This repository contains the code for the Medical Examiner Open Data Pipeline.
We currently fetch data from the following sources:
The results of this data are used in various other analysis here on GitHub:
This repo exists mainly to take advantage of GitHub actions for automation.
The actions workflow is located in .github/workflows/pipeline.yml
and is triggered weekly or manually.
This workflow fetches data from the configured data sources inside config.json
,
geocodes addresses (when available) using ArcGIS, extracts drugs using the drug extraction toolbox
and then compiles and zips up the results into the GitHub Releases page.
The data is then available for download from the releases page page.
Further, the entire workflow effectively runs a series of commands using the CLI application opendata-pipeline
which is located in the src
directory.
This is also available via a docker image hosted on ghcr.io. The benefits of using the CLI via a docker image is that you don't have to have Python3.10 or the drug toolbox on your local machine 🙂.
We utilize async methods to speed up the large number of web requests we make to the data sources.
It is important to regularly fetch/pull from this repo to maintain an updated
config.json
We currently do not guarantee Windows support unfortunately. If you want to help make that a reality, please submit a new Pull Request
There is further API-documentation available on the GitHub Pages website for this repo if you want to interact with the CLI.
I would recommend using the docker image as it is easier to use and always referring to the CLI --help
for more information.
The workflow can best be described by looking at the pipeline.yml
file.
The following table shows the fields that we add to the original datafiles:
Column Name | Description |
---|---|
CaseIdentifier |
A unique identifier across all the datasets. |
death_day |
Day of the Month death occurred |
death_month |
Month Name death occurred |
death_month_num |
Month Number death occurred |
death_year |
Year death occurred |
death_day_of_week |
Day of week death occurred. Starting with 0 on Monday. Weekends are 5 (Saturday) & 6 (Sunday). |
death_day_is_weekend |
Death occurred on weekend day |
death_day_week_of_year |
Week of the year (of 52) that death occurred |
geocoded_latitude |
Geocoded latitude. |
geocoded_longitude |
Geocoded longitude. |
geocoded_score |
Confidence of geocoding. 70-100. |
geocoded_address |
The address that the geocoded results correspond to. Not the address provided to the geocoder. |
In addition to providing the extracted drugs as a separate file in each release, we also convert this data to wide-form for each dataset. This adds the following columns in the subsequent pattern:
Column Name/Pattern | Description |
---|---|
*_1 |
* drug found in first search column provided in drug configuration |
*_2 |
* drug found in second search column provided in drug configuration |
*_meta |
Drug of * category/class found in this record across any search column. |
To install the python cli I recommend using pipx.
pipx install opendata-pipeline
To install the docker image, you can use the following command:
docker pull ghcr.io/uk-ipop/opendata-pipeline:latest
Usage is very similar to any other command line application. The most important thing is to follow the workflow defined above.
Pull requests are welcome. For major changes, please open an issue first to discuss what you would like to change.
Help me write some tests!
If you use this software or the enhanced data, please cite this repository:
@software{Anthony_Medical_Examiner_OpenData_2022,
author = {Anthony, Nicholas},
month = {9},
title = {{Medical Examiner OpenData Pipeline}},
url = {https://github.com/UK-IPOP/open-data-pipeline},
version = {0.2.1},
year = {2022}
}
Thank you.