Predictive models on patients' eligibility for peritoneal dialysis.

Area of Science:

• Nephrology
• Renal Replacement Therapy

Background:

• Peritoneal dialysis (PD) offers cost-effectiveness and comparable outcomes to hemodialysis (HD).
• PD eligibility assessment is subjective and varies significantly across renal programs.
• Accurate PD eligibility determination is crucial for patient management.

Purpose of the Study:

• To develop a predictive model for PD eligibility in patients initiating HD.
• To identify key predictors influencing PD eligibility.
• To assess the transition rate from HD to PD among eligible patients.

Main Methods:

• Retrospective cohort study of 598 patients initiating HD in Alberta, Canada (2016-2018).
• Logistic regression modeling incorporating 27 predictors (demographics, labs, comorbidities).
• Model selection via Akaike information criterion; performance evaluated using confusion matrix and ROC curve.

Main Results:

• 65.4% (391/598) of patients were eligible for PD.
• The predictive model demonstrated high sensitivity (91.3%) and moderate accuracy (0.68).
• Factors reducing PD eligibility included older age, lower BMI, ICU initiation, and polycystic kidney disease.

Conclusions:

• A majority of patients starting HD are eligible for PD.
• The developed predictive model shows high sensitivity for screening PD candidates post-HD initiation.
• The model can assist in identifying suitable patients for PD therapy.