Multisensory interaction and predictive modelling of indoor environmental comfort for pregnant women in obstetrics departments in Chinese hospitals

Abstract

Indoor environmental comfort in healthcare buildings has traditionally been evaluated using general population models that insufficiently account for the physiological and psychological particularities of vulnerable groups. Pregnancy-related physiological adaptation and heightened emotional sensitivity can modify comfort-related physiological and perceptual responses, yet these responses remain underexplored in hospital environments. This research addresses the lack of an integrated framework for understanding and predicting indoor environmental comfort for pregnant women in outpatient healthcare settings.

This study establishes a multidimensional comfort framework integrating physical environmental parameters, physiological responses, psychological states, and adaptive behaviours. Field investigations were conducted in hospital outpatient departments, combining objective environmental measurements, physiological monitoring, and structured questionnaires. Statistical analyses were first employed to examine reliability, distributional patterns, and group differences across pregnancy stages and accompanying individuals. Subsequently, structural equation modelling (PLS-SEM) was applied to identify path relationships, mediation and moderation mechanisms, and structural variations across early, mid, and late pregnancy. Findings indicate that thermal factors exert the strongest direct and indirect influence on overall environmental comfort, with psychophysiological variables, particularly emotional state and pulse rate, serving as significant mediators. Gestational-stage-dependent structural differences were observed, indicating progressive strengthening of psychophysiological mediation in later pregnancy. Comparative analysis further revealed perceptual and structural divergences between pregnant women and accompanying individuals.

Building upon the identified dominant role of thermal factors, a stage-sensitive thermal comfort model was developed for pregnant women in hospitals in the cold-climate region of China. Neutral and acceptable temperature ranges were derived, metabolic rates were recalibrated based on individual basal metabolic rate (BMR) estimation, and adjusted PMV and adaptive models were validated. The adjusted PMV model incorporating BMR-based metabolic estimation demonstrated improved predictive alignment with observed thermal sensation, particularly in late pregnancy, achieving a 6.6% improvement in prediction accuracy compared with the standard PMV model.

To advance personalised comfort prediction, multiple machine learning algorithms were developed and benchmarked. Among the evaluated models, a DL-MLP ensemble combined with SMOTE achieved the best balanced predictive performance. Model interpretability analysis using SHAP revealed nonlinear interactions among environmental, physiological, and behavioural variables, as well as seasonal and gestational dependencies. The findings further identified an integrated multisensory comfort envelope associated with comparatively higher overall comfort, comprising an operative temperature of approximately 17.7–24°C, illuminance above 200 lx, correlated colour temperature below 5500 K, CO₂ concentration below 1000 ppm, noise level below 70 dB, air velocity below 0.2 m/s, and relative humidity below 65%. The integration of mechanism-based modelling and data-driven prediction enabled the construction of a personalised comfort prediction framework. Finally, building energy simulation was used to compare comfort-informed temperature setpoint scenarios, indicating potential trade-offs between thermal comfort improvement and energy efficiency in the studied hospital context.

Overall, this thesis contributes a gestational-stage-sensitive, psychophysiology-informed, and data-driven framework for evaluating and predicting indoor environmental comfort for pregnant women in healthcare settings. The findings provide theoretical advancement in multisensory comfort modelling and practical guidance for evidence-based hospital environmental design and operation.

Date of Award	18 Jul 2026
Original language	English
Awarding Institution	University of Nottingham
Supervisor	Jun Lu (Supervisor), Wu Deng (Supervisor), Zhiang Zhang (Supervisor) & Paolo Beccarelli (Supervisor)