Forecasts are shown for mortality, severe cases requiring hospital care, proportion of working days lost, and case detection by tests (rapid tests and/or RT-PCR) in Dhaka District. These metrics are shown both as time series and summarized over 12 months (barplots to the right).
The costs effectiveness of the intervention scenarios is described in the bottom three barplots. Total costs include costs of implementing the selected interventions (e.g. from advertising, community support teams, mask distribution, and tests) and providing hospital care for COVID-19 patients. The cost per death averted is calculated by comparison with a scenario with no interventions implemented. % return on investment (ROI) (in terms of saved healthcare costs) is also estimated relative to no interventions.
All plots show outputs for both a baseline scenario and a comparison scenario so the impact of interventions can be compared. The interventions applied can be adjusted via the 'Comparison' and 'Baseline' tabs in the sidebar.

Excess daily mortality

2020


Hospital Demand

2020


Working days lost

2020


Cases & detection

2020


Total Costs

Cost/Death Averted

%ROI



This tab can be used to generate forecasts for 2021 in populations of different sizes and age distributions (defaults represent Dhaka District).
The number of days to forecast following 1st March 2021 (beginning of the third COVID-19 wave) can be selected (minimum 30), along with starting infections and the percentage of the population immune (from prior infection or vaccination) at the forecast start. Interventions applied during the forecast period can be selected in the 2021 tab of the sidebar. Once selected, inputs need to be confirmed and the model run using the grey button below. The parameters that can be adjusted in this tab can be returned to their default settings by pressing the 'Reset defaults' button
The proportions of cases that are symptomatic, hospitalised and fatal increase as the average age of the population increases. The age distribution can be adjusted at the bottom of this tab.

Defaults to the estimated population of Dhaka District in 2020.

Forecast


Forecast deaths


Age distribution

Defaults to the overall age distribution for Dhaka District estimated by the 2011 census.



We use an SEIR model to explore impacts of control measures on COVID-19 transmission in Dhaka District. R0 and the impact of lockdown were tuned to match the trend in deaths prior to June 2020 (see early epidemic forecasts below). Other epidemiological and population parameters were obtained from the literature. Parameters and their sources are detailed in the tables below. Most parameters describing the interventions, including timing, compliance, and impacts on transmission, can be adjusted in the sidebar tabs.
A detailed description of the model and of how this app has been used in Bangladesh, along with further analyses, can be found in our preprint. R code for the model and app can be found in our Github repository.

Forecast vs reports (early epidemic)

Forecast vs reports (2020)


Forecast vs reported deaths (early epidemic)

Forecast vs reported deaths (2020)


Model Schematic


Epidemiological Parameters


Intervention/Testing Parameters


Age-dependent severity of infection


Population parameters






Background

This tool was developed in collaboration with multiple partners in early 2020. A detailed description of the model and of how this app has been used in Bangladesh, along with further analyses, can be found in our preprint. Code for the model and app is in our Github repository.
To understand how the epidemic could progress in Bangladesh and the potential impacts of different responses we developed a relatively simple deterministic SEIR framework. We assume persons infected with the virus are either asymptomatic for the duration of their infectious period or enter a pre-symptomatic infectious state before progressing to symptomatic infection. A proportion of symptomatic people are hospitalised and/or die. The rates and probabilities of movement between states are detailed in the technical details tab, informed by published studies.
We assume that transmission can be reduced through response measures, but the size of the reduction depends on how well interventions are implemented, adhered to and enforced. i.e. as a result of investment, technological and sociological capacity, communications, trust etc.
The timing and duration of the epidemic and associated responses will have major impacts on morbidity, mortality and the economy, with potential to overwhelm health systems and cause catastrophic consequences for families, communities and society as a whole. These knock on effects are beyond the scope of our model. However, we attempt to lay out short-term impacts on morbidity and hospital demand relative to capacity. Immediate economic costs, from intervention implementation, testing, and healthcare provision, and from loss of work due to illness or restrictions, are also explored. We summarize these impacts over 12 months to better understand longer-term consequences of decisions that need to be taken quickly. We further calibrate the model to the resurgence in 2021 to better understand factors (R0 of new variants, prior immunity, NPIs) that led to the resurgence and the potential for control measures to mitigate impacts.
There is considerable uncertainty in quantitative predictions therefore we focus on order of magnitude impacts. We caveat that while this framework helps us to understand the consequences of different decisions, outcomes depend on how interventions are delivered and complied with. We use publicly available data on the course of the pandemic to calibrate the model, and caution that recorded cases and deaths are underestimated in most countries. Caution is therefore needed in comparing model trajectories and data.