Kidney Week

Abstract: FR-PO0040

Markov Multistate Modeling of Longitudinal Kidney Replacement Therapy Patterns in a Canadian Population with Advanced CKD

Session Information

• Artificial Intelligence and Digital Health at the Bedside
  November 07, 2025 | Location: Exhibit Hall, Convention Center
  Abstract Time: 10:00 AM - 12:00 PM

Category: Artificial Intelligence, Digital Health, and Data Science

• 300 Artificial Intelligence, Digital Health, and Data Science

Authors

• Klamrowski, Martin M., The Ottawa Hospital Research Institute, Ottawa, Ontario, Canada

Martin M. Klamrowski, MS, MSc

• Masucci, Lisa, The Ottawa Hospital Research Institute, Ottawa, Ontario, Canada

Lisa Masucci, MSc

• Klein, Ran, The Ottawa Hospital Research Institute, Ottawa, Ontario, Canada

Ran Klein, PhD

• McCudden, Chris, Eastern Ontario Regional Laboratory Association, Ottawa, Ontario, Canada

Chris McCudden, PhD

• Green, James, Carleton University, Ottawa, Ontario, Canada

James Green, PhD

• Amooei, Elmira, The Ottawa Hospital Research Institute, Ottawa, Ontario, Canada

Elmira Amooei

• McArthur, Eric, ICES, Toronto, Ontario, Canada

Eric McArthur

• Garg, Amit X., ICES, Toronto, Ontario, Canada

Amit X. Garg, MD, PhD

• Thavorn, Kednapa, The Ottawa Hospital Research Institute, Ottawa, Ontario, Canada

Kednapa Thavorn, PhD

• Edwards, Cedric A.W., University of Ottawa, Ottawa, Ontario, Canada

Cedric A.W. Edwards, MD

• Rashidi, Babak, University of Ottawa, Ottawa, Ontario, Canada

Babak Rashidi, MD

• Akbari, Ayub, The Ottawa Hospital Research Institute, Ottawa, Ontario, Canada

Ayub Akbari, MD, MS, MSc, FASN

• Hundemer, Gregory L., The Ottawa Hospital Research Institute, Ottawa, Ontario, Canada

Gregory L. Hundemer, MD, MPH

Background

Chronic kidney disease (CKD) is a condition characterized by the gradual loss of kidney function. As computer decision aids and other practice-changing tools become more widespread, estimates of kidney replacement therapy patterns are needed to support cost-effectiveness research and guide tool selection.

Methods

Data were acquired from ICES in Ontario, Canada. Separate multi-state models were developed in patients with, 1) a 2-year kidney failure risk equation (KFRE) prediction ≥ 10%, and 2) eGFR < 30 mL/min/1.72m² with urine albumin-to-creatinine ratio (UACR) ≥ 30 mg/g. Indexing criteria required a second confirmatory test within one year of the first. A model was developed to describe the sequence of possible transitions from CKD to kidney transplantation, peritoneal dialysis, hemodialysis, and death. The state space was further augmented to represent important transition pathways, including patients who initiated dialysis in an unplanned manner (during a hospital admission), and those who underwent pre-emptive kidney transplantation. Individual parametric hazard models were fit to estimate transition-specific hazards. Model fit was compared for both time-homogeneous hazards and semi-Markov hazards using the Akaike Information Criterion (AIC).

Results

33,494 patients with mean (SD) age 69 (15) years, eGFR 23 (7) ml/min per 1.73 m² were included as part of the KFRE cohort, and 49,457 patients with mean (SD) age 72 (13) years, eGFR 24 (5) ml/min per 1.73 m² were included as part of the eGFR + UACR cohort. Patient state was not static, with potentially multiple transitions occurring throughout follow-up. The semi-Markov models had improved AIC values compared to the time-homogeneous models.

Conclusion

These multi-state models reveal the dynamic nature of kidney replacement therapy and may facilitate economic evaluations towards guiding the effective selection of decision-support tools in CKD.

Funding

• Government Support – Non-U.S.

Digital Object Identifier (DOI)

• doi: 10.1681/ASN.202519qp91z4