Original articleAccuracy of administrative data to assess comorbidity in patients with heart disease: an Australian perspective
Introduction
Comorbidity can be defined as other additional conditions that are unrelated to the patient's principal diagnosis (e.g., in a patient admitted to hospital with an acute myocardial infarction, unstable angina would be a complication and diabetes would be a comorbidity) [1]. Comorbidity has been shown to be associated with mortality, length of hospital stay and disability 1, 2, 3. When assessing a person's functional ability it is necessary to take into account their total disease burden [4]. As the aged population increases in our society, the assessment of comorbidity has become progressively important, particularly for health policy makers and administrators in regard to health costs and health planning 4, 5.
In most hospitals, patient comorbidities are recorded as discharge diagnosis codes for administrative purposes. In the Hunter Region of Australia up to 10 ICD-9-CM codes are used. This coding is usually done by coding clerks, based on the patient's medical records.
Increasingly, because of their accessibility and cost-effectiveness, these discharge codes are being used to derive indices of comorbidity (most often the Charlson Index), for research and other health planning purposes 6, 7, 8, 9. However, because these discharge codes were primarily created for administrative purposes, they have been found to be limited in the amount of clinical information they can provide 10, 11. Previous work has questioned the accuracy of using administrative data for the purpose of measuring comorbidity using the Charlson Index 10, 12. Recent studies have found conflicting results when comorbidities based on discharge codes were associated with risk of death. Iezzoni et al. found that diabetes, which would be expected to increase the risk of death, was associated with lower death rates [10]. A plausible explanation is that there is a bias against coding of chronic disease when patients die or a critical illness occurs during the admission 10, 12.
Considering the importance of measuring comorbidity and the popular use of administrative datasets, relatively little research has been done to determine the accuracy of administrative data for this purpose 9, 11, 13, 14, 15. Some studies have reported that administrative data underestimate the presence of comorbid conditions 9, 11, 14, 15. Waite et al. compared three comorbidity indices (including the Charlson) derived from the medical records to two indices derived from the administrative data [13]. Kieszak et al. also made this comparison using the Charlson comorbidity index [15]. Romano et al. compared the prevalence of 14 comorbidities and the relative risk estimates for short-term mortality derived from medical record data to those derived from administrative data of similar subjects [11]. Only two studies have validated administrative data directly by a comparison of discharge codes with comorbidities recorded in the patient's medical record 9, 14. Hawker et al.'s study was restricted to 175 patients admitted to hospital for knee replacement surgery [14]. The data for Malenka et al.'s study were collected from 1974 to 1984 and they recommended further research using more current data [9].
The aim of this study was to determine the accuracy of administrative data (by use of hospital discharge codes) for measuring comorbidity in patients with heart disease.
Section snippets
Data source
The Hunter Region in the state of NSW is situated 150 kilometres north of Sydney and has a total population of 521,785 [16]. The region comprises a major industrial centre, the city of Newcastle. The area is also supported by the rural and coal mining district of the Hunter Valley. There are 23 hospitals ranging from tertiary care hospitals to community nonacute, including private hospitals, in the region.
The subjects in our study included patients between 20 and 85 years of age who had been
Results
Table 1 shows the characteristics of the 1765 study subjects who were admitted to hospital with AMI, unstable angina, angina pectoris, chronic IHD or heart failure and whose records were available for study.
Discussion
The accurate measurement of comorbidity cannot be understated when trying to assess the health burden on society [13].
Previous research comparing administrative data and medical record data found comorbidity to be underreported in the administrative data 9, 11, 14, 15. Hawker et al. found consistently high false-negative rates of recording of comorbidities in the administrative data (mean 63%), low false-positive rates (less than 3%), except for muskuloskeletal conditions (8.4%), and good mean
Conclusion
This study confirms an entrenched underreporting of comorbidity throughout administrative databases based on hospital discharge codes. Further research is needed into why this occurs, and standards in coding practice should be introduced throughout hospitals. Training for medical practitioners in accurate recording of all comorbidity on the discharge abstract as well as complications is required. Administrative databases are a valuable resource for both researchers and health policy makers and,
References (19)
Assessing the impact of comorbidity in the older population
Ann Epidemiol
(1996)- et al.
Comorbidity of chronic disease in general practice
J Clin Epidemiol
(1993) - et al.
A new method of classifying prognostic comorbidity in longitudinal studiesDevelopment and validation
J Chron Dis
(1987) - et al.
Adapting a clinical comorbidity index for use with ICD-9-CM administrative datadiffering perspectives
J Clin Epidemiol
(1993) - et al.
Using administrative data to describe casemixa comparison with the medical record
J Clin Epidemiol
(1994) - et al.
A comparison of administrative versus clinical datacoronary artery bypass surgery as an example
J Clin Epidemiol
(1994) - et al.
Development and validation of a claims based index for adjusting for risk of mortalitythe case of acute myocardial infarction
J Clin Epidmeiol
(1995) - et al.
Lack of association between patients' burden of disease and risk for hospital readmission
J Clin Epidemiol
(1994) - et al.
Accuracy of administrative data for assessing outcomes after knee replacement surgery
J Clin Epidemiol
(1997)