A Sensor Network for Real-Time Measurement of Aerosol Movement and Persistence for Infectious Disease Monitoring in ICUs and ORs

Abstract

Since the COVID-19 pandemic, air quality has become a public concern due to infectious diseases spread by aerosols. A methodology is presented to map the movement and size of particles using a network of low-cost particulate matter (PM) sensors to determine aerosol persistence in operating rooms (ORs) and intensive care units (ICUs). Mimicking aerosol generation by a patient, we nebulized saline solution to create NaCl particles that function as tracers for potentially infectious aerosols. The ORs rapidly returned to the background after the aerosolization, with a minimum of 94 seconds and a maximum of 254 seconds due to the high air exchange rate. The ICUs had more variability. Positive pressure (door closed) and neutral pressure (door open) ICUs took the longest to clear with a maximum of 830 seconds. Negative-pressure rooms returned to background level to background levels of aerosols on average 1.5-2 times faster than positive-pressure and neutral-pressure rooms (~500 sec). In positive and neutral-pressure rooms, the aerosol plume exfiltrates from the room ~1-2 minutes after the start of aerosolization, with up to 7% of total generated aerosols escaping the ICUs in positive pressure rooms and up to 17% -- in neutral pressure rooms. The outside sensors in negative-pressure ICUs remained at background levels in each experiment. After the initial spike at the aerosolization source (head of the bed), the rooms become 'well-mixed' after 100sec, i.e., the aerosol concentration becomes uniform. The PM level reduces the exponential decay rate. Negative pressure ICUs consistently had the fastest decay rate, with an average of 1.5-2 x faster than positive or neutral pressure rooms. Statistical K-means analysis of aerosol spatio-temporal distribution was performed; the clustering algorithm suggested that ICU could be divided into three distinct zones of similar aerosol levels. This gave similar results to zoning the room spatially by sorting the room into (i) the zone near the aerosolization source (the head of the bed), (ii) the periphery of the room, and (ii) the antechamber and directly outside the room. The K-means clustering did not work for ORs since some sensors never saw a rise in aerosol, and the room never became well-mixed. Instead, the OR zones were chosen based on the proximity to the aerosol source. In future studies, long-term monitoring can be attempted, and the current study can be used to inform sensor location and grid density in the network. For example, in ICU, a single sensor instead of using a grid of 16-20 sensors to show aerosol decay could be beneficial to simplify and expedite the experimental process. Experiments in ORs with a prolonged aerosolization period (5 minutes) could show more detailed airflow patterns than those done here (only 1 minute). This research was limited by a small data set (5 ORs and 4 ICUs); additional work monitoring more ICUs and ORs would help validate this work and generalize the monitoring approach. This network monitoring approach can also be extended to other medical settings, such as emergency departments, ambulatory clinics, etc.