NSF Award
2406439
Medical errors are one of the leading causes of death in the United States. Healthcare professionals (HPs) perform critical tasks in complex environments, facing incomplete and inconsistent information, fluid team dynamics, and complex interactions with technology, among other distractions. Consequently, healthcare presents a high-risk setting for errors. Errors are driven by multiple interrelated factors, such as patient acuity, human factors, team composition, and clinical context. Addressing these factors in isolation fails to capture their interconnected nature. This project adopts a proactive systems approach, leveraging artificial intelligence (AI) techniques to analyze diverse data sources, including cognitive and emotional states, behavioral patterns, and environmental conditions, to predict errors. Error prediction will leverage AI to determine assistive interventions for HPs that will contribute to the avoidance or reduction of medical errors. This project aligns with the National Science Foundation's mission to support innovative research that benefits society by advancing national health. It builds on ongoing interdisciplinary research at the nexus of human performance, sensing technologies, and human-AI collaboration. Additionally, it offers opportunities for graduate and undergraduate students to gain critical research skills for advancing smart and safe healthcare delivery in the age of artificial intelligence. The outcomes of this project may also impact other high-stakes fields like aviation, military training, and emergency response. <br/><br/>This project will develop a novel system to identify cognitive and affective states, behavioral patterns, and contextual factors contributing to medical errors. The system will provide a transparent risk assessment of medical errors and prevent them through AI collaboration with HPs. The research team will implement multi-modal machine-learning algorithms leveraging data from wearable sensors, audio, video, and other contextual sources to predict medical errors. The research involves three main thrusts: (1) investigating the impact of neurophysiological processes on medical error rates, (2) developing multi-modal representation learning to predict healthcare professionals' cognitive states and potential errors, and (3) designing adaptive AI interactions considering human readiness and cognitive resources. A key component is human subject research (HSR), collecting multi-modal data (facilitating system testing and validation) from HPs across various simulated emergency scenarios with varying distractions and error-inducing conditions. The HSR will also compare how healthcare professionals perform under three conditions: (i) no AI support, (ii) constant AI support, and (iii) context-aware AI support based on predicted errors. This strategy will yield quantitative and qualitative data to assess the effectiveness of the context-aware, adaptive Human-AI teaming framework developed in this project.<br/><br/>This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
