Abstract
Aim: To assess agreement of comorbidity recording between the ANZDATA Registry and the NSW Admitted Patient Data Collection (APDC).
Background: Coding quality of administrative data has been thought to be poor for comorbidity recording. Co‐morbidities recorded by ANZDATA are thought to be reliable but it is time consuming to audit registries.
Methods: The ANZDATA Registry identified all incident patients in NSW receiving renal replacement therapy between 01/07/2000 and 31/07/2010. These patients were linked to NSW APDC which records up to 55 medical conditions impacting upon a hospital admission. We calculated agreement between the two datasets for patients with cardiovascular disease, diabetes, chronic lung disease (CLD) and peripheral vascular disease (PVD) and agreement beyond chance using the kappa statistic (κ). A kappa of >0.75 indicates excellent agreement, 0.40–0.75 is fair to good agreement and <0.40 is poor agreement.
Results: ANZDATA identified 11,036 patents, 531 did not match within NSW APDC or had missing data, leaving 10,505 patients with 2,403,678 hospitalisations including 2,282,850 dialysis or daystay admissions. ANZDATA recorded diabetes in 3199 (30.5%) while APDC recorded diabetes in 1846 (17.6%; κ = 0.53) patients. Cardiovascular disease was recorded in 3756 (35.8%) patients in ANZDATA and in 1457 patients in APDC (13.9%; κ = 0.22). ANZDATA recorded PVD in 2500 (23.8%) patients while APDC recorded PVD in 4 patients (<0.01%; κ = 0.002). CLD was recorded in 1564 (14.9%) patients in ANZDATA and in 239 patients in APDC (0.02%; κ = 0.13).
Conclusions: Poor agreement in co‐morbidity recording exists between ANZDATA and the APDC, with APDC recording far fewer co‐morbidities. These findings have implications for future health care funding models and the ability to use administrative data in clinical risk adjustment tools.
Background: Coding quality of administrative data has been thought to be poor for comorbidity recording. Co‐morbidities recorded by ANZDATA are thought to be reliable but it is time consuming to audit registries.
Methods: The ANZDATA Registry identified all incident patients in NSW receiving renal replacement therapy between 01/07/2000 and 31/07/2010. These patients were linked to NSW APDC which records up to 55 medical conditions impacting upon a hospital admission. We calculated agreement between the two datasets for patients with cardiovascular disease, diabetes, chronic lung disease (CLD) and peripheral vascular disease (PVD) and agreement beyond chance using the kappa statistic (κ). A kappa of >0.75 indicates excellent agreement, 0.40–0.75 is fair to good agreement and <0.40 is poor agreement.
Results: ANZDATA identified 11,036 patents, 531 did not match within NSW APDC or had missing data, leaving 10,505 patients with 2,403,678 hospitalisations including 2,282,850 dialysis or daystay admissions. ANZDATA recorded diabetes in 3199 (30.5%) while APDC recorded diabetes in 1846 (17.6%; κ = 0.53) patients. Cardiovascular disease was recorded in 3756 (35.8%) patients in ANZDATA and in 1457 patients in APDC (13.9%; κ = 0.22). ANZDATA recorded PVD in 2500 (23.8%) patients while APDC recorded PVD in 4 patients (<0.01%; κ = 0.002). CLD was recorded in 1564 (14.9%) patients in ANZDATA and in 239 patients in APDC (0.02%; κ = 0.13).
Conclusions: Poor agreement in co‐morbidity recording exists between ANZDATA and the APDC, with APDC recording far fewer co‐morbidities. These findings have implications for future health care funding models and the ability to use administrative data in clinical risk adjustment tools.
| Original language | English |
|---|---|
| Article number | 023 |
| Pages (from-to) | 22-22 |
| Number of pages | 1 |
| Journal | Nephrology |
| Volume | 19 |
| Issue number | Suppl. 4 |
| DOIs | |
| Publication status | Published - 20 Jul 2014 |
