Abstract

Healthcare-associated infections (HAIs) represent a significant burden to healthcare provision; in the United States alone, it is estimated that approximately 2 million patients acquire HAIs each year. As part of a larger effort to understand how HAIs spread, we deployed a wireless sensor network in the Medical Intensive Care Unit of the University of Iowa Hospitals and Clinics. We used data reported by the network to estimate healthcare worker movement, interactions between healthcare workers, and adherence to hand sanitization policies. Our experiment joins the growing yet still small collection of sensor network deployments in healthcare settings. This work contributes to this body of research by presenting a comprehensive approach to pre-processing the collected sensor data, thereby reducing errors and increasing robustness. We provide two main contributions: (i) a simple and theoretically sound calibration method for sensor signals that eliminates biases in pairwise sensor communication and (ii) filters that increase the reliability of signal strength from stationary sensors. We validate our methods by comparing visits of healthcare workers to rooms, as discovered from the sensor data, to ground truth room occupancy data collected in notes.