Analyzing the Variations in Emergency Department Boarding and Testing the Transferability of Forecasting Models across COVID-19 Pandemic Waves in Hong Kong: Hybrid CNN-LSTM approach to quantifying building-level socioecological risk (preprint)

Emergency department's (ED) boarding (defined as ED waiting time greater than four hours) has been linked to poor patient outcomes and health system performance. Yet, effective forecasting models is rare before COVID-19, lacking during the peri-COVID era. Here, a hybrid convolutional neural network (CNN)-Long short-term memory (LSTM) model was applied to public-domain data sourced from Hong Kong's Hospital Authority, Department of Health, and Housing Authority. In addition, we sought to identify the phase of the COVID-19 pandemic that most significantly perturbed our complex adaptive healthcare system, thereby revealing a stable pattern of interconnectedness among its components, using deep transfer learning methodology. Our result shows that 1) the greatest proportion of days with ED boarding was found between waves four and five; 2) the best-performing model for forecasting ED boarding was observed between waves four and five, which was based on features representing time-invariant residential buildings' built environment and sociodemographic profiles and the historical time series of ED boarding and case counts, compared to during the waves when best-performing forecasting is based on time-series features alone; and 3) when the model built from the period between waves four and five was applied to data from other waves via deep transfer learning, the transferred model enhanced the performance of indigenous models.

Quantifying the Risk of General Health and Early COVID-19 Spread in Residential Buildings with Deep Learning and Expert-augmented Machine Learning (preprint)

Buildings' built environment has been linked to their occupants' health. It remains unclear whether those elements that predisposed its residents to poor general health before the two SARS pandemics also put residents at risk of contracting COVID-19 during early outbreaks. Relevant research to uncover the associations is essential, but there lacks a systematic examination of the relative contributions of different elements in one's built environment and other non-environmental factors, singly or jointly. Hence, the current study developed a deep-learning approach with multiple input channels to capture the hierarchical relationships among an individual's socioecology's demographical, medical, behavioral, psychosocial, and built-environment levels. Our findings supported that 1) deep-learning models whose inputs were structured according to the hierarchy of one's socioecology outperformed plain models with one-layered input in predicting one's general health outcomes, with the model whose hierarchically structured input layers included one's built environment performed best; 2) built-environment features were more important to general health compared to features of one's sociodemographic and their health-related quality of life, behaviors, and service utilization; 3) a composite score representing built-environment features' statistical importance to general health significantly predicted building-level COVID-19 case counts; and 4) building configurations derived from the expert-augmented learning of granular built-environment features that were of high importance to the general health were also linked to building-level COVID-19 case counts of external samples. Specific built environments put residents at risk for poor general health and COVID-19 infections. Our machine-learning approach can benefit future quantitative research on sick buildings, health surveillance, and housing design.