COVID19_undetected

Data on COVID19 cases observed outside of China. Includes Hubei and China travel history information.

variables explained

repatriated -- whether the case is from a repatriation flight from Wuhan. This doesn't include repatriates from the Diamond princess date_report_dd_mm_yyyy -- date which the case was reported in the press/ministry of health reports. there is usually a lag following admission to hospital and/or confirmation country -- country where the case is reported and confirmed. mode_of_transport_plain_train -- if the case travelled from Wuhan, by which mode of transport did they travel. travel_history_to_hubei_y_n -- yes/no variable indicating whether that individual had any travel history to Hubei province local_transmission_y_n -- local transmission is defined as any transmission outside of China.

Clarification

If both travel history and local transmission variables are "n", this means that the case was imported from China but not from Hubei province. For some cases, it is known they are imported from China but not from where. In this case travel history would be noted as "unknown".

Running the analysis

To reproduce the analysis, first clone or download this repository. Navigate to the downloaded repository on your computer. To process the list of exported cases, in a new R session, source the R script data_processing.R.

source('data_processing.R')

This will read the file exported_cases.csv, and aggregate the rows into a list of cases detected overseas (see Methods in the main text). Then, use rmarkdown to render the file analysis.Rmd.

rmarkdown::render('analysis.Rmd')

This analysis uses a list of "reference" countries (see Methods). Reference countries can be changed by editing the object called possible_ref_countries (Line 49). We then read in data on the volume of air travel between countries. Our analysis relied on data from IATA, which cannot be shared publicly. We have instead provided a file with the same structure as used for our work, but with dummy numbers.

Finally, the data on exported cases is matched with travel volume data and used to estimate the surveillance sensitivity as described in the paper. The likelihood function (called likelihood) is implemented in the file likelihood.R.