Providers may face challenges in reaching ESRD Treatment Choices goals
In September 2020, CMS published the final rule for ESRD Treatment Choices, a new demonstration designed to test incentives for increasing use of home dialysis and kidney transplantation for patients with end-stage kidney disease.
The model is mandatory for approximately 30% of the Medicare-approved dialysis facilities in the United States, selected to participate by CMS, and will run through Dec. 31, 2027.
Originally, CMS proposed to include 50% of all dialysis facilities in the ESRD Treatment Choices (ETC) demonstration, stating a sample that size was needed for the model to have the statistical power necessary to detect meaningful differences in outcomes between the model participants and the “control” or comparison group (those facilities not required to participate). In the final rule, the mandatory participation sample was reduced to 30%, which CMS stated would be based on random selection of the 306 hospital referral regions (HRRs) in the United States.
HRRs are the regional markets for tertiary medical care derived from Medicare claims data as defined by the Dartmouth Atlas Project.
CMS stated that other changes to the methodology made between the proposed and final rules would compensate for any lack of statistical power lost by reducing the sample size from 50% to 30% of facilities.
The random selection process for HRR participation in ETC would include regional stratification across the four census-defined geographic regions: Northeast, South, Midwest and West.
CMS said this stratification would control for regional practice pattern variations and would help ensure ETC participants were geographically dispersed across the United States. Some commenters on the proposed rule had suggested CMS employ a more rigorous covariate-based random sampling approach that would ensure comparability across the treatment (ETC) and control groups. The commenters stated that simple random selection, as proposed, would not address other covariates that impact home dialysis and transplant rates, such as current rates of home dialysis and transplantation, rurality, insurance status and the availability of transplant centers.
CMS dismissed these suggestions, stating stratification by census region was adequate.
Given the reduced sample size of 30% to be included in ETC and the relatively simple random selection process employed, we examined whether CMS’ random sample for mandatory inclusion of facilities in ETC was a representative sample from the entire population and whether there are any inherent biases that resulted in the selection of the treatment group vs. the control group on variables that might affect performance under ETC.
General facility characteristics
Looking at a list of general dialysis facility characteristics, it appears CMS’ stratified random sample by region provided a good balance on most of these factors (see Table 1). Facilities in Puerto Rico and the U.S. territories are excluded (not in ETC and not counted in the comparison group).
There are a couple of observations to note in Table 1:
- the proportion of ETC facilities participating in ESRD Seamless Care Organizations (ESCOs) is higher (22% vs. 13% in the control group);
- the percent of treatments furnished to traditional Medicare patients is higher (62.4% vs. 58.7%); and
- the percent of facilities affiliated with the two largest dialysis chains is higher (79.6% vs. 77.3% in the control group). Most differences in proportions between ETC participants and non-ETC participants are small, however.
Perhaps of greater interest and concern is how ETC facilities compare with the control group on historical performance on home dialysis and transplantation. If the comparison group has better performance on these measures, then the bar will be set higher for ETC participants at the start of the demonstration, making it that much harder to succeed under the incentives created by ETC.
It will also be harder to measure whether the ETC model incentives create meaningful differences in outcomes between the treatment and control groups.
Table 2 reviews recent historical performance on a number of home dialysis and transplantation metrics, comparing average performance between ETC dialysis facilities and non-ETC facilities. Two measures in Table 2 closely align to the performance measures that are created under ETC: the ETC home dialysis rate and the ETC transplant rate. The home dialysis rate will be based on the percentage of patient years where patients receive dialysis at home. The 50th percentile on this measure for benchmark year 1 (based on data from July 1, 2019 to June 30, 2020) was 12.75%. The proxy for the home dialysis rate in Table 2 is 2019 home treatments as a percentage of total treatments for Medicare fee-for-service patients, which was 12.4% for ETC facilities and 13.3% for non-ETC facilities.
For all the home dialysis measures in Table 2, ETC facilities have lower scores than non-ETC facilities.
The ETC transplant rate is the sum of the transplant waitlist rate and the living donor transplant (LDT) rate.
The LDT rate is a generally small contributor to this rate, as the national average across all facilities in 2018 was 0.1%, and 1.5% of all facilities had any living donor transplants, according to FY2020 Dialysis Facility Reports.
Thus, the percent of prevalent patients waitlisted from Table 2 (found in the Dialysis Facility Compare) is a reasonably good proxy for the ETC transplant rate. In the most recent period available, ETC facilities had a waitlist rate of 17.5% compared to 19.2% for non-ETC facilities. As with the home dialysis measures, ETC facilities had historically worse performance on almost all the transplantation measures.
Based on this analysis, it looks like CMS’ random selection process did not adequately control for these most important covariates.
Social determinants of health
A third category of interest in this study was to examine how ETC participants compared to non-ETC facilities on the social determinants of health (SDOH) that have been shown to impact home dialysis rates and transplantation rates.
These underlying social and economic factors, including race, age, income, education and insurance status, can impact the chances for relative success or failure of efforts to increase home dialysis and transplant rates, depending on whether disparities exist between ETC and non-ETC facilities on these factors. Table 3 compares some key SDOH measures between ETC and non-ETC facilities.
At the county and HRR level, which is the level of aggregation of the socio demographic measures in Table 3, ETC facilities look generally similar to non-ETC facilities, with the possible exception that ETC facilities are in counties with a higher proportion of Black patients (18% vs. 15%). Being a Black patient has been shown to be associated with a lower likelihood of getting a kidney transplant.
More granular data (eg, at the postal ZIP code level) may show greater differences than county and/or HRR-aggregated data.
From a certain point of view, it looks as if CMS’ random selection methodology for ETC mandatory participation produced a good balance between the treatment group and the control group for this demonstration model. General facility characteristics and the socioeconomic backdrops for ETC participants are similar to the facilities in the comparison group.
However, recent historical performance on a range of home dialysis and transplantation measures shows ETC facilities have under-performed those in the comparison group. Reasons for this are not clear, but by failing to account for these measures as covariates in the random selection process, CMS appears to have introduced a disadvantage right from the start of the demonstration for ETC facilities as they struggle to meet or exceed the historical benchmarks established from the comparison group of non-ETC facilities.
- Reddy Y, et al. JASN.2020;doi.org/10.1681/ASN.2020101466.
- Wesselman H, et al. CJASN.2021;doi:10.2215/CJN.04860420.
- For more information:
- J. Mark Stephens is the founder of Prima Health Analytics, based in Weymouth, Massachusetts, and has worked in chronic disease research since 2009. He can be reached at email@example.com.