Data Science in Clinical Care (2025-2026)
Background
Clinical decision support (CDS) tools are data science innovations that aim to provide timely information to help providers make informed decisions and provide the right treatment to the right patient at the right time. Typically, CDS tools consist of computerized algorithms that draw from electronic health record data.
CDS tools that achieve good metrics when tested in the lab do not necessarily perform as well when implemented in real world settings. This disconnect occurs because target users may not use the tools as expected. While some reasons have to do with the tools themselves (e.g., providers not trusting the scores), others have to do with systemic factors (e.g., annual staff turnover).
Revealing these structural factors is important because the system within which the tools function has a bearing on their quality. Implementation of complex CDS tools without understanding the context within which they operate could end up perpetuating structural biases, such as inequities in access to health services. Understanding system structure provides insight into how to intervene to achieve consistent and equitable use and better outcomes
Project Description
Building on the work of a previous team, this project team will engage with Duke Health staff to illuminate system-level factors that affect their clinical decision support tool adoption. Members will employ system dynamics modeling, which shows the connections between different factors in a complex system and how these relationships change over time. The overarching goal of this project is to create a foundational model of CDS adoption that draws from different use cases.
This team will focus on using CDS risk scores to identify patients with peripheral arterial disease (PAD) who are at high risk for wound infection after surgery, addressing a gap in consistent real-time intervention guidance. The team will work on group-based modeling workshops with staff, in which facilitators guide the participants to share stories about their views about factors and their behavior over time.
The team will use a participatory approach to create system dynamics models by collecting data from approximately 30 staff at Duke Health who work with clinical decision support tools around cases of lower extremity revascularization. These staff will include clinical decision support developers and owners and clinical providers (nurses, medical assistants, doctors).
Members will also guide participants on how to make connections between these factors to develop a causal loop diagram that reveals how the factors affect each other in feedback loops, revealing a theory about an underlying causal structure that explains why there is inconsistency in risk assessment for peripheral arterial disease (PAD) surgical patients, leveraging points for intervention and how a CDS tool can help improve the process outcomes. This team will work on developing PAD risk analysis alongside a computerized version of the model that can serve as a simulation model for CDS implementation.
Anticipated Outputs
Simulation model; publication; development of a clinical decision support tool for PAD risk assessment
Student Opportunities
Ideally, this project team will include 1- 3 graduate students and 2-4 undergraduate students with backgrounds and interests in computer science, engineering, the intersection of data science, policy, healthcare and implementation.
Students will learn how to execute participatory system dynamics modeling techniques (including variable elicitation, graphs over time and causal loop diagramming scripts) through in-person workshops with clinical staff. Members will contribute to ongoing development of a generalizable online simulation model of factors that affect implementation of clinical decision support tools by incorporating findings from this team’s use case.
Team members will prepare reports, papers and presentations to demonstrate this innovative approach to studying clinical decision support implementation and communicate findings and implications to operational, academic and policy audiences. This team will also attend a week-long training institute at Systems Science for Social Impact in St. Louis prior to the start of the fall semester.
Team members will meet weekly to develop project goals and deliverables and learn new research skills in participatory system dynamics modeling. Returning students will have the chance to take the lead in training new team members.
In Fall 2025, the team will meet on Tuesdays from 4-6 p.m.
A graduate student will be selected as project manager.
Timing
Summer 2025- Spring 2026
- Summer 2025: Complete training completion at the Systems Science for Social Impact Institute
- Fall 2025: Submit IRB for approval; schedule, plan and conduct group model building workshops; refine casual loop diagrams
- Spring 2026: Analyze fall semester data; write a case study; program a system dynamics model based on casual loop diagram; identify measures for programmed version of the system dynamics model
Crediting
Academic credit available for fall and spring semesters
See earlier related team, Data Science in Clinical Care (2024-2025).