TY - JOUR
T1 - Evaluating data quality in the Australian and New Zealand dialysis and transplant registry using administrative hospital admission datasets and data-linkage
AU - Palamuthusingam, Dharmenaan
AU - Pascoe, Elaine M.
AU - Hawley, Carmel M.
AU - Johnson, David W.
AU - Ratnayake, Gishan
AU - McDonald, Stephen
AU - Boudville, Neil
AU - Jose, Matthew
AU - Fahim, Magid
PY - 2023/9
Y1 - 2023/9
N2 - Background: Clinical quality registries provide rich and useful data for clinical quality monitoring and research purposes but are susceptible to data quality issues that can impact their usage. Objective: This study assessed the concordance between comorbidities recorded in the Australia and New Zealand Dialysis and Transplant (ANZDATA) Registry and those in state-based hospital admission datasets. Method: All patients in New South Wales, South Australia, Tasmania, Victoria and Western Australia recorded in ANZDATA as requiring chronic kidney replacement therapy (KRT) between 01/07/2000 and 31/12/2015 were linked with state-based hospital admission datasets. Coronary artery disease, diabetes mellitus, cerebrovascular disease, chronic lung disease and peripheral vascular disease recorded in ANZDATA at each annual census date were compared overall, over time and between different KRT modalities to comorbidities recorded in hospital admission datasets, as defined by the International Classification of Diseases (ICD-10-AM), using both the kappa statistic and logistic regression analysis. Results: 29, 334 patients with 207,369 hospital admissions were identified. Comparison was made at census date for every patient comparison. Overall agreement was “very good” for diabetes mellitus (92%, k = 0.84) and “poor” to “fair” (21–61%, k = 0.02–0.22) for others. Diabetes mellitus recording had the highest accuracy (sensitivity 93% (±SE 0.2) and specificity 93% (±SE 0.2)), and cerebrovascular disease had the lowest (sensitivity 54% (±SE 0.2) and specificity 21% (±SE 0.3)). The false positive rates for cerebrovascular disease, peripheral vascular disease and chronic airway disease ranged between 18 and 33%. The probability of a false positive was lowest for kidney transplant patients for all comorbidities and highest for patients on haemodialysis. Conclusions and Implications: Agreement between the clinical quality registry and hospital admission datasets was variable, with the prevalence of comorbidities being higher in ANZDATA.
AB - Background: Clinical quality registries provide rich and useful data for clinical quality monitoring and research purposes but are susceptible to data quality issues that can impact their usage. Objective: This study assessed the concordance between comorbidities recorded in the Australia and New Zealand Dialysis and Transplant (ANZDATA) Registry and those in state-based hospital admission datasets. Method: All patients in New South Wales, South Australia, Tasmania, Victoria and Western Australia recorded in ANZDATA as requiring chronic kidney replacement therapy (KRT) between 01/07/2000 and 31/12/2015 were linked with state-based hospital admission datasets. Coronary artery disease, diabetes mellitus, cerebrovascular disease, chronic lung disease and peripheral vascular disease recorded in ANZDATA at each annual census date were compared overall, over time and between different KRT modalities to comorbidities recorded in hospital admission datasets, as defined by the International Classification of Diseases (ICD-10-AM), using both the kappa statistic and logistic regression analysis. Results: 29, 334 patients with 207,369 hospital admissions were identified. Comparison was made at census date for every patient comparison. Overall agreement was “very good” for diabetes mellitus (92%, k = 0.84) and “poor” to “fair” (21–61%, k = 0.02–0.22) for others. Diabetes mellitus recording had the highest accuracy (sensitivity 93% (±SE 0.2) and specificity 93% (±SE 0.2)), and cerebrovascular disease had the lowest (sensitivity 54% (±SE 0.2) and specificity 21% (±SE 0.3)). The false positive rates for cerebrovascular disease, peripheral vascular disease and chronic airway disease ranged between 18 and 33%. The probability of a false positive was lowest for kidney transplant patients for all comorbidities and highest for patients on haemodialysis. Conclusions and Implications: Agreement between the clinical quality registry and hospital admission datasets was variable, with the prevalence of comorbidities being higher in ANZDATA.
KW - administrative datasets
KW - ANZDATA
KW - chronic
KW - clinical quality registries
KW - comorbidities
KW - comorbidity
KW - data accuracy
KW - data-linkage
KW - dialysis
KW - end-stage kidney disease
KW - health care
KW - information storage and retrieval
KW - kidney failure
KW - kidney transplantation
KW - outcome and process assessment
KW - process assessment
KW - quality of health care
KW - registries
KW - Supplementary Keywords
UR - http://www.scopus.com/inward/record.url?scp=85131735116&partnerID=8YFLogxK
U2 - 10.1177/18333583221097724
DO - 10.1177/18333583221097724
M3 - Article
C2 - 35695032
AN - SCOPUS:85131735116
SN - 1833-3583
VL - 52
SP - 212
EP - 220
JO - Health Information Management Journal
JF - Health Information Management Journal
IS - 3
ER -