Volume 200 - Issue 1
How well do NSW hospital data identify cases of heart failure?
Authors: Jane Robertson, Sallie-Anne Pearson and John R Attia
Med J Aust 2014; 200 (1): 25. || doi: 10.5694/mja13.10207
Published online: 20 January 2014
Published online: 20 January 2014
To the Editor: Large administrative datasets are increasingly being used for health research; validation studies are critical to understanding data quality, in particular that coding accurately reflects the clinical condition under study. We compared heart failure coding and the associated comorbidity burden using the New South Wales Admitted Patient Data Collection (APDC) with the clinical data in patient medical records (the “gold standard”). The APDC ...
To the Editor: Large administrative datasets are increasingly being used for health research; validation studies are critical to understanding data quality, in particular that coding accurately reflects the clinical condition under study. We compared heart failure coding and the associated comorbidity burden using the New South Wales Admitted Patient Data Collection (APDC) with the clinical data in patient medical records (the “gold standard”). The APDC contains about 26 million records generated since 2000 of all public and private hospital admissions, with reasons for admission and clinical diagnosis.
We identified a random sample of 581 patients from the APDC (mean age, 78.9 years; 54% female) who were treated in 13 hospitals and who had an index admission for heart failure in 2008. The hospitals were in two NSW Area Health Services, treating around 22% of admissions for heart failure in NSW. The index admission for heart failure required for eligibility was defined as an International Classification of Diseases, 10th revision (ICD-10) hospital separation code of I-50 or I-51 (principal diagnosis position), and no admissions with heart failure as a primary or secondary diagnosis in the previous 2 years. There was a median length of stay of 6 days (interquartile range, 3–10) for the index admission, and patient and clinical characteristics were similar to those in our previous NSW study of 29 000 heart failure patients (all eligible heart failure separations in 2002–2007).1
We calculated positive predictive values (PPV) for heart failure coded as a principal diagnosis in the APDC from patient medical records (PPV, 92%) and clinical diagnostic measures of heart failure using modified Framingham criteria2 (PPV, 86%), and Boston criteria scores > 43 (PPV, 85%) (Box 1). These PPVs were lower than reported in Western Australia in 2008 (PPV, 99.5%, using medical chart diagnosis and 10-year look-back)3 and in some international studies.2 PPVs were lower when more stringent clinical criteria (Framingham, Boston) were applied. This approach does not provide sensitivity or specificity of the diagnosis.
The burden of comorbidity, summarised by the Charlson Comorbidity Index, was determined from the medical records at the index admission only or for all admissions in the periods 12 and 24 months before the index admission. Estimates of the burden of comorbidity were similar based on the APDC4 and information from patient medical records (Box 2); this contrasts with previous studies showing that administrative data underestimated the burden of comorbidity.2,5 A look-back window of up to 2 years enhanced the detection of comorbidity, with increased Charlson Comorbidity Index scores indicating that comorbidity adjustment relying on a single admission only will be suboptimal.
This method will not identify all patients with heart failure, but provides reassurance that those coded with heart failure are likely to have the disease.
1 Positive predictive values (PPVs) of Admitted Patient Data Collection principal separation code compared with chart review and selected clinical criteria for 581 patients*
2 Comorbidity burden expressed as Charlson Comorbidity Index* scores
Competing interests
No relevant disclosures.
References
- Robertson J, McElduff P, Pearson SA, et al. The health services burden of heart failure: an analysis using linked population health data-sets. BMC Health Serv Res 2012; 12: 103.
- Lee DS, Donovan L, Austin PC, et al. Comparison of coding of heart failure and comorbidities in administrative and clinical data for use in outcomes research. Med Care 2005; 43: 182-188. 0_i1142965
- Teng TH, Finn J, Hung J, et al. A validation study: how effective is the Hospital Morbidity Data as a surveillance tool for heart failure in Western Australia? Aust N Z J Public Health 2008; 32: 405-407. 0_i1142967
- Sundararajan V, Henderson T, Perry C, et al. New ICD-10 version of the Charlson comorbidity index predicted in-hospital mortality. J Clin Epidemiol 2004; 57: 1288-1294. 0_i1142969
- Powell H, Lim LL, Heller RF. Accuracy of administrative data to assess comorbidity in patients with heart disease: an Australian perspective. J Clin Epidemiol 2001; 54: 687-693. 0_i1142973