Healthcare is the most expensive operations in the United States costing the providers and government. According to National Healthcare Expenditures in 2016, United States spends $3.3 trillion (which amounts to $10,438 per person). 32% of this amount (i.e. $1.1 trillion) is spent on hospital operations. Many operational processes in the hospital are manual, reactive and inefficient. Every healthcare provider is looking at how to drive quality for the patient and reduce cost. This project seeks to leverage IOT, AI and visual analytics on data streams to build a pilot of a Digital-twin for one of the hospital to automate and improve operational processes in the hospital. We specifically seek to understand the factors and activities that contribute to stress for nurses as part of this study. We plan to create a continuous data stream processing, analytics and visualization at the edge to satisfy the lower latency, privacy, and cost requirements. The proposed project builds on prior research in the area of distributed stream processing at the edge for smart buildings, deep learning approaches to improve time series forecasting, and ED volume forecasting.
Specific objectives of the project include the following: (1) Analyze stress index level of providers through biometric feedback from wearables, (2) Real-time activity detection of nurses from wearables, and (3) visualization of activities & stress data in a 3D environment. Accomplishing these objectives require many data processing and machine learning tasks including experimenting with different wearable devices, using machine learning to detect stress and predict activity from different sensors, high-resolution indoor positioning. In addition, we will conduct a formal study to measure the reliability of automated measurements.