This repository contains code for the paper "Algorithmic Changes Are Not Enough: Evaluating the Removal of Race Adjustment from the eGFR Equation" by Marika Cusick, Glenn Chertow, Doug Owens, Michelle Williams, and Sherri Rose (2024), Proceedings of the Conference on Health, Inference, and Learning. [Link] [Policy Brief]
While the analytical code is made available, our data is generated from Stanford Health Care electronic health record data that includes patient identifiers and protected health information. These data cannot be accessed without approval from the Stanford Institutional Review Board. We have replaced all references to data tables with 'X'.
Our code is written in SQL (BigQuery) and python version 3.9.1. Required python packages include pandas (version 2.0.3), google.cloud (3.3.5), matplotlib (3.3.3), numpy (version 1.25.1), and statsmodels (0.14.0).
-
Create tables for analysis Code: OMOP_queries.sql runs the SQL code to generate tables required for our analysis. Data from SHC is in the Observational Medical Outcomes Partnership (OMOP) common data model format.
-
Analysis of nephrology referrals Code: kidney_clinic_referral_1_31_24.py runs the interrupted time series regression to assess changes to the quarterly rates of nephrology referrals at SHC after the implementation of the eGFR formula without race adjustment (CKD-EPI 2021).
-
Analysis of nephrology visits Code: kidney_clinic_visit_1_31_24.py runs the interrupted time series regression to assess changes to the quarterly rates of nephrology visits at SHC after the implementation of the eGFR formula without race adjustment (CKD-EPI 2021).