Metadata-Version: 2.4
Name: CDSL-MEDS
Version: 0.0.2
Summary: An ETL pipeline to extract the COVID Data for Shared Learning dataset into MEDS.
Author: Medical Event Data Standard
Project-URL: Homepage, https://physionet.org/content/covid-data-shared-learning/1.0.0/
Classifier: Programming Language :: Python :: 3
Classifier: License :: OSI Approved :: MIT License
Classifier: Operating System :: OS Independent
Requires-Python: <3.14,>=3.11.4
Description-Content-Type: text/markdown
License-File: LICENSE
Requires-Dist: meds-extract[download]<0.8,>=0.7.1
Dynamic: license-file

# CDSL MEDS ETL

Configuration-first [MEDS-Extract](https://github.com/mmcdermott/MEDS_extract) ETL for
[COVID Data for Shared Learning (CDSL) v1.0.0](https://physionet.org/content/covid-data-shared-learning/1.0.0/).
It maps the tabular EHR data and DICOM metadata, never the `IMAGES/` tree.

````
## Run

Three CDSL files are Latin-1 encoded, while MEDS-Extract's CSV reader requires UTF-8.
The installed `meds-extract-run` stages a UTF-8 view automatically, without changing
the downloaded files:

```bash
uv run meds-extract-run spec=CDSL output_dir=cdsl_meds \
	do_download=false input_dir=raw_cdsl
````

Set `input_dir` to the directory that directly contains `patient_01.csv` if your
download layout differs. The internal staging step hard-links every other source file
and transcodes only `atc7.csv`, `medication_05.csv`, and `vital_signs_04.csv`.

## Mapping

CDSL's dates are deidentified but preserve within-patient chronology and are emitted
directly. Diagnosis, procedure, and neoplasm tables supply no timestamps, so those events
use the explicit `1970-01-01` pseudotime anchor. See
[docs/MEDS_ETL_MAPPING.md](docs/MEDS_ETL_MAPPING.md) for the table-level mapping and
limitations.
