Identification of Maintenance Dialysis Recipients in Administrative Health Data: A Systematic Review

Administrative health data are increasingly used to identify individuals receiving maintenance dialysis. We systematically reviewed literature on the accuracy of administrative data in recording dialysis treatments and the performance of case definitions for identifying recipients of maintenance dialysis.

Systematic review of studies included in MEDLINE and Embase from inception to June 13, 2025.

Studies of individuals with kidney failure treated with maintenance dialysis.

Studies assessed the accuracy of administrative data in recording dialysis treatments or in identifying recipients of maintenance dialysis.

Two investigators independently screened studies and extracted data.

Narrative synthesis of study characteristics, data sources, case definitions, and reference standards. Performance metrics included sensitivity, positive predictive value, and F1 score (range of 0-1, the higher the better).

Nine studies were included. One study reported high accuracy in documenting initial and 90-day dialysis modalities in claims versus medical records. Eight studies evaluated various case definitions for maintenance dialysis using inpatient data (n = 1), outpatient claims (n = 2), or both (n = 5). Three studies used procedural codes alone, and 5 used procedural and diagnostic codes. Two studies included incident cases only. Reference standards were kidney replacement therapy registries or data from kidney care programs in 7 studies and from medical records in 1. The median positive predictive value was 85.9% (25^th and 75^th percentiles, 74.4%-92.9%; 37 case definitions), median sensitivity was 75.8% (63.5%-83.8%; 18 case definitions), and median F1 score was 79.5% (70.2%-81.9%; 18 case definitions).

Findings were based on studies from high-income countries; lack of studies using contemporary data.

Performance of case definitions for identifying maintenance dialysis recipients in administrative data is context-dependent and varies across definitions. Existing definitions may guide the development of updated, locally adapted methods, ideally validated against medical records as the reference standard for future studies.

Registered at PROSPERO with identification code CRD42024582507.

Administrative health data such as hospital records and billing information are often used to identify individuals receiving long-term dialysis. Different methods have been developed to identify these patients in the data, but it is unclear how accurate these methods are, and a critical assessment of the literature is lacking. We conducted a systematic review to compare the accuracy of these methods. We found that most methods were reasonably accurate, but many were tested using outdated data or were compared against a reference standard, which may not have accurately identified all patients. Our findings inform the development of more reliable methods for accurately identifying dialysis recipients in administrative data and for planning health care services.