Translate ALT key measure to ehrQL #24
Conversation
| analysis/measure_definition.py | ||
| --output output/alt_tests/measures.arrow | ||
| -- | ||
| --codelist codelists/opensafely-alanine-aminotransferase-alt-tests.csv |
There was a problem hiding this comment.
I didn't think of this when we were pairing (you probably did), but the easiest way to pass a codelist as a parameter to the measure definition is, well, to pass a codelist as a parameter!
| run: > | ||
| ehrql:v1 generate-measures | ||
| analysis/measure_definition.py | ||
| --output output/alt_tests/measures.arrow |
There was a problem hiding this comment.
The name of the sub-directory need not match the name of the codelist, because the action tells the reader which sub-directory is associated with which codelist. If a downstream action needs to associate a sub-directory with a codelist, then we may revisit this design decision.
| outputs: | ||
| highly_sensitive: | ||
| dataset: output/dataset.csv.gz | ||
| measures: output/alt_tests/measures.arrow |
There was a problem hiding this comment.
We can read an Arrow (Feather) file into a Pandas data frame with:
import pandas
pandas.read_feather("output/alt_tests/measures.arrow")If we don't have Pandas installed on our system, then we can start JupyterLab in a container that does have Pandas installed with:
opensafely jupyter
The measure definition is parametrized by a codelist, such that it can be applied to any number of codelists. The measures file is written to a sub-directory of the `outputs` directory, because we anticipate a downstream transformation action reading the measures file and writing a transformed file, suitable for releasing and publishing. Co-authored-by: Alice Wong <[email protected]>
iaindillingham
force-pushed
the
iaindillingham/alt-key-measure
branch
from
bce6e72 to
d9dd78f
Compare
| def num_months(from_, to_): | ||
| """Returns the number of months between the given dates.""" | ||
| return abs(to_.year - from_.year) * 12 + abs(from_.month - to_.month) | ||
|
|
||
|
|
||
| # same year, same month | ||
| assert num_months(date(2025, 1, 1), date(2025, 1, 31)) == 0 | ||
| # different year, different month | ||
| assert num_months(date(2024, 1, 1), date(2025, 2, 1)) == 13 | ||
| # to_ before from_ | ||
| assert num_months(date(2025, 2, 1), date(2024, 1, 1)) == 13 |
There was a problem hiding this comment.
I realised that there was a bug in num_months. I've fixed it, and have added some lightweight tests: namely, the three asserts here. I'd like to highlight that they're lightweight tests. One day, we may choose to move them to test functions, which we may choose to call with, for example, pytest. Not today, however 🙂
The measure definition is parametrized by a codelist, such that it can be applied to any number of codelists. The measures file is written to a sub-directory of the
outputsdirectory, because we anticipate a downstream transformation action reading the measures file and writing a transformed file, suitable for releasing and publishing.