Health Services/Policy
Evaluating Lack of Treatment Arm Exchangeability and Conditional Exchangeability in the End-Stage Renal Disease Treatment Choices (ETC) Medicare Payment Model Randomized Controlled Trial Ariana Mora* Ariana Mora Kelsey Drewry Daeho Kim Kalli Koukounas Amal Trivedi Adam Wilk Rachel Patzer
Objective: Evaluate lack of treatment arm exchangeability for Medicare’s largest RCT, ESRD Treatment Choices (ETC), with particular focus on disparities and underlying ESRD trends by race/ethnicity prior to policy implementation.
Methods: Construction of a pre-policy cohort (2015-2019) using the United States Renal Data System national registry using ETC evaluation metrics (patients attributed to dialysis facilities as patient-months (p-m); ESRD outcomes: dialysis, waitlisting, and living donor kidney transplant). Conditional exchangeability assessed using multilevel fixed effects models.
Results: 25,797,647 p-m (n=784,777 patients) met inclusion for ETC pre-implementation analysis (ETC: 8,203,393 p-m; Control: 17,594,254 p-m). Patients treated at ETC and Control facilities had similar age, sex, BMI, pre-ESRD nephrology care, and dialysis modality, but otherwise were non-exchangeable. (See figure) Disproportionate ETC allocation within ESRD Networks occurred as a function of the randomization unit (ZIP Code clusters), leading to geospatial clustering and higher/lower than expected ethno-racial distribution by ETC assignment and ESRD Network. Adjusting for patient- and facility-level covariates and spatial-temporal fixed effects was unable to achieve conditional exchangeability by race/ethnicity for ESRD outcomes. Adjusted odds ETC:Control assignment differed by race/ethnicity: dialysis (ETC 5-8% higher), waitlisting (ETC 3-21% lower), and living donor transplant (ETC 4-33% lower).
Conclusions: There are underlying differences between ETC and Control groups prior to policy implementation, particularly with respect to race/ethnicity, that have resulted in non-exchangeability and confounding at baseline. Future evaluation of ETC Policy must account for cluster randomization failure and particular attention given to avoid exacerbating underlying ethno-racial disparities that exist between ETC and Control patient populations at baseline prior to policy implementation.
