science, pre-med, science or engineering backgrounds. Graduate and professional students may have training in statistics, computer science, population health or medicine, including medical students and resident physicians.

Team members will learn how to design and carry out applied research that informs the development and implementation of machine learning-enabled clinical decision support tools. Students will gain experience facilitating group model building sessions, synthesizing causal loop diagrams, working with clinical stakeholders and understanding system dynamics concepts. Additional skills include defining research questions, conducting interdisciplinary collaboration, writing case studies and communicating findings through posters and publications.

The team will meet weekly as part of the Vascular Informatics Lab, with additional work occurring outside scheduled meetings. A student project manager will be recruited to help coordinate activities and support team organization. This project involves engagement with clinicians and patients and will operate under appropriate institutional review board approvals. No travel beyond the Triangle is expected.

Timing

Fall 2026 – Spring 2027

Fall 2026:

Training in facilitation and group model building methods
Onboarding to institutional review board protocols
Introduction to clinical systems and research context

Spring 2027:

Facilitation of group model building sessions
Development and synthesis of system dynamics models
Preparation of presentations and written outputs

Crediting

Academic credit available for fall and spring semesters

See earlier related team, Data Science in Clinical Care (2025-2026).