AI Framework for Acuity Assessment in Psychiatry Units

Objectives

This interdisciplinary project led by Qiwei Britt He from the Data Science and Analytics program, Karan Kverno from the School of Nursing, and Mihriye Mete from the Psychiatry Department and Medstar Research Institute aimed to address challenges in transferring patients from Emergency Departments (ED) to psychiatry inpatient units. 

Due to the high demand for transfers and limited bed availability, decision-making in patient admissions relies on various factors, including patient and unit acuity. However, inpatient psychiatry lacks validated acuity rating tools, partly due to the absence of objective diagnostic measures and the complex nature of shared living environments, fluctuating patient behaviors, and rigid staffing ratios. 

To address these challenges, the project piloted a multi-phase initiative in 2023-2024 to develop an AI/ML-based model capable of estimating patient and unit acuity. By incorporating survey responses and patient severity descriptions, the model aimed to optimize decisions on patient admissions. The long-term goal is to implement this tool in the MedStar electronic medical record system, assisting clinicians and administrators in making informed decisions for improved patient care and safety. 

Outcomes

The project achieved significant advancements in research capabilities and student skill development. The team conducted complex research and performed advanced linguistic analyses that would have been difficult without AI tools. Students were thus able to develop a robust understanding of complex concepts like LLMs, text mining, NLP, item response models, and Bayesian networks, which is crucial for further exploration in the intersection of data science, nursing, and psychiatry. 

The research team engaged with students on cutting-edge research, guiding them through contemporary studies in language and AI. On the quantitative side, students learned how to generate new sample data to enhance data resources for contrastive analysis and gained hands-on experience in inputting raw data into AI tools for automatic text extraction. Qualitatively, students developed a richer appreciation for the challenges and opportunities present in current research. They acquired a more nuanced perspective of AI’s ethical and cultural implications in linguistics, expanding their comprehension of the broader impact of their work.

One of the key outcomes of the project was the successful collaboration between Data Science and Nursing students, which highlighted the power of interdisciplinary teamwork. Data Science students brought technical expertise to the project, handling data transfer, applying AI models, and conducting quantitative analysis, while Nursing students provided essential validation and contextual interpretation of the AI-processed data. The partnership bridged the gap between technological capabilities and real-world clinical knowledge, ensuring that the data was both accurate and applicable to patient care. 

Overall, the project underscored that AI can be an effective educational tool that enhances analytical skills and supports interdisciplinary learning. The experience provided students with practical skills, ethical awareness, and a nuanced understanding of AI’s impact on research and practice, paving the way for more engaged and informed research contributions.

Team