The goal of this dashboard is to provide an epidemiological overview of the COVID-19 pandemic, incorporating data from multiple sources.
It is being actively developed by the data science team at Epicentre MSF.
The tool is built using the R programming language and the shiny web framework. Packages used include the tidyverse, leaflet and highcharter.
If you are part of the MSF network, you can access further Epicentre COVID-19 content on our Sharepoint site.
Trends are estimated daily on the daily number of cases and deaths in JHU CSSE data observed over a period of 14 days. The most recent 2 days of data are not considered in the 14 day period, as counts may not yet be complete due to reporting delays. For example, if data is available as of the 16th of a given month, trends will be calculated on the period from 1st-14th of that month.
On the 14-day time-series we first carry a smoothing using a 3-day moving average. As a result, we obtain 12-day smoothed values for which we run a linear regression of the values in the natural logarithm scale using the following formula:
lm(ln(smoothed values) ~ 12 days)
The standard error of the model was used to calculate the confidence intervals.
Trends presented in the report were defined using the coefficients of the linear regression as follows:
To ensure reliable estimates we estimate the trend only if the cumulative number cases (or deaths) during the 14-day period was higher than 30.