Kidney Week

Abstract: TH-PO285

Analytical Methods for Assessing Clinic-Level Outcomes

Session Information

• Dialysis: Cost, Socioeconomics, Quality of Life
  October 25, 2018 | Location: Exhibit Hall, San Diego Convention Center
  Abstract Time: 10:00 AM - 12:00 PM

Category: Dialysis

• 701 Dialysis: Hemodialysis and Frequent Dialysis

Authors

• Pankratz, V. Shane, UNM Health Sciences Center, Albuquerque, New Mexico, United States

V. Shane Pankratz,

• Jiang, Huan, Dialysis Clinic, Inc., Albuquerque, New Mexico, United States

Huan Jiang,

• Paine, S., Dialysis Clinic, Inc., Albuquerque, New Mexico, United States

S. Paine,

• Gul, Ambreen, Dialysis Clinic, Inc., Albuquerque, New Mexico, United States

Ambreen Gul,

• Miskulin, Dana, Tufts Medical Center, Boston, Massachusetts, United States

Dana Miskulin,

• Zager, Philip, Dialysis Clinic, Inc., Albuquerque, New Mexico, United States

Philip Zager,

Background

Patient outcomes from dialysis units are important indicators of the quality of care provided by each unit. Clinical outcomes for dialysis providers are typically assessed on a per-clinic level. However, the number of patients in each clinic varies greatly, and this can impact the validity of ratings. Different analytical approaches might enhance the estimated ratings.

Methods

Using 2009-2017 data from Dialysis Clinic, Inc. (DCI) clinics, as well as simulated data, we assessed the impact of different estimation approaches (fixed vs. random clinic effects) on ratings based on clinic-level mortality.

Results

Simulations demonstrate that random effects estimates are shrunk towards the overall mean, and have lower standard errors than fixed effects estimates, and that random effects estimates more reliably provide estimates that agree with truth, particularly for smaller clinics. Estimates of mortality from DCI clinics illustrate the degree of shrinkage that is induced through the application of random effects methods (Figure 1).

Conclusion

Random effects approaches are more likely to provide correct rankings of clinics than fixed effects estimates, and might reliably be used to obtain stabilized estimates of mortality rates. However, obtaining estimates in this way may require larger numbers of patients per unit than are often treated within individual clinics. Enhancing information content, either by aggregating across years or by incorporating other data types into an aggregated rating measure, would enable more accurate estimation of clinic performance.

Figure 1. Distribution of unit-specific mortality rates in 2014

Funding

• Clinical Revenue Support