COVID-19 planning tools
Since the emergence of the new coronavirus (COVID-19) in December 2019, we have adopted a policy of immediately sharing research findings on the developing pandemic. This page provides access to the planning tools developed by the Imperial College COVID-19 Response Team.
Planning tools
Background
The multiple planning tools presented on this page, provide alternative analyses of COVID-19 transmissibility and forecasts future reported cases or deaths for many countries. The analyses rely on distinct models and data, and therefore results may vary accordingly. Each forecast comes with its own strengths and caveats and they should all be interpreted with caution.
The models
LMIC short-term forecasts dashboard (find out more)
Daily updated estimates for Europe (find out more)
Weekly short-term global forecasts (find out more)
Why use more than one model?
Evidence in science rarely comes from a single study but is rather a collaborative effort. The more results, based on independent data and models, accumulate and converge toward the same interpretation, the greater the confidence in the results.
The Imperial College London COVID-19 Response Team is analysing data using multiple models for countries around the world, producing multiple planning tools. See below for a note on the UK.
To help interpret those forecasts, we outline below some of the key assumptions underlying the different models:
Key model characteristics
In epidemiological model can use a combination of mechanistic and statistical approaches.
| Type of model | Explicitly accounts for interventions | Model parameters shared across countries | |
|---|---|---|---|
| LMIC short-term forecasts dashboard | Mechanistic +++ Statistical + |
Yes | Yes |
| Daily updated estimates for Europe | Mechanistic ++ Statistical ++ |
Yes | Yes |
| Weekly short-term global forecasts | Mechanistic + Statistical +++ |
No | No |
| The number of "+" signs indicate the relative focus on mechanistic and/or statistical approaches | |||
Mechanistic model: Explicitly accounts for the underlying mechanisms of diseases transmission and attempt to identify the drivers of transmissibility. Rely on more assumptions about the disease dynamics.
Statistical model: Do not explicitly model the mechanism of transmission. Infer trends in either transmissibility or deaths from patterns in the data. Rely on fewer assumptions about the disease dynamics.
Mechanistic models can provide nuanced insights into severity and transmission but require specification of parameters – all of which have underlying uncertainty. Statistical models typically have fewer parameters. Uncertainty is therefore easier to propagate in these models. However, they cannot then inform questions about underlying mechanisms of spread and severity.
Data used
| Reported deaths | Reported cases | Country demography | ICU / Hospitalisation | |
|---|---|---|---|---|
| LMIC short-term forecasts dashboard | Yes (ECDC) | Yes (ECDC) | Yes | No |
| Daily updated estimates for Europe | Yes (ECDC) | Yes (ECDC) | Yes | No |
| Weekly short-term global forecasts | Yes (ECDC) | 1 of 3 approaches (ECDC) | No | No |
Strengths and caveats
Further detailed description of the methods underlying each model and links to the code are available on their respective pages.
| Strengths | Caveats | Method description and code | |
|---|---|---|---|
| LMIC short-term forecasts dashboard | Explores alternative scenarios for future interventions. Forecasts health system demand as well as future deaths. |
Projections of healthcare demand assumes reliable data on cumulative deaths. | Website |
| Daily updated estimates for Europe | Robust statistical and mechanistic framework to evaluate the ipact of control strategies. Provide estimate of the true epidemic size. |
Model outputs are sensitive to assumptions about intervention efficacy. | Website |
| Weekly short-term global forecasts |
Agnostic about control strategies and how effectively they are implemented. More robust to rapid changes in reporting. |
Does not predict the impact of control, therefore subject to greater delay in estimating their impact (due to delay between infections and deaths). | Website |
Note on the United Kingdom
The advice to the UK government has been based on multiple models and multiple data streams. Those models have tended to be more complex, reflecting:
- The wealth of alternative data available from the UK, e.g. ICU demand and capacity; and
- The range of questions addressed, e.g. impact of lifting some control measures.
Reference to those models can be found here.