Data quality of the A ustralia and N ew Z ealand D ialysis and T ransplant R egistry: A pilot audit

Aim Most clinical registries in A ustralia, including the A ustralia and N ew Z ealand D ialysis and T ransplant R egistry ( ANZDATA ), do not audit submitted data. Inaccurate data can bias registry analysis. This study aimed to audit data submitted to ANZDATA from a single region. Methods A retrospective audit of individual haemodialysis patient data recorded by ANZDATA at 31 D ecember 2009 was completed by nephrologists in a blinded fashion. Original data were recorded by nursing staff. Patients received treatment at a public hospital, two affiliated satellite haemodialysis units, and three private haemodialysis units. Results Fifty‐one audits were completed of a total 175 patients (29.1%) undertaking haemodialysis in 2009. Primary renal disease was correct in 86.3% (95% CI : 74.3–93.2), although errors in type of glomerulonephritis were common. Date of first dialysis (± 1‐month error margin) was correct for 93.6%. Creatinine at first dialysis (± 10% error margin) was correct in 74.4%. Baseline comorbidity accuracy included: peripheral vascular disease (sensitivity 36.4% (95% CI : 24.6–50.1), specificity 82.8% (95% CI : 70.2–90.7)), ischaemic heart disease (sensitivity 69.2% (95% CI : 55.6–80.2), specificity 88.0% (95% CI : 76.3–94.3)), chronic lung disease (sensitivity 25.0% (95% CI : 15.2–38.3), specificity 93.6% (95% CI : 83.4–97.7)), diabetes (sensitivity 86.4% (95% CI : 74.4–93.2), specificity 96.6% (95% CI : 87.5–99.1)), cerebrovascular disease (sensitivity 75.0% (95% CI : 61.7–84.8), specificity 95.3% (95% CI : 85.8–98.6)), and ever smoked (sensitivity 83.3% (95% CI : 70.3–91.4), specificity 71.4% (95% CI : 57.3–82.3)). Non‐melanoma skin cancer was under‐reported and inaccurate. Conclusion Data accuracy was favourable compared with other renal registry validation studies. Data accuracy may be improved by education and training of collectors. A larger audit is necessary to validate ANZDATA .