Original articlesA Comparison of the Charlson Comorbidity Index Derived from Medical Record Data and Administrative Billing Data
Introduction
Risk adjustment in clinical and health services research studies has been advocated for several reasons 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13. For example, the effects of confounding factors in nonrandomized studies can be removed or reduced by risk adjusting the outcome measures. Failing to adjust for patient risk factors, including comorbidity, may call into question the scientific validity of the entire study and subsequent results.
Before development of the risk adjustment measure known as the Charlson comorbidity index, the only available method for classifying comorbidity was one developed by consensual criteria. The Charlson comorbidity index was originally developed using data abstracted from the medical record to provide a prognostic taxonomy for comorbid conditions that singly or in combination might alter the risk of short-term mortality for patients enrolled in longitudinal studies [14].
Charlson and colleagues [14] investigated the relationship between potential prognostically important variables and survival using Cox’s regression for life-table data. They calculated unadjusted and adjusted relative risks for each comorbid condition (controlling for all coexistent comorbid conditions, illness severity, and reason for admission) and used the adjusted relative risks as the basis for assigning weights to the comorbid conditions. The weighted index was tested for its ability to predict mortality in a cohort of women with histologically proved primary cancer of the breast. With each increased level of the comorbidity index, there was a stepwise increase in the cumulative mortality attributable to comorbid disease [14].
After the development of the Charlson comorbidity index in 1987, investigators attempted to adapt the index for use with administrative data based on recorded ICD-9-CM diagnoses. Although several ICD-9-CM adaptations of the Charlson index exist, the two most common versions are those developed by Deyo et al. [15] and the Dartmouth-Manitoba group [10]. These two adaptations differ slightly in their selection of ICD-9-CM codes chosen to represent the various comorbidities. When the two versions were compared through application to data from coronary artery bypass surgery, identical Charlson comorbidity scores were assigned to 90% of the cases, and kappa values for specific comorbidities were generally high [16]. In this study, we constructed an ICD-9 index based upon the codes selected by Deyo et al.
Using data from a large study that examined the practice patterns for carotid endarterectomy among Georgia Medicare beneficiaries, we evaluated the ability of the Charlson comorbidity index to predict each of several outcomes. We compared the Charlson comorbidity index derived from medical record data (Chart Index) with the same index derived from the ICD-9-CM codes contained in the Medicare administrative data (ICD-9 Index) to determine how well each predicted inpatient mortality, 30-day mortality, complications of stroke or myocardial infarction, and length of stay. There have been few studies designed to compare the ability of clinical and administrative databases to predict patient outcomes. Hannan et al. [17] compared the ability of a clinical database, the Cardiac Surgery Reporting System, and an administrative database, the Statewide Planning and Research Cooperative System, to predict in-hospital mortality and assess hospital performance for coronary artery bypass graft surgery (CABG) patients. The clinical database was substantially better at predicting case-specific mortality.
Section snippets
Data Collection
These data were collected as part of a project initiated by the Georgia Medical Care Foundation (GMCF) under the auspices of the Health Care Financing Administration’s (HCFA’s) Health Care Quality Improvement Program. The project was designed to assess the appropriateness and outcomes of carotid endarterectomy among Medicare beneficiaries in the state of Georgia. The GMCF, the Medicare peer review organization for the state of Georgia, collaborated with Case Mix Research at Queens University in
Results
Slightly more than half the Medicare beneficiaries undergoing carotid endarterectomy were male (53.3%). Most of the patients in this study were white (90.8%), with 4.7% patients identified as black, 2.0% other, and 2.5% unknown. Their mean age was 72.32 ± 6.80.
Sixty-eight Georgia hospitals performed a carotid endarterectomy and contributed data during the study period. The number of carotid endarterectomy procedures performed at each hospital during the study year varied from 1 to 223.
The mean
Conclusions
As more studies are conducted using administrative data, the need to have a valid and reliable measure derived from this data source for use as a risk adjuster becomes increasingly important. The advantages of using a risk adjuster based on administrative data include lower cost and reduced data collection time required. These advantages must be weighed against the possible loss in predictive ability and/or credibility associated with the use of administrative data for risk adjustment. In this
Acknowledgements
The analyses upon which this publication is based were performed under Contract Number 500-93-0704, entitled, “Operation of Utilization and Quality Control Peer Review Organization (PRO) for the State of Georgia,” sponsored by the Health Care Financing Adminsitration (HCFA), Department of Health and Human Services. The conclusions and opinions expressed, and methods used herein are those of the authors. They do not necessarily reflect HCFA policy. The authors assume full responsibility for the
References (23)
- et al.
Validation of a combined comorbidity index
J Clin Epidemiol
(1994) The pre-therapeutic classification of co-morbidity in chronic disease
J Chronic Dis
(1970)- et al.
The value of measuring severity of disease in clinical research of acutely ill patients
J Chronic Dis
(1984) - et al.
Adapting a clinical comorbidity index for use with ICD-9-CM administrative dataDiffering perspectives
Clin Epidemiol
(1993) - et al.
A new method of classifying prognostic comorbidity in longitudinal studiesDevelopment and validation
J Chronic Dis
(1987) - et al.
Adapting a clinical comorbidity index for use with ICD-9-CM administrative data bases
J Clin Epidemiol
(1992) - et al.
Searching for an improved clinical comorbidity index for use with ICD-9-CM administrative data
J Clin Epidemiol
(1996) Risk adjusting health care outcomesA methodologic review
Med Care Rev
(1986)- et al.
Flaws in mortality dataThe hazards of ignoring comorbid disease
JAMA
(1988) Risk adjustment for medical effectiveness researchAn overview of conceptual and methodological considerations
J Invest Med
(1995)
Interpreting hospital mortality dataThe role of clinical risk adjustment
JAMA
Cited by (188)
Prognosis of Myocarditis Developing After mRNA COVID-19 Vaccination Compared With Viral Myocarditis
2022, Journal of the American College of CardiologyPancreatic ductal adenocarcinoma and invasive intraductal papillary mucinous tumor: Different prognostic factors for different overall survival
2022, Digestive and Liver DiseaseCitation Excerpt :Comorbidities are important in cancer and the physical status of patients is evaluated before planning tailored therapy. CACI [19,20,43] has been shown to be a predictor of survival in patients with digestive system cancers who have undergone resection [44]. Poor preoperative functional reserve, as certified by high values of both ASA [24] and CACI [45], has also been shown to reduce survival in patients with PDAC [24,34,45–47].
Association Between Social and Economic Needs With Future Healthcare Utilization
2020, American Journal of Preventive MedicineNephrotic Syndrome is Associated with Increased Risk of Ischemic Stroke
2019, Journal of Stroke and Cerebrovascular DiseasesCitation Excerpt :CCI is the most extensively studied comorbidity index for predicting mortality.17,18 Predictive validity of CCI has been confirmed by reports based on various outcome criteria, such as mortality, disability, readmissions, and length of stay.18-22 High CCI score has been used to predict poor outcome and death in patients with spontaneous intracerebral hemorrhage and ischemic stroke.23
Individual Patient Comorbidities and Effect on Cochlear Implant Performance
2024, Otology and Neurotology