ASN's Mission

ASN leads the fight to prevent, treat, and cure kidney diseases throughout the world by educating health professionals and scientists, advancing research and innovation, communicating new knowledge, and advocating for the highest quality care for patients.

learn more

Contact ASN

1401 H St, NW, Ste 900, Washington, DC 20005

email@asn-online.org

202-640-4660

The Latest on Twitter

Kidney Week

Abstract: TH-PO789

Identifying Factors Associated with Mortality in Pediatric Hemodialysis Patients Using a Machine Learning Approach

Session Information

  • Pediatric CKD
    November 07, 2019 | Location: Exhibit Hall, Walter E. Washington Convention Center
    Abstract Time: 10:00 AM - 12:00 PM

Category: Pediatric Nephrology

  • 1700 Pediatric Nephrology

Authors

  • Gotta, Verena, University of Basel Children's Hospital, Basel, Switzerland
  • Marsenic Couloures, Olivera, Yale University School of Medicine, New Haven, Connecticut, United States
  • Tancev, Georgi, University of Basel Children's Hospital, Basel, Switzerland
  • Vogt, Julia, ETH Zurich, Zurich, Switzerland
  • Pfister, Marc, University of Basel Children's Hospital, Basel, Switzerland
Background

Mortality in pediatric end-stage renal disease patients is ≥30 times higher than in healthy children, and higher on chronic dialysis than after kidney transplantation. We aimed to explore factors associated with mortality on chronic hemodialysis (HD) in patients having started HD at pediatric age.

Methods

Data used originate from a cohort of patients <30 years on chronic HD since childhood, having received thrice-weekly HD between 2004 and 2016 in outpatient DaVita dialysis centres. Patients with 5-year follow-up since initiation of HD, or death within 5 years, were included. 106 variables (“features”) relating to demographics, HD treatment and laboratory measurements were considered as predictors for 5-year mortality using a machine learning approach (random forest). Among correlated features (ρ>0.7) only the variable with higher clinical significance was retained. Accuracy was evaluated by 30 bootstraps.

Results

363 patients were included in the analysis (n=84 <12 years and n=279 of 12-19 years at initiation of HD). Albumin and lactate-dehydrogenase were retained as the two most important features of 5-year mortality, other features retained in the final model included: lymphocyte count, red blood cell distribution width, red blood cell count, hemoglobin, z-score weight for age, post-HD systolic blood pressure, albumin/globulin ratio, ultra-filtration rate, creatinine and total spKt/V. Mortality was predicted with an accuracy of 81% (standard deviation: ±5%).

Conclusion

Mortality in paediatric patients on chronic HD is associated with multifactorial unspecific markers of nutrition, inflammation, anemia, cardio-vascular risk and dialysis dose. This highlights importance of multimodal intervention strategies besides adequate HD treatment. The association with lactate-dehydrogenase was not expected, but may indicate relevance of blood-membrane interactions, organ-malperfusion or metabolic changes during chronic HD treatment.