Patterns of peritoneal dialysis catheter practices and technique failure in peritoneal dialysis: A nationwide cohort study

Abstract

Introduction: Our objective was to assess whether clusters of centers with similar peritoneal dialysis (PD) catheter related practices were associated with differences in the risk of technique failure.

Methods: Patients on incident PD in French centers contributing to the French Language PD Registry from 2012 to 2016 were included in a retrospective analysis of prospectively collected data. Centers with similar catheter cares practices were gathered in clusters in a hierarchical analysis. Clusters of centers associated with technique failure were evaluated using Cox and Fine and Gray models. A mixed effect Cox model was used to assess the influence of a center effect, as explained by the clusters.

Results: Data from 2727 catheters placed in 64 centers in France were analyzed. Five clusters of centers were identified. After adjustment for patient-level characteristics, the fourth cluster was associated with a lower risk of technique failure (cause specific-HR 0.70, 95%CI 0.54-0.90. The variance of the center effect decreased by 5% after adjusting for patient characteristics and by 26% after adjusting for patient characteristics and clusters of centers in the mixed effect Cox model. Favorable outcomes were observed in clusters with a greater proportion of community hospitals, where catheters were placed via open surgery, first dressing done 6 to 15 days after catheter placement, and local prophylactic antibiotics was applied on exit-site.

Conclusion: Several patterns of PD catheter related practices have been identified in France, associated with differences in the risk of technique failure. Combinations of favorable practices are suggested in this study.

Fig 1. Flowchart.

Fig 2. Type of administrative structures.

Fig 3. Dendrogram defining the clusters of centers.

Each PD-center is represented by a black dot at the bottom of the dendrogram, with the number of catheter registered in it. In the virtual space of practices, distances between each center are computed. Centers are joined by nodes placed at a height that is proportional to the distance between the centers. Five clusters of centers are determined to optimize the variability between clusters while minimizing complexity.

Fig 4. Distribution of practices in the five clusters of centers.

The distribution of the modalities used for each practices is represented in the five clusters of centers. Statistically significant differences in the distribution of modalities with cluster 1 (taken as the reference) are marked with a red star.

Fig 5. Survival Kaplan-Meier curves according to the cluster of centers. 5A. Technique failure. 5B. Peritonitis.

Fig 6. Association between covariates and technique failure. Results of the multivariate Cox model (A) and Fine and Gray model (B).

The class of references are: female for the sex, 18–39 years old for the age, 2–3 for the modified Charlson comorbidity index, and cluster number 1 for the clusters of centers. CCI: modified Charlson Comorbidity Index, cs-HR: cause specific hazard-ratio, 95%CI: 95% confidence interval, sd-HR: subdistribution specific hazard-ratio.