Overview of the Canadian pediatric end-stage renal disease database

The Canadian Organ Replacement Register (CORR) is the sole national organ failure registry in Canada and is maintained by the Canadian Institute for Health Information [1]. All pediatric and adult dialysis and transplant centers in Canada submit data regularly to CORR using standardized paper forms. CORR data elements include demographics (age, date of birth, gender, personal health information number, ethnicity, postal code of residence), ESRD treatment modalities, change in treatment modalities with dates, treatment facility information, renal transplant details, selected comorbidities and outcomes (death, allograft loss). Height and weight at time of initiation of renal replacement are also recorded in the registry. Individual patient data is updated every 6 months.

The registry data have been used extensively in multiple clinical studies in patients with kidney failure [3, 19‐25]. A data quality study on the Canadian Organ Replacement Register was published in July 2009 [26]. This study involved a chart review for a sample of 1351 patient charts to assess documentation and coding practices at dialysis clinics, and the participation rate in CORR. In addition, CORR data were compared with data available in the national hospitalization database Canadian Institute for Health Information Discharge Abstract Database (CIHI DAD). It was determined that CORR captures 98.5% of renal transplants performed in Canada. Patient demographics were found to be coded with high reliability in CORR. In contrast, co-morbidities were recorded in CORR with low to moderate sensitivity, indicating a tendency to under-report co-morbidities to CORR. Nevertheless, co-morbidities were recorded with high specificity. The agreement rate between CORR data and chart review data on the primary renal disease code was 59%, indicating that primary renal disease could be more accurately reported to CORR.

The CIHI DAD is the national Canadian in-patient discharge abstract database containing demographic, clinical and institutional data [1]. Hospital discharge diagnoses are coded using International Statistical Classification of Diseases and Related Health Problems 9^th and 10^th revisions. In-hospital deaths and dates of death are also recorded. Each hospital discharge abstract record contains key demographic variables (date of birth, gender), personal health number assigned by relevant health authority, dates of admission, dates of discharge, up to 16 discharge diagnoses codes (most responsible diagnosis, pre- and post-admission co-morbidities) and up to 10 intervention codes. Special care unit length of stay is also documented. This dataset is available from the Canadian Institute for Health Information (CIHI) starting from fiscal year 1992 and data can be released to investigators with appropriate ethics approval and safeguards for privacy.

Ethical approval was obtained from the University of Calgary to perform data linkages.

The CORR database was used to identify a cohort of incident pediatric ESRD patients in Canada except for the Province of Quebec (0-21 years) for the years 1992-2007. The Province of Quebec has separate privacy legislation and application for data release has been submitted.

Each patient identified in the registry was linked to their relevant hospital admissions in CIHI DAD using deterministic linkage methods. Deterministic linkage is accomplished by comparing identifying data fields data from two data sets leading to a judgment on whether two records refer to the same patient [27]. The data element used in the linkage algorithm for creation of this database consisted of only personal health card number and an exact match of health card number between two datasets was required to match a CORR registrant to corresponding hospitalization records in CIHI DAD [27]. Personal health information numbers are assigned by provincial health authorities and edit checks are performed during CORR data entry process. Personal health card numbers recorded in CORR are validated for length and alpha numeric characters and a check sum is computed for each health card number to detect accidental errors that may occur during data entry. CIHI DAD data has similar checking mechanisms to ensure the accuracy of health card numbers.

Linkage was performed by personnel within CIHI to comply with privacy legislation. After linkage, each patient was assigned a unique identification number and the dataset was stripped of all personal identifying information. The merged database (Canadian Pediatric End-Stage Renal Disease Database)therefore comprises linked registry information and hospitalization records as shown in Figure 1. Selected variables and outcomes available in the database are shown in tables 1 and 2. The database is arranged longitudinally in time and includes key clinical information (index renal replacement therapy date, modality and changes in modalities of renal replacement, date of first transplant, clinical details of transplant, facilities of ESRD care, renal allograft loss dates and cause of renal allograft loss, hospitalizations, death date, cause of death).

The merged research database was supplemented with the addition of two key variables: a measure of socioeconomic status and geography for each patient. The six digit postal code for each study subject was obtained from CORR and linked to data from the 2006 Canadian Census using the Postal Code Conversion File (PCCF) to determine median neighborhood income quintile as one measure of socioeconomic status for each patient. PCCF changes with each census and socioeconomic variables are updated with time. The latest available PCCF was from year 2006. Residence location was defined as the distance between the residence of each patient and the nearest ESRD care centres. For these distance calculations, a comprehensive dictionary of postal codes for all Canadian ESRD care centres including satellite dialysis centres was created. Then, geographic coordinates for the center of the polygon defined by each 6 digit postal code (of patient residence and ESRD care centres) was determined using the PCCF and Canadian Postal Data Software (MelissaData, CA). These coordinates were then entered into ArcGIS 9 software (ESRI, CA) to determine the shortest distance by road between patient residence and nearest ESRD care centers. Using average posted speed, estimated travel times for individual patients to their nearest dialysis or transplant facilities were also calculated. Furthermore, we classified each patient residence location as urban or rural also using the PCCF available through Statistics Canada [28]. Investigators on this team have prior experience using these measures of socioeconomic status and residence location in studies involving the adult chronic kidney disease population [20, 22, 24, 25].