MFL_COVID19: Quantifying Country-based Factors affecting Case Fatality Rate in Early Phase of COVID-19 Epidemic via Regularised Multi-task Feature Learning (preprint)

Recent outbreak of COVID-19 has led a rapid global spread around the world. Many countries have implemented timely intensive suppression to minimize the infections, but resulted in high case fatality rate (CFR) due to critical demand of health resources. Other country-based factors such as sociocultural issues, ageing population etc., has also influenced practical effectiveness of taking interventions to improve morality in early phase. To better understand the relationship of these factors across different countries with COVID-19 CFR is of primary importance to prepare for potentially second wave of COVID-19 infections. In the paper, we propose a novel regularized multi-task learning based factor analysis approach for quantifying country-based factors affecting CFR in early phase of COVID-19 epidemic. We formulate the prediction of CFR progression as a ML regression problem with observed CFR and other countries-based factors. In this formulation, all CFR related factors were categorized into 6 sectors with 27 indicators. We proposed a hybrid feature selection method combining filter, wrapper and tree-based models to calibrate initial factors for a preliminary feature interaction. Then we adopted two typical single task model (Ridge and Lasso regression) and one state-of-the-art MTFL method (fused sparse group lasso) in our formulation. The fused sparse group Lasso (FSGL) method allows the simultaneous selection of a common set of country-based factors for multiple time points of COVID-19 epidemic and also enables incorporating temporal smoothness of each factor over the whole early phase period. Finally, we proposed one novel temporal voting feature selection scheme to balance the weight instability of multiple factors in our MTFL model.

The effect of multiple interventions to balance healthcare demand for controlling COVID-19 outbreaks: a modelling study (preprint)

BackgroundRecent outbreak of a novel coronavirus disease 2019 (COVID-19) has led a rapid global spread around the world. For controlling COVID-19 outbreaks, many countries have implemented two non-pharmaceutical interventions: suppression like immediate lock-downs in cities at epicentre of outbreak; or mitigation that slows down but not stopping epidemic for reducing peak healthcare demand. Both interventions have apparent pros and cons; the effectiveness of any one intervention in isolation is limited. It is crucial but hard to know how and when to take which level of interventions tailored to the specific situation in each country. We aimed to conduct a feasibility study for robustly accessing the effect of multiple interventions to control the number and distribution of infections, growth of deaths, peaks and lengths of COVID-19 breakouts in the UK and other European countries, accounting for balance of healthcare demand. MethodsWe developed a model to attempt to infer the impact of mitigation, suppression and multiple rolling interventions for controlling COVID-19 outbreaks in the UK. Our model assumed that each intervention has equivalent effect on the reproduction number R across countries and over time; where its intensity was presented by average-number contacts with susceptible individuals as infectious individuals; early immediate intensive intervention led to increased health need and social anxiety. We considered two important features: direct link between Exposed and Recovered population, and practical healthcare demand by separation of infections into mild, moderate and critical cases. Our model was fitted and calibrated with date on cases of COVID-19 in Wuhan to estimate how suppression intervention impacted on the number and distribution of infections, growth of deaths over time during January 2020, and April 2020. We combined the calibrated model with data on the cases of COVID-19 in London and non-London regions in the UK during February 2020 and April 2020 to estimate the number and distribution of infections, growth of deaths, and healthcare demand by using multiple interventions. We applied the calibrated model to the prediction of infection and healthcare resource changes in other 6 European countries based on actual measures they have implemented during this period. FindingsWe estimated given that 1) By the date (5th March 2020) of the first report death in the UK, around 7499 people would have already been infected with the virus. After taking suppression on 23rd March, the peak of infection in the UK would have occurred between 28th March and 4th April 2020; the peak of death would have occurred between 18th April and 24th April 2020. 2) By 29th April, no significant collapse of health system in the UK have occurred, where there have been sufficient hospital beds for severe and critical cases. But in the Europe, Italy, Spain and France have experienced a 3 weeks period of shortage of hospital beds for severe and critical cases, leading to many deaths outside hospitals. 3) One optimal strategy to control COVID-19 outbreaks in the UK is to take region-level specific intervention. If taking suppression with very high intensity in London from 23rd March 2020 for 100 days, and 3 weeks rolling intervention between very high intensity and high intensity in non-London regions. The total infections and deaths in the UK were limited to 9.3 million and 143 thousand; the peak time of healthcare demand was due to the 96th day (12th May, 2020), where it needs hospital beds for 68.9 thousand severe and critical cases. 4) If taking a simultaneous 3 weeks rolling intervention between very high intensity and high intensity in all regions of the UK, the total infections and deaths increased slightly to 10 million and 154 thousand; the peak time of healthcare occurs at the 97th day (13th May, 2020), where it needs equivalent hospital beds for severe and critical cases of 73.5 thousand. 5) If too early releasing intervention intensity above moderate level and simultaneously implemented them in all regions of the UK, there would be a risk of second wave, where the total infections and deaths in the UK possibly reached to 23.4 million and 897 thousand. InterpretationConsidering social and economic costs in controlling COVID-19 outbreaks, long-term suppression is not economically viable. Our finding suggests that rolling intervention is an optimal strategy to effectively and efficiently control COVID-19 outbreaks in the UK and potential other countries for balancing healthcare demand and morality ratio. As for huge difference of population density and social distancing between different regions in the UK, it is more appropriate to implement regional level specific intervention with varied intensities and maintenance periods. We suggest an intervention strategy to the UK that take a consistent suppression in London for 100 days and 3 weeks rolling intervention in other regions. This strategy would reduce the overall infections and deaths of COVID-19 outbreaks, and balance healthcare demand in the UK.

Feasibility of Controlling COVID-19 Outbreaks in the UK by Rolling Interventions (preprint)

Background: Recent outbreak of a novel coronavirus disease 2019 (COVID-19) in China has led a rapid global spread around the world. For controlling COVID-19 outbreaks, many countries have implemented two non-pharmaceutical interventions: suppression like immediate lockdowns in cities at epicentre of outbreak; or mitigation that slows down but not stopping epidemic for reducing peak healthcare demand. Both interventions have apparent pros and cons; the effectiveness of any one intervention in isolation is limited. We aimed to conduct a feasibility study for robustly estimating the number and distribution of infections, growth of deaths, peaks and lengths of COVID-19 breakouts by taking multiple pharmaceutical interventions in London and the UK, accounting for reduction of healthcare demand. Methods: We developed a model to attempt to infer the impact of mitigation, suppression and multiple rolling interventions for controlling COVID-19 outbreaks in London and the UK. Our model assumed that each intervention has equivalent effect on the reproduction number R across countries and over time; where its intensity was presented by average-number contacts with susceptible individuals as infectious individuals; early immediate intensive intervention led to increased health need and social anxiety. We considered two important features: direct link between Exposed and Recovered population, and practical healthcare demand by separation of infections into mild and critical cases. Our model was fitted and calibrated with data on cases of COVID-19 in Wuhan and Hubei to estimate how suppression intervention impacted on the number and distribution of infections, growth of deaths over time during January 2020, and April 2020. We combined the calibrated model with data on the cases of COVID-19 in London, the UK (non-London) and the UK during February 2020 and March 2020 to estimate the number and distribution of infections, growth of deaths, and healthcare demand by using multiple interventions. Findings: We estimated given that multiple interventions with an intensity range from 3 to 15, one optimal strategy was to take suppression with intensity 3 in London from 23rd March for 100 days, and 3 weeks rolling intervention with intensity between 3 and 5 in non-London regions. In this scenario, the total infections and deaths in the UK were limited to 2.43 million and 33.8 thousand; the peak time of healthcare demand was due to the 65th day (April 11th), where it needs hospital beds for 25.3 thousand severe and critical cases. If we took a simultaneous 3 weeks rolling intervention with intensity between 3 and 5 in all regions of the UK, the total infections and deaths increased slightly to 2.69 million and 37 thousand; the peak time of healthcare kept the same at the 65th day, where it needs equivalent hospital beds for severe and critical cases of 25.3 thousand. But if we released high band of rolling intervention intensity to 6 or 8 and simultaneously implemented them in all regions of the UK, the COVID-19 outbreak would not end in 1 year and distribute a multi-modal mode, where the total infections and deaths in the UK possibly reached to 16.2 million and 257 thousand.

Feasibility Study of Mitigation and Suppression Intervention Strategies for Controlling COVID-19 Outbreaks in London and Wuhan (preprint)

Recent outbreaks of coronavirus disease 2019 (COVID-19) has led a global pandemic cross the world. Most countries took two main interventions: suppression like immediate lockdown cities at epicentre or mitigation that slows down but not stopping epidemic for reducing peak healthcare demand. Both strategies have their apparent merits and limitations; it becomes extremely hard to conduct one intervention as the most feasible way to all countries. Targeting at this problem, this paper conducted a feasibility study by defining a mathematical model named SEMCR that can access effectiveness of mitigation, suppression and hybrid interventions for controlling COVID-19 outbreaks in London and Wuhan. It extended traditional SEIR (Susceptible-Exposed Infectious-Recovered) model by adding two key features: a direct connection between Exposed and Recovered populations, and separating infections into mild and critical cases. It defined parameters to classify two stages of COVID-19 control: active contain by isolation of cases and contacts, passive contain by suppression or mitigation. The model was fitted and evaluated with public dataset containing daily number of confirmed active cases including Wuhan and London during January, 2020 and March 2020. The simulated results showed that 1) Immediate suppression taken in Wuhan significantly reduced the total exposed and infectious populations to 119610, but it has to be consistently maintained at least 90 days (by the middle of April 2020). Its success heavily relied on sufficiently external support from other places of China. This mode were not suitable to other countries that have no sufficient health resources. 2) In London, it is possible to take a hybrid intervention of suppression and mitigation for every 2 or 3 weeks over a longer period to balance the total infections and economic loss. While the total infectious populations in this scenario would be possibly 2 times than the one taking suppression, economic loss and recovery of London would be less affected. 3) Both in Wuhan and London cases, one important issue of fitting practical data was that there were a large portion (probably 62.9% in Wuhan) of self-recovered populations that were asymptomatic or mild symptomatic. These people might think they have been healthy at home and did not go to hospital for COVID-19 tests. Early release of intervention intensity potentially increased a risk of the second outbreak. One limitation of our model was that our prediction of infections and deaths depended on a parameter estimation of intervention intensity that presented by average number contacts with susceptible individuals as infectious individuals in a certain region. It assumed that each intervention had equivalent effects on the reproduction number R in different regions over time. Practical effectiveness of implementing intervention intensity might be varied with respect to cultures or other issues of certain county.