CLARA - Global Engineering Programs and Partnerships - Purdue University

AI-Powered Clinical Skill Assessment to Elevate the Nursing Education

THE CHALLENGE

The U.S. nursing shortage is not only a workforce problem but a training capacity crisis. In 2025, nursing schools turned away 93,176 qualified applications, up from 80,162 in 2024 [1,2], largely due to limited faculty and clinical training resources [1]. One time-intensive requirement is the clinical skills “checkoff,” which requires individual competency validation. For a cohort of 40 students, validating a single eight-minute procedure takes over five hours of faculty observation (40 × 8 min ≈ 5.3 hours), excluding setup, instructor debriefing, and retakes.

Time is only part of the problem. Clinical competence requires recognizing subtle errors, such as sterile-field breaches or incorrect landmarking, and evaluating them in context. These assessments demand expert judgment and individualized feedback, making it difficult for educators to maintain consistent evaluation quality as student enrollment grows. Without scalable assessment support, institutions face a difficult tradeoff between training capacity and clinical competency assurance. This raises a critical question: How can we scale skills validation while preserving feedback quality and patient safety?

OUR AI SOLUTION

To address the need, we have developed CLARA (Clinical Learning and Assessment Reasoning Agent), an MLLM-powered clinical skills assessment system. Built on authentic student skill validation videos and expert-grounded annotations, CLARA is designed to learn from real-world errors and subtle “grey-area” deviations. By combining procedural video understanding with multidimensional assessment of correctness, performance quality, and contextual factors, CLARA moves beyond basic step recognition toward fine-grained nursing skills evaluation. These assessments are further grounded in instructor-defined checkoff rubrics and expert reviews to provide explainable evaluations and personalized, actionable feedback (Figs. 1–2). In practice, CLARA supports the clinical learning process through three interconnected functions: assessing performance, guiding skill improvement, and tracking learning progress over time.

Together, these functions create a continuous learning cycle, empowering nursing students to practice independently, learn from mistakes, and progressively strengthen their clinical skills.

PROGRESS SINCE PHASE 1

WHY IT MATTERS

PEOPLE

Students gain specific coaching on incomplete or unsafe actions. Preceptors focus attention on uncertain or high-risk steps, while retaining all formal evaluation decisions. In the long term, structured refresher training could support practicing nurses in maintaining their clinical competencies.

EXPECTED IMPACTS

CLARA aims to reduce faculty review time by ≥70% while achieving ≥95% agreement on safety-critical error detection, enabling more efficient and consistent clinical skills assessment. Supported by preliminary results of 92.4% step-recognition accuracy and 85.1% agreement with instructor-annotated assessments, CLARA has the potential to expand training capacity and personalized feedback without increasing faculty workload.

PHASE 2 GOALS 

The requested $50k in ShahLab.AI Phase 2 funding will support platform engineering, student-facing web development, community pilot testing, expert-led clinical validation, and deployment preparation over 12 months. 

Months 1–3 | Platform Development: 

- Convert the existing annotation tool and AI models into an automated video analysis pipeline. 
- Develop a student-facing web portal for video uploads, step-level assessment reports, and actionable AI feedback. 

Months 3–6 | Community Pilot Testing: 

- Deploy a supervised pilot with a target of 20–30 nursing students and 10–12 instructors/preceptors at Purdue Nursing and NCNC. 
- Gather user perspectives to evaluate platform usability, clinical relevance, and learning experience. 

Months 6–9 | Clinical Validation & AI Refinement: 

- Benchmark CLARA against expert assessments, measuring accuracy, safety-critical error detection, clinical guidance quality, and faculty time savings. 
- Incorporate expert-validated corrections into the clinical RAG knowledge base through human-in-the-loop refinement. 

Months 9–12 | Strategic Partnerships & Sustainability: 

- Strengthen CLARA's deployment readiness through platform improvements, data privacy safeguards, instructor oversight protocols, and faculty onboarding resources. 
- Leverage Purdue Nursing Lab's existing relationship with ATI to explore integration into its established educational ecosystem. 
- Engage with Laerdal to assess SimCapture compatibility and explore pilot collaborations and licensing opportunities to support broader adoption beyond ShahLab.AI funding. 

PATH TO SCALE

Building on Phase 1's research foundation and Phase 2's planned clinical validation, CLARA will progress from a research prototype toward institutional deployment and broader adoption in nursing education. Its web-based architecture will enable nursing programs to integrate AI-assisted assessment into existing training workflows, expanding access to personalized learning without proportionally increasing faculty workload. Our adoption strategy follows three stages: skill-based nursing assessment, simulation-based learning, and clinical competency development during clinical rotations. Beginning with Purdue's Nursing Skills Lab, CLARA will expand into nursing simulation centers and eventually support clinical training environments, such as NCNC.

For long-term sustainability, we envision two complementary revenue models: institutional subscriptions for nursing schools and clinical training centers, and platform licensing through established educational technology providers. Potential collaborations with ATI and Laerdal could accelerate regional and national adoption through existing institutional networks. Looking ahead, our vision is to support nursing education across these three learning environments, enabling every nurse to train with insight, grow with confidence, and deliver better care.

WORKING WITH PARTNERS 

Purdue Nursing | Data & Lab Validation
Connection: Libby Richards, School Head; Dagne Hartman

Continue nursing video collection, expert annotation, and rubric development. Faculty and students will lead initial CLARA testing, sharing insights to refine clinical assessments.

NCNC | Clinical Validation
Connection: Jennifer Coddington, Medical Director

Clinical preceptors will evaluate CLARA in practice-based training, validate assessment quality, and identify safety concerns and improvements for clinical workflows.

THE TEAM

• Shen Chang
• Dr. Nan Kong - Purdue University, Biomedical Engineering
• Clinical partners: Purdue School of Nursing
• NCNC