Abstract: FR-PO812

Use of Poisson Regression Analysis with Restricted Cubic Splines Facilitates Cross-Sectional Comparisons of Mortality in a Large Provider

Session Information

Category: Dialysis

  • 601 Standard Hemodialysis for ESRD

Authors

  • Harford, Antonia, UNM, Albuquerque, New Mexico, United States
  • Schrader, R., DCI, Albuquerque, New Mexico, United States
  • Paine, S., DCI, Albuquerque, New Mexico, United States
  • Gul, Ambreen, DCI, Albuquerque, New Mexico, United States
  • Zager, Philip, UNM, Albuquerque, New Mexico, United States
Background

Dialysis Clinic, Inc. (DCI) is a large not-for-profit provider. Assessing mortality across geographic regions served by a large provider is complicated by differences in demographics and local variations in overall healthcare. The use of Poisson regression analysis with restricted cubic splines may facilitate cross-sectional comparisons of mortality that are not influenced by differences in distributions of race, sex, age, vintage and diabetes.

Methods

We assessed mortality for HD and PD patients treated in DCI facilities in different geographic regions across the country for the years 2009 to 2016 combined. We used ESRD Networks to define the geographic regions. We conducted Poisson regression analysis using the rms package in R. Instead of using USRDS categories we used restricted cubic splines, which have flexible shapes, determined by the data. Models contained categorical (race, sex, and diabetes) and continuous (age, vintage, and year) variables. We fit all 2-way interactions among race, sex, diabetes status, and interactions between age and vintage. Network 8, which has the largest number of DCI patients, was the referent.

Results

Standardized mortality ratios by ESRD Networks for HD and PD patients for the years 2009-2016 are shown. Results are expressed as ratios of mortality within a given Network to that in the referent Network. Among HD patients, clinics in Networks 1, 2, 3, 15, and 17 had lower mortality ratios vs. Network 8. Among PD patients, clinics in Networks 2 and 3 had lower mortality ratios vs. Network 8.

Conclusion

Within DCI there were significant variations in mortality across ESRD Networks for both HD and PD patients. The use of Poisson regression analysis, with restricted cubic splines, facilitates assessing mortality across different geographic regions served by a large provider. This technique provides estimates that are not influenced by demographic differences.