Exploring Missing Data in the Enhanced Tuberculosis Surveillance System

Sam Abbott, Hannah Christensen, Ellen Brooks-Pollock

Background

The Enhanced Tuberculosis Surveillance (ETS) system is a routine surveillance system that collects data on all notified tuberculosis (TB) cases in England and is routinely used in research. Routine data often has a large amount of missing data which may introduce bias to analyses.

Aim

Explore potential sources of bias from missing data for key outcomes in the ETS system.

Methods

We used data on all notifications in the ETS for England (2000-2015). We summarised the proportion of missing data and then used logistic regression to explore the associations between missingness and demographic variables for the following outcomes: drug resistance; BCG status and year; date of symptom onset, diagnosis, notification, starting treatment, finishing treatment and death; and cause of death. For all date variables, we visualised the distribution annually and by month, identifying variables at risk of bias.

Results

All demographic variables considered were associated with data being missing for multiple outcomes. Missingness was not associated with all demographic variables for all outcomes. Associations that were present did not all have the same direction of effect. We found that the date of symptom onset and the date of ending treatment had a high proportion of cases occurring in January and on the 1st and 14th of each month indicating the presence of potential bias.

Conclusions

Missingness in outcomes was associated with multiple demographic variables. These associations could not be generalised, preventing a systematic approach to dealing with missingness and indicating that domain knowledge is required. The identified associations may induce bias, meaning that multiple imputation may be beneficial for analyse using this data source - or other similar surveillance data. Our findings should be used to motivate the use of imputation and to inform the specification of imputation models that include auxiliary variables.

Keywords

Tuberculosis, surveillance, missingness, imputation

Reproducibility

Repository structure

The repository is structured as an R package. It has the following structure:

data-raw: Raw data processing.
data: Processed data.
R: Supporting R functions.
docs: Documentation for R code.
vignettes: Analysis paper, results, and analysis plan.
peer-review: Documentation required for peer review.

Manual install

Install R (analysis run with 3.6.1) and Rstudio (alternatively use Docker as outlined below).
Download the analysis folder from https://github.com/seabbs/ETSMissing/archive/master.zip or use git clone, as follows, in the command line (not the R terminal).

git clone https://github.com/seabbs/ETSMissing.git

Once this has been downloaded click on the project file (ETSMissing.Rproj).
Install the analysis dependencies and build the package using the following. To enable more robust reproducibility consider using the checkpoint package versioned locked to R 3.6.1.

#install.packages("devtools")
# To build locally
devtools::install_dev_deps(dependencies = TRUE)
devtools::install()
# Alternatively to remote install
devtools::install_github("seabbs/ETSMissing", dependencies = TRUE)

Load the analysis results by running vignettes/paper.Rmd. Alternatively the complete analysis (along with documentation) can be reconstructed using make in the project root directory.
See data-raw for data processing and the documentation for implementation details.

Docker

This analysis was developed in a docker container based on the tidyverse docker image. To run the docker image run:

docker run -d -p 8787:8787 --name etsmissing -e USER=etsmissing -e PASSWORD=etsmissing seabbs/etsmissing

The rstudio client can be found on port :8787 at your local machines ip. The default username:password is assessbcgpolicychange:assessbcgpolicychange, set the user with -e USER=username, and the password with - e PASSWORD=newpasswordhere. The default is to save the analysis files into the user directory.

If you have access to the required underlying raw data (see tbinenglanddataclean) then the entire analysis can be reproduced from scratch by adding the following to the docker run command, with the data saved into data/tb_data. The data requirements, and structure, can be found here.

--mount type=bind,source=$(pwd)/data/tb_data,target=/home/ETSMissing/data/tb_data

Alternatively the analysis environment can be accessed via binder.