Dr Neeru Agarwal1,2, Dr JESSICA SUN2, Dr JAMES SAUNDERS3, Dr KRISTY ROBLEDO1, Dr BRENDAN SMYTH1,4, A/Prof ANGUS RITCHIE2, Prof MEG JARDINE1,2
1NHMRC Clinical Trials Centre, Faculty of Medicine & Health, University Of Sydney, Camperdown, Australia, 2Concord Repatriation General Hospital, Concord, Australia, 3Royal Prince Alfred Hospital, Camperdown, Australia, 4St George Hospital, Kogarah, Australia
Biography:
Dr Neeru Agarwal is a PhD Candidate at the NHMRC Clinical Trials Centre, University of Sydney. She is a Nephrologist with an interest in researching the use of routinely collected electronic healthcare data to understand kidney disease. Neeru completed her medical school at the University of New South Wales. She completed her advanced training in Nephrology at Royal Prince Alfred Hospital and Concord Hospital and was a Clinical Fellow at the Imperial College of London, Hammersmith Hospital.
Aim:
To evaluate concordance of clinical outcome reporting in ANZDATA Registry and hospital administrative sources compared with audited chart review, among in-centre haemodialysis patients.
Background:
Systematic collections of clinical events through methods such as patient interviews, chart reviews, registries, or hospital administrative sources produce some variation in accuracy, although this variation may be comparable to that between clinicians making individual diagnoses.
Methods:
We conducted a retrospective cohort study at Concord Repatriation Hospital of participants registered in the RESOLVE quality assurance study between 25/06/2016 and 31/12/2022, with follow-up to 25/06/2023. We evaluated concordance of registry data and hospital discharge ICD-10-AM codes with an audited chart review encompassing results from all sources, for mortality (registry only), acute myocardial infarction (AMI), heart failure and stroke. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and overall agreement using kappa statistics were calculated according to a 14-day window.
Results:
When comparing registry data to chart review for mortality, sensitivity, specificity, PPV and NPV were 93.9%[88.4-97.4], 100%[97.4-100], 100%[97.1-100], and 94.6%[89.6-97.6], respectively, using a 14-day window with concordance classified as “almost perfect”. Sensitivity and NPV improved without the 14-day window. Concordance between registry data and chart review was “moderate” for AMI and stroke, and “fair” for heart failure. Hospital discharge codes showed low sensitivity and high specificity when limited to the primary diagnosis field, with concordance classified as “fair” for AMI and heart failure, and “substantial” for stroke.
Conclusions:
Registry data captured mortality and non-fatal outcomes well. Non-fatal registry and hospital primary diagnosis outcomes showed high specificity, lower sensitivity and variable concordance. Routinely collected data may be suitable for research and quality assurance studies with thoughtful design of extraction algorithms.
Presentation Slides PDF – Click Here