wusm-lgm-informatics-section / adlm-fairlabs-competition Goto Github PK

The dashboard is very busy and dense with information. It would be less overwhelming if you organized the dashboard with additional tabs. I would suggest the following:

1. Home page - welcome, project overview, developer team
2. Methods
3. Demographic parity
4. Predictive parity
5. equalized odds
6. equalized outcomes
7. statistics

The FairLabs Challenge was an especially difficult ADLM data analytics competition. FairLabs asked participants to go beyond simple data analysis to create a sharable tool that effectively presents fairness metrics and actionable information to drive meaningful improvement in health equity. Competitors were also incentivized to engage with local experts and datasets to use their tools to further health equity at their own institutions.

The FairLabs Challenge closed with seven outstanding entries. These were carefully reviewed by judges representing ADLM’s Health Equity & Access Division, Informatics Division, and Data Analytics Steering Committee. All the entries were very impressive, and we thank everyone for participating.

I am very pleased to announce and congratulate the University of Washington Department of Laboratory Medicine & Pathology Team as the winner of the FairLabs Challenge. (See their entry here.) The judges also recognized the Tricore Clinical Innovations Team as first runners-up (here) and the Oregon Health & Science University team as second runners-up (here).

Be sure to attend the Health Equity & Access Division breakfast meeting during ADLM 2024 to learn more about applying data science to health equity and the winning entry.

University of Washington Department of Laboratory Medicine & Pathology FairLabs Team
Nathan Breit – Analytics Lead
Jing Zhang – QI Data Analyst
Joyce Liao – Assistant Professor
Kate Crawford – Clinical Pathology Resident

For the winning submission, the statistical tables are helpful in better understanding the likely impacts of the intervention. However the odds ratios as currently constructed are a bit challenging to interpret because it is often easier to conceptualize increased odds rather than decreased odds, using the majority population as the reference population. So my recommendation would be to adjust the statistics tables to use the white population as the reference. In addition using the log of the OR also adds a little more complexity so it would be good to evaluate whether the OR without a log transformation makes sense.

Thank you for participating in the FairLabs competition! I just wanted to remind you that your contributions are due by May 15th, 2024. Please submit your contributions by opening a pull request on the parent repo. If you have any issues or questions on how to do so, we are happy to help you out.

@acarriker
@ditopcu
@Downssi
@drnalan
@jvonreusner
@liskut
@MDodd13
@mkbohn
@mogulboy88
@njr-ahmed
@nspies13
@oronila
@pmunk894
@rebeccagreenblatt
@rwcitek
@Samwise327
@zhangly811
@MarkZaydman