Data-Driven Evidence-Based Patient-Centered Optimal Initiation Time for Dialysis Treatment

Area of Science:

• Nephrology
• Biomedical Informatics
• Data Science

Background:

• Determining the optimal time to initiate dialysis for end-stage chronic kidney disease (CKD) is challenging due to the lack of national guidelines.
• Current clinical practices for dialysis initiation lack a standardized, data-driven approach, leading to suboptimal patient outcomes.

Purpose of the Study:

• To develop and evaluate a novel decision-making framework to optimize the initiation time for dialysis treatment in end-stage CKD patients.
• To provide a data-driven perspective for personalized dialysis timing, improving mortality, utility rewards, and cost-effectiveness.

Main Methods:

• A framework integrating natural language processing (NLP), machine learning (ML), and stochastic modeling was developed.
• Electronic Health Record (EHR) data was processed through a pipeline for extraction, de-identification, and standardization.
• An informatics toolkit was used for NLP, event mapping, clustering, and ML to predict prognosis and treatment effects.
• A personalized stochastic model was created to determine optimal dialysis initiation timing.

Main Results:

• The proposed optimal initiation-time policies demonstrated a potential 20.0%–54.7% mortality reduction compared to current clinical policy.
• An increase of 6.4%–14.7% in overall utility reward and a reduction of 9.6%–16.7% in overall cost were observed.
• A clinical practice guideline (CPG) was developed, tested in clinics, and showed promising initial results.

Conclusions:

• The novel decision-making framework offers significant improvements in mortality, utility, and cost for dialysis initiation in CKD patients.
• The developed CPG, based on a data-driven approach, has the potential to become a national standard for end-stage renal disease (ESRD) care.
• Further clinical trials are planned to validate the overall effectiveness and impact on patient care.