Factors Associated with Elective Withdrawal of Maintenance Hemodialysis: A Case-Control Analysis

Background: Associations of demographic factors with elective dialysis withdrawal and setting of death, patterns of illness trajectories preceding death, and how illness trajectories, particularly worsening putative disability, are associated with elective withdrawal are poorly understood. Methods: Using United States Renal Data System data, we performed a case-control analysis of hemodialysis patients who died in 2010-2015. A disability proxy score characterized disability; logistic regression identified characteristics associated with death from withdrawal and with death setting; and group-based trajectory models characterized the trajectory of disability in the months preceding death. Results: We identified 14,571 (9.2%) patients who withdrew and 144,305 (90.8%) who died of a non-withdrawal cause. Women were more likely than men to withdraw (OR 1.19, 95% CI 1.15-1.24). The most rural patients were more likely to withdraw than the most urban (OR 1.37, 95% CI 1.25-1.50). Medicaid coverage (a marker for impoverishment) was associated with less withdrawal (OR 0.90, 95% CI 0.86-0.94). Disability proxy score was strongly related to withdrawal: the OR for patients in the highest score category was 31.16 (95% CI 28.40-34.20) versus those with a score of 0. Women and whites (vs. blacks) were overrepresented in the worst, versus better, proxy disability score trajectory. In-hospital death and death in the intensive care unit were more common in women and minorities than in men and whites, but less common in the most rural patients. Conclusions: Important differences separate patients who electively withdraw from those who die of non-withdrawal causes. Worsening disability, in particular, may be a marker for withdrawal.