Incremental Dialysis Decision Model Based on Expert-Guided Machine Learning

Machine Learning Based on Expert Knowledge to Build and Validate a Decision Model for Incremental Dialysis

This observational prospective study combined clinical expert knowledge with machine learning to develop and validate a predictive model for incremental hemodialysis decision-making. The aim of the predictive model is to assist clinicians in developing individualized incremental dialysis treatment plans to optimize patient outcomes.

By collecting patients' clinical and biochemical parameters and combining them with experts' judgments of dialysis timing and frequency, the model can dynamically assess patients' risk of needing to increase the frequency of dialysis, thus assisting physicians in formulating individualized incremental dialysis regimens to optimize dialysis outcomes and improve patients' prognosis.

Inclusion Criteria:

1. New hemodialysis patients (Apr 2010-Jun 2024), started within 3 months, including transfers.
2. Age ≥18, stable hemodialysis >6 months.

Exclusion Criteria:

1. Incomplete/unreliable data.
2. Twice-weekly palliative dialysis.
3. No baseline urine output or ≤200 mL/24h.
4. Liver disease, heart failure, or severe comorbidities.