Maintenance release

Fixes #21, #25, #26, #28 - handle cases where there is no inpatient data. Also fixed some install issues related to having different versions of various dependencies.

Improved PyPI install

Fixes #24

Support for newer versions of Pandas

Fixes #21 - import error when using the latest version of Pandas

Fixes Excel input files

• Fixes handling of Excel files for input
• Fixes some broken links in the README.

All adults model

This release is a significant update

All ages model

This release incorporates a new model that is now appropriate for adults ages 18 and over. This model was trained on a combination of the original CMS Medicare data along with additional data provided by HealthFirst.

The original Medicare only 'xgboost' model is still available by adding a -m xgboost option to cv19index. However, even for Medicare populations we recommend moving to the new xgboost_all_ages model. This model is now the default.

Other updates

• The README has been completely rewritten to make the usage more clear.
• We have added prebuilt whl files to make instlalation on windows easier.
• Documented the input and output formats more clearly.
• Removed the old preprocessing code from version 1.0.0
• Corrected a bug with the "# of Admissions" feature.
• Corrected a bug where 3 character ICD-10 codes would not be mapped to CCSR.
• Added a "features.csv" option to enable users to see the result of preprocessing.
• Added a "run_cv19index.py" script that neables running the package without installing from PyPI.
• Several asserts have been added to verify that data types and row counts are correct through the code.

Acknowledgements

Many thanks to HealthFirst for being one of the first users of the model and for allowing us to use their data in order to create a model for all ages.

Simplified Usage

We have simplified the library to take in just 2 files. A demographics file and a claims files. These need to have a few key columns, which is outlined in cv19index/resources/xgboost/demographics.schema.json and cv19index/resources/xgboost/claims.schema.json. The column names must match and are case sensitive. These file can have other columns. The core goal is for a simple dump of these two datasets (demographics and claims) to be a basis for building the model with minimal changes.

SageMaker updates

Converted the model name from model_medium to xgboost and added support for SageMaker