Global short-term forecasting of COVID-19 cases

Abstract

The continuously growing number of COVID-19 cases pressures healthcare services worldwide. Accurate short-term forecasting is thus vital to support country-level policy making. The strategies adopted by countries to combat the pandemic vary, generating different uncertainty levels about the actual number of cases. Accounting for the hierarchical structure of the data and accommodating extra-variability is therefore fundamental. We introduce a new modelling framework to describe the pandemic’s course with great accuracy and provide short-term daily forecasts for every country in the world. We show that our model generates highly accurate forecasts up to seven days ahead and use estimated model components to cluster countries based on recent events. We introduce statistical novelty in terms of modelling the autoregressive parameter as a function of time, increasing predictive power and flexibility to adapt to each country. Our model can also be used to forecast the number of deaths, study the effects of covariates (such as lockdown policies), and generate forecasts for smaller regions within countries. Consequently, it has substantial implications for global planning and decision making. We present forecasts and make all results freely available to any country in the world through an online Shiny dashboard.

Dynamic hierarchical state space forecasting

“…(1.3) For parametric true functions, a sine function with unknown period and unknown magnitude and AR (1) error process with unknown AR(1) parameters is used. (2) Scenario 2: The sum of i) an AR (1) shared by all the units with the AR(1) coefficient 𝜙 1 = 0.5 and AR innovation variance 𝜎 2 a1 = 1, and ii) a unit-specific AR(1) with 𝜙 2 = −0.5 and 𝜎 2 a2 = 1, with and without an additional error term e ij ∼ N(0, 𝜎 2 e = 0.1). (2.1) For the proposed method, the true model is used, that is, an AR(1) for the shared component and another AR(1) for the unit-specific component.…”

“…One motivating example is the forecasting of new cases in the early stages of the COVID-19 pandemic, which are crucial to health care resource planning and policy making. Various forecasting methods have been applied, modified, or proposed in response to this public health emergency, including the classical autoregressive integrated moving average models (ARIMA), 1 epidemiology compartmental models such as the susceptible-infectious-removed (SIR) model and its variants, 2 machine learning methods, 3 classical regression methods using linear or nonlinear curves, [4][5][6] and hybrids of different models, 7 which all have focused on forecasting a single unit into the future time. Figure 1 displays the logarithms of new cases per 100 000 people for a few U.S. states.…”

Attributable Fraction and Forecasting for COVID-19 Confirmed Cases in Nigeria Using Facebook- Prophet Machine Learning Model

“…[26] Mathematical model and machine learning model Study conducted in one country and forecasts for more than 48 days were not considered. [27] State-space hierarchical models The model cannot identify the days of the week that we have a higher number of covid-19 confirmed cases.…”

COVID-19: average time from infection to death in Poland, USA, India and Germany

“…Often, the issues of confirmed case curves and the number of deaths appear when solving the predictions of virus spread. For example, Oliveira and Moral (2020) make forecasts (short-term) based on data from countries grouped for predictive purposes. A slightly different approach was proposed by Medeiros et al (2020) using data from countries where the virus appeared earlier in forecasts for Brazil.…”