Developing Artificial Intelligence Framework for Acuity Assessment in Psychiatry Units

Project Overview

The high demand for transfers from Emergency Department to psychiatric inpatient units, and the limited supply of available beds imposes serious challenges to the management and care of patients with acute psychiatric conditions. Hospital admission decisions are based on a variety of factors, most importantly patient acuity and receiving unit acuity. Although acuity rating scales are prevalent in critical care, there are no validated acuity rating tools in inpatient psychiatry, possibly due to a paucity of objective diagnostic and biological predictive measures. Measuring unit acuity is complicated by unpredictable threats to safety such as violent behavior, fluctuating acuity during shifts, shared living environments, and inadequate flexibility in patient-nurse staffing ratios.

The purpose of this research project is to initiate data collection efforts at two Medstar Health inpatient psychiatry units, and to develop an innovative data-driven Artificial Intelligence/Machine Learning (AI/ML) model for estimating psychiatric patient and unit acuity. The acuity framework will consist of patient-level and unit-level models that will eventually be combined to create a dynamic model to guide the admission decision and nursing staffing needs. The research team will train the data from shift acuity, admission request, transfer request logs and patient characteristics to score patient and unit acuity and to establish a link between these scores and admit/denial decisions. Notes from the patient records will also be mined using natural language processing and OpenAI to gain further insights about the patients’ conditions at the transfer request time. The long-term goal is to implement this tool into the Medstar electronic medical record as a decision support tool to aid clinicians and administrators in making optimal admission decisions for safe and high-quality psychiatric care. This project will also give students an opportunity to understand the clinical workflow and get better equipped to translate their knowledge into AI/ML modeling.

Team