Forecasting indoor air quality in high-risk operating theatres in hospitals using Artificial Neural Networks

Airborne particulate matter is a vehicle for transmitting infectious diseases. Therefore, operating theatres must undergo periodic audits to verify that their indoor air quality remains within acceptable ranges. However, the months that may pass between on-site data collections mean that the temporal discretisation of the audit sample is not representative of the room's indoor air quality. This research aimed to forecast the particulate matter in high-risk operating theatres as a novel approach to continuously monitor indoor air quality in hospitals employing artificial neural networks. For that purpose, three types of artificial neural network architectures were analysed: Multilayer Perceptron, Long Short-Term Memory, and Nonlinear Autoregressive Exogenous. Temperature, relative humidity, airflow supply and extraction, and the renewal rate in high-risk operating theatres were used as input variables for each model. Airborne particulate matter data were used as the predicted parameter. The networks were trained and validated using a curated dataset obtained from 20 hospitals in Spain over five years. The Nonlinear Autoregressive Exogenous model exhibited better performance metrics in forecasting both 0.5 and 5 μm particles. This research demonstrates that real-time HVAC variables can accurately forecast the presence of airborne particles in operating theatres, thereby reducing the risk of nosocomial infection. Findings are extremely useful for hospital managers, easily applicable in the operation and maintenance of critical facilities. • Airborne particulate matter is a vehicle for transmitting nosocomial diseases. • Hygienic inspections of operating theatres suffer from overly discrete time intervals. • Artificial Neural Network predictive approach supports surgical site management. • Nonlinear Autoregressive Exogenous model shows excellent forecasting capability. • Validation with actual data supports the practical application of these findings.