Peritoneal Dialysis–Associated Peritonitis Trends Using Medicare Claims Data, 2013-2017

Abstract

The occurrence and consequences of peritoneal dialysis (PD)-associated peritonitis limit its use in populations with kidney failure. Studies of large clinical populations may enhance our understanding of peritonitis. To facilitate these studies we developed an approach to measuring peritonitis rates using Medicare claims data to characterize peritonitis trends and identify its clinical risk factors. Retrospective cohort study of PD-associated peritonitis. US Renal Data System standard analysis files were used for claims, eligibility, modality, and demographic information. The sample consisted of patients receiving PD treated at some time between 2013 and 2017 who were covered by Medicare fee-for-service (FFS) insurance with paid claims for dialysis or hospital services. Peritonitis risk was characterized by year, age, sex, race, ethnicity, vintage of kidney replacement therapy, cause of kidney failure, and prior peritonitis episodes. The major outcome was peritonitis, identified using ICD-9 and ICD-10 diagnosis codes. Closely spaced peritonitis claims (30 days) were aggregated into 1 peritonitis episode. Patient-level risk factors for peritonitis were modeled using Poisson regression. We identified 70,271 peritonitis episodes from 396,289 peritonitis claims. Although various codes were used to record an episode of peritonitis, none was used predominantly. Peritonitis episodes were often identified by multiple aggregated claims, with the mean and median claims per episode being 5.6 and 2, respectively. We found 40% of episodes were exclusively outpatient, 9% exclusively inpatient, and 16% were exclusively based on codes that do not clearly distinguish peritonitis from catheter infections/inflammation ("catheter codes"). The overall peritonitis rate was 0.54 episodes per patient-year (EPPY). The rate was 0.45 EPPY after excluding catheter codes and 0.35 EPPY when limited to episodes that only included claims from nephrologists or dialysis providers. The peritonitis rate declined by 5%/year and varied by patient factors including age (lower rates at higher ages), race (Black > White>Asian), and prior peritonitis episodes (higher rate with each prior episode). Coding heterogeneity indicates a lack of standardization. Episodes based exclusively on catheter codes could represent false positives. Peritonitis episodes were not validated against symptoms or microbiologic data. PD-associated peritonitis rates decline over time and were lower among older patients. A claims-based approach offers a promising framework for the study of PD-associated peritonitis.