This is a single-center prospective cohort study that included 175 patients with end-stage renal disease (ESKD) who received maintenance hemodialysis at the hemodialysis center of Huashan Hospital from April 2010 to June 2024. The ESKD patient population was comprised of 175 cases in total. All patients retained some residual kidney function (RKF), and their dialysis records and regular laboratory test results were integrated as input features for the machine learning model. The primary objective of the model was twofold: first, to integrate expert knowledge with machine learning to predict when a switch from lower frequency incremental dialysis (I-HD) to higher frequency dialysis should be made; and second, to identify key variables affecting the risk of adverse outcomes over a two-year period.

An incremental dialysis event is defined as an increase in a patient's dialysis frequency (e.g., from 1 session per week to 2 sessions per week, or from 2 to 3 sessions per week, etc.) due to clinical considerations such as decreased residual renal function, fluid overload, or other physician-determined criteria. At each monthly visit (up to 5 years from enrollment), investigators will record whether each participant experiences an incremental event. We will quantify the primary outcome as the number and proportion of participants who transition to a higher dialysis frequency per month, as well as the cumulative incidence over time.