Real-time COVID-19 hospital admissions forecasting with leading indicators and ensemble methods in England (preprint)

Hospitalisations from COVID-19 with Omicron sub-lineages have put a sustained pressure on the English healthcare system. Understanding the expected healthcare demand enables more effective and timely planning from public health. We collect syndromic surveillance sources, which include online search data, NHS 111 telephonic and online triages. Incorporating this data we explore generalised additive models, generalised linear mixed-models, penalised generalised linear models and model ensemble methods to forecast over a two-week forecast horizon at an NHS Trust level. Furthermore, we showcase how model combinations improve forecast scoring through a mean ensemble, weighted ensemble, and ensemble by regression. Validated over multiple Omicron waves, at different spatial scales, we show that leading indicators can improve performance of forecasting models, particularly at epidemic changepoints. Using a variety of scoring rules, we show that ensemble approaches outperformed all individual models, providing higher performance at a 21-day window than the corresponding individual models at 14-days. We introduce a modelling structure used by public health officials in England in 2022 to inform NHS healthcare strategy and policy decision making. This paper explores the significance of ensemble methods to improve forecasting performance and how novel syndromic surveillance can be practically applied in epidemic forecasting.

Quantifying the risk of workplace COVID-19 clusters in terms of commuter, workplace, and population characteristics (preprint)

Objectives: To identify and quantify risk factors that contribute to clusters of COVID-19 in the workplace. Methods: We identified clusters of COVID-19 cases in the workplace and investigated the characteristics of the individuals, the workplaces, the areas they work, and the methods of commute to work, through data linkages based on Middle Layer Super Output Areas (MSOAs) in England between 20/06/2021 and 20/02/2022. We estimated associations between potential risk factors and workplace clusters, adjusting for plausible confounders identified using a Directed Acyclic Graph (DAG). Results: For most industries, increased physical proximity in the workplace was associated with increased risk of COVID-19 clusters, while increased vaccination was associated with reduced risk. Commuter demographic risk factors varied across industry, but for the majority of industries, a higher proportion of black/african/caribbean ethnicities, and living in deprived areas, was associated with increased cluster risk. A higher proportion of commuters in the 60-64 age group was associated with reduced cluster risk. There were significant associations between gender, work commute methods, and staff contract type with cluster risk, but these were highly variable across industries. Conclusions: This study has used novel national data linkages to identify potential risk factors of workplace COVID-19 clusters, including possible protective effects of vaccination and increased physical distance at work. The same methodological approach can be applied to wider occupational and environmental health research.

Understanding the leading indicators of hospital admissions from COVID-19 across successive waves in the UK (preprint)

Following the UK Government's Living with COVID-19 Strategy and the end of universal testing, hospital admissions are an increasingly important measure of COVID-19 pandemic pressure. Understanding leading indicators of admissions at National Health Service (NHS) Trust, regional and national geographies help health services plan capacity needs and prepare for ongoing pressures. We explored the spatio-temporal relationships of leading indicators of hospital pressure across successive waves of SARS-CoV-2 incidence in England. This includes an analysis of internet search volume values from Google Trends, NHS triage calls and online queries, the NHS COVID-19 App, lateral flow devices and the ZOE App. Data sources were analysed for their feasibility as leading indicators using linear and non-linear methods; granger causality, cross correlations and dynamic time warping at fine spatial scales. Consistent temporal and spatial relationships were found for some of the leading indicators assessed across resurgent waves of COVID-19. Google Trends and NHS queries consistently led admissions in over 70% of Trusts, with lead times ranging from 5-20 days, whereas an inconsistent relationship was found for the ZOE app, NHS COVID-19 App, and rapid testing, that diminished with granularity, showing limited autocorrelation of leads between -7 to 7 days. This work shows that novel syndromic surveillance data has utility for understanding the expected hospital burden at fine spatial scales. The analysis shows at low level geographies that some surveillance sources can predict hospital admissions, though care must be taken in relying on the lead times and consistency between waves.

Forecasting influenza hospital admissions within English sub-regions using hierarchical generalised additive models (preprint)

Background: Seasonal influenza causes a substantial burden on healthcare services over the winter period when these systems are already under pressure. Policies during the COVID-19 pandemic supressed the transmission of season influenza, making the timing and magnitude of a potential resurgence difficult to predict. Methods: We developed a hierarchical generalised additive model (GAM) for the short-term forecasting of hospital admissions with a positive test for the influenza virus sub-regionally across England. The model incorporates a multi-level structure of spatio-temporal splines, weekly seasonality, and spatial correlation. Using multiple performance metrics including interval score, coverage, bias, and median absolute error, the predictive performance is evaluated for the 2022/23 seasonal wave. Performance is measured against an autoregressive integrated moving average (ARIMA) time series model. Results: The GAM method outperformed the ARIMA model across scoring rules at both high and low-level geographies, and across the different phases of the epidemic wave including the turning point. The performance of the GAM with a 14-day forecast horizon was comparable in error to the ARIMA at 7 days. The performance of the GAM is found to be most sensitive to the flexibility of the smoothing function that measures the national epidemic trend. Interpretation: This study introduces a novel approach to short-term forecasting of hospital admissions with influenza using hierarchical, spatial, and temporal components. The model is data-driven and practical to deploy using information realistically available at time of prediction, addressing key limitations of epidemic forecasting approaches. This model was used across the winter for healthcare operational planning by the UK Health Security Agency and the National Health Service in England.

Nowcasting the 2022 mpox outbreak in England (preprint)

In May 2022, a cluster of mpox cases were detected in the UK that could not be traced to recent travel history from an endemic region. Over the coming months, the outbreak grew, with over 3000 total cases reported in the UK, and similar outbreaks occurring worldwide. These outbreaks appeared linked to sexual contact networks between gay, bisexual and other men who have sex with men. Following the COVID-19 pandemic, local health systems were strained, and therefore effective surveillance for mpox was essential for managing public health policy. However, the mpox outbreak in the UK was characterised by substantial delays in the reporting of the symptom onset date and specimen collection date for confirmed positive cases. These delays led to substantial backfilling in the epidemic curve, making it challenging to interpret the epidemic trajectory in real-time. Many nowcasting models exist to tackle this challenge in epidemiological data, but these lacked sufficient flexibility. We have developed a novel nowcasting model using generalised additive models to correct the mpox epidemic curve in England, and provide real-time characteristics of the state of the epidemic, including the real-time growth rate. This model benefited from close collaboration with individuals involved in collecting and processing the data, enabling temporal changes in the reporting structure to be built into the model, which improved the robustness of the nowcasts generated.

Novel methods for estimating the instantaneous and overall COVID-19 case fatality risk among care home residents in England (preprint)

The COVID-19 pandemic has had high mortality rates in the elderly and frail worldwide, particularly in care homes. This is driven by the difficulty of isolating care homes from the wider community, the large population sizes within care facilities (relative to typical households), and the age/frailty of the residents. To quantify the mortality risk posed by disease, the case fatality risk (CFR) is an important tool. This quantifies the proportion of cases that result in death. Throughout the pandemic, CFR amongst care home residents in England has been monitored closely. To estimate CFR, we apply both novel and existing methods to data on deaths in care homes, collected by Public Health England and the Care Quality Commission. We compare these different methods, evaluating their relative strengths and weaknesses. Using these methods, we estimate temporal trends in the instantaneous CFR (at both daily and weekly resolutions) and the overall CFR across the whole of England, and dis-aggregated at regional level. We also investigate how the CFR varies based on age and on the type of care required, dis-aggregating by whether care homes include nursing staff and by age of residents. This work has contributed to the summary of measures used for monitoring the UK epidemic.

EpiBeds: Data informed modelling of the COVID-19 hospital burden in England (preprint)

The first year of the COVID-19 pandemic put considerable strain on the national healthcare system in England. In order to predict the effect of the local epidemic on hospital capacity in England, we used a variety of data streams to inform the construction and parameterisation of a hospital progression model, which was coupled to a model of the generalised epidemic. We named this model EpiBeds. Data from a partially complete patient-pathway line-list was used to provide initial estimates of the mean duration that individuals spend in the different hospital compartments. We then fitted EpiBeds using complete data on hospital occupancy and hospital deaths, enabling estimation of the proportion of individuals that follow different clinical pathways, and the reproduction number of the generalised epidemic. The construction of EpiBeds makes it straightforward to adapt to different patient pathways and settings beyond England. As part of the UK response to the pandemic, EpiBeds has provided weekly forecasts to the NHS for hospital bed occupancy and admissions in England, Wales, Scotland, and Northern Ireland.

Using statistics and mathematical modelling to understand infectious disease outbreaks: COVID-19 as an example (preprint)

During an infectious disease outbreak, biases in the data and complexities of the underlying dynamics pose significant challenges in mathematically modelling the outbreak and designing policy. Motivated by the ongoing response to COVID-19, we provide a toolkit of statistical and mathematical models beyond the simple SIR-type differential equation models for analysing the early stages of an outbreak and assessing interventions. In particular, we focus on parameter estimation in the presence of known biases in the data, and the effect of non-pharmaceutical interventions in enclosed subpopulations, such as households and care homes. We illustrate these methods by applying them to the COVID-19 pandemic.

Challenges in control of Covid-19: short doubling time and long delay to effect of interventions (preprint)

The unconstrained growth rate of COVID-19 is crucial for measuring the impact of interventions, assessing worst-case scenarios, and calibrating mathematical models for policy planning. However, robust estimates are limited, with scientific focus on the time-insensitive basic reproduction number R0. Using multiple countries, data streams and methods, we consistently estimate that European COVID-19 cases doubled every three days when unconstrained, with the impact of physical distancing interventions typically seen about nine days after implementation, during which time cases grew eight-fold. The combination of fast growth and long detection delays explains the struggle in countries' response better than large values of R0 alone, and warns against relaxing physical distancing measures too quickly. Testing and tracing are fundamental in shortening such delays, thus preventing cases from escalating unnoticed.

Challenges in control of Covid-19: short doubling time and long delay to effect of interventions (preprint)

Early assessments of the spreading rate of COVID-19 were subject to significant uncertainty, as expected with limited data and difficulties in case ascertainment, but more reliable inferences can now be made. Here, we estimate from European data that COVID-19 cases are expected to double initially every three days, until social distancing interventions slow this growth, and that the impact of such measures is typically only seen nine days - i.e. three doubling times - after their implementation. We argue that such temporal patterns are more critical than precise estimates of the basic reproduction number for initiating interventions. This observation has particular implications for the low- and middle-income countries currently in the early stages of their local epidemics.