Premium
Early withdrawal and non‐withdrawal death in the months following hemodialysis initiation: A retrospective cohort analysis
Author(s) -
Wetmore James B.,
Roetker Nicholas S.,
Gilbertson David T.,
Liu Jiang
Publication year - 2019
Publication title -
hemodialysis international
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.658
H-Index - 47
eISSN - 1542-4758
pISSN - 1492-7535
DOI - 10.1111/hdi.12723
Subject(s) - medicine , hemodialysis , retrospective cohort study , cause of death , cohort , proportional hazards model , survival analysis , logistic regression , comorbidity , disease
: Whether and how factors associated with elective hemodialysis withdrawal differ from those associated with non‐withdrawal death soon after maintenance hemodialysis initiation have not been well studied. Methods : A retrospective cohort analysis was performed using USRDS data from 2011 to 2014. Patients were randomly categorized 2:1 into training and validation samples. Elective withdrawal deaths were identified using the Death Notification form. Multinomial logistic regression was used to fit a prediction model for three outcome categories (withdrawal, non‐withdrawal death, survival at 6 months) as a function of demographic, comorbidity, and functional status. Findings : The training sample comprised 80,284 hemodialysis patients. Mean age was 71.7 ± 11.4 years, 44.9% were female, 72.9% were white, and 22.8% were black. Within 6 months, 19.1% died, of whom 2099 (2.6%) withdrew and 13,223 (16.5%) died of a non‐withdrawal cause; 13.7% of all deaths were withdrawals. Baseline characteristics and event rates were similar among the 40,142 patients in the validation sample. The model was calibrated adequately and could discriminate moderately well between withdrawal and survival (area under ROC curve [AUC]: 0.77) and between non‐withdrawal death and survival (AUC: 0.73). However, discrimination between withdrawal and non‐withdrawal death was relatively low (AUC: 0.62). Older age and white, compared with non‐white, race were each associated with greater odds of death, and these associations were stronger for withdrawal than for non‐withdrawal death. Discussion : Advanced age and white, as opposed to black, race were most strongly associated with early elective hemodialysis withdrawal compared with non‐withdrawal death. However, it is difficult to differentiate between patients who will experience early withdrawal vs. non‐withdrawal death, as many factors are similarly associated with both outcomes.