Abstract: FR-PO488
Developing a Predictive Model for Adverse Outcomes for Patients on Peritoneal Dialysis (PD)
Session Information
- Peritoneal Dialysis: Current Topics
November 04, 2022 | Location: Exhibit Hall, Orange County Convention Center‚ West Building
Abstract Time: 10:00 AM - 12:00 PM
Category: Dialysis
- 702 Dialysis: Home Dialysis and Peritoneal Dialysis
Authors
- Blair, Alex, Icahn School of Medicine at Mount Sinai, New York, New York, United States
- Dai, Yang, Icahn School of Medicine at Mount Sinai, New York, New York, United States
- Wen, Huei Hsun, Icahn School of Medicine at Mount Sinai, New York, New York, United States
- Wu, Eric, Icahn School of Medicine at Mount Sinai, New York, New York, United States
- Nadkarni, Girish N., Icahn School of Medicine at Mount Sinai, New York, New York, United States
- Sharma, Shuchita, Icahn School of Medicine at Mount Sinai, New York, New York, United States
- Uribarri, Jaime, Icahn School of Medicine at Mount Sinai, New York, New York, United States
- Chan, Lili, Icahn School of Medicine at Mount Sinai, New York, New York, United States
Background
Peritonitis, cardiac events, and inadequate dialysis are leading causes of morbidity, mortality, and technique failure in PD. New automated cyclers provide remote data that may improve prediction of events
Methods
This is a retrospective single-center study of PD patients at the Mount Sinai Kidney Center between 2019 and 2021. We included patients >18 years of age using cyclers with data available on Sharesource. The outcome of interest was a composite of time to first peritonitis, death, or technique failure. Models were developed using Cox regression with covariates selected via univariate testing (p-value ≤0.1) or clinical input. Correlated covariates were removed. Model 1 included demographic variables (age, gender, race/ethnicity) and Elixhauser Comorbidity Index (ECI). Model 2 included model 1 + mean and standard deviation (SD) of 30-day Sharesource Data (blood pressure (BP), pulse, total ultrafiltration (UF), and night fill volume). Model 3 was similar to model 2 but included Sharesource data as time-varying covariates. To evaluate each model, the cohort was split into 70/30 testing and training sets and iterated 100 times. Model performance was assessed using area under the receiver operator characteristic curves (AUC) over time
Results
We identified 95 patients with complete remote data, 59 with composite endpoint and 36 controls. Patients with the composite outcome were older (59 vs. 53), more often male (59% vs. 39), and had higher ECI (7.9 vs 5.8). With univariate testing, ECI, BP SD, and total UF SD were significantly associated with events. Model 3 which included the Sharesource data as time-varying covariates had the best performance with an average AUC of 0.724 (Figure 1). In all models, performance was best during the first 6 months of follow-up
Conclusion
Addition of Sharesource data as a time-varying covariate improved model performance for prediction of a composite clinical outcome. This model could be used to identify new start PD patients who are at high risk of adverse outcomes