Original ArticleNew ICD-10 version of the Charlson comorbidity index predicted in-hospital mortality
Introduction
The Charlson comorbidity index [1] has been a useful tool for health researchers in their effort to measure comorbid disease status or casemix in health care databases. Charlson et al. [1] defined numerous clinical conditions through reviewing hospital charts and assessed their relevance in the prediction of 1-year mortality. A weighted score was assigned to each of 17 comorbidities, based on the relative risk of 1-year mortality. As a consequence, the sum of the index score is an indicator of disease burden, and a strong estimator of mortality. Since then, the Charlson Index has been validated in various larger populations [2], [3]. These studies consistently demonstrate that the Charlson Index is a valid prognostic indicator. Coding algorithms in ICD-9CM were later developed for each of the variables in the Charlson Index by other researchers [4], [5], [6].
The first version of the International Classification of Disease was adopted in 1900 to monitor and compare mortality statistics and causes of death. Now under the auspices of the World Health Organization, the classification has been revised periodically to accommodate new knowledge of disease and health [7]. The United States modified the ninth version, ICD-9, by specifying many categories and extending coding rubrics to describe the clinical picture in more detail, known as ICD-9-CM [8].
ICD-10, the newest version of this nosology, has been used by many European countries for coding mortality and/or morbidities since 1994 [9], [10], [11]. Canada, Australia, and the United States enhanced the ICD-10 by adding new codes and have developed their own versions [12], [13], [14]. Advantages of ICD-10 include the fact that its coding structure leaves room for future expansion and allows the coding of richer clinical information [14].
Despite these potential advantages of ICD-10, the use of such data in health services research initiatives has been limited to date, perhaps because of the lack of familiarity and agreement among researchers on coding algorithms for defining clinical conditions in ICD-10, such as the above-mentioned ICD-9-CM coding algorithms that define the 17 comorbidities that constitute the Charlson Index. Clearly, the development and validation of similar algorithms for ICD-10 would represent a contribution to the field of health services research, at a time when much of the world is shifting to ICD-10 coding for hospital discharges.
The State of Victoria, Australia, has maintained administrative data on all hospital admissions for a number of years. Prior to July 1, 1998, all discharges were coded in ICD-9-CM format, but from that date forward, all hospital discharges have been coded in ICD-10-AM, the Australian version of ICD-10. Given the current availability of several years of ICD-10 data, we initiated this methodological study, with specific objectives being (1) to develop an ICD-10 coding algorithm that permits definition—in ICD-10 data—of the 17 variables that constitute the Charlson Index, (2) to validate the algorithm by assessing the prevalence of the resulting comorbidity variables in ICD-10 data relative to prevalence of the same variables in the earlier ICD-9-CM data, and (3) to further validate the algorithm by assessing the association between the resulting comorbidity variables and in-hospital mortality.
Section snippets
Data source for validation
Victoria is Australia's second largest state, with a population of more than 4.5 million [15]. As part of a universal health system, each state maintains administrative data on all hospital admissions. The Victorian Admitted Episodes Dataset (VAED), maintained by the Victorian Department of Human Services, is based upon hospital data compiled by individual public and private hospitals in Victoria [16]. The dataset contains demographic and clinical information on each discharge. The diagnostic
Results
There were more than 400,000 multiple-day hospitalizations (overnight stay or longer) for each year of data used in our analysis (Table 2).
On average, the prevalence of the specific disease categories during multiple-day hospitalizations did not shift dramatically between the fiscal years 1997–98 and 1998–99, when the change from ICD-9-CM to ICD-10-AM occurred in Victoria.
The prevalence of congestive heart failure, peripheral vascular disease, chronic pulmonary disease, connective tissue
Discussion
The ICD-9-CM coding algorithms developed by Deyo et al. [4] and by the Dartmouth–Manitoba groups in the early 1990s were important developments; they provided a methodological foundation for a large number of studies based on administrative data [24], [25], [26], [27], [28], [29], [30], [31], [32].
We have presented the first ICD-10 version of the Charlson comorbidity index to be developed and tested on a large population-based dataset. In comparison with a well-established ICD-9-CM coding
Acknowledgments
H.Q. is a Population Health Investigator of Alberta Heritage Foundation for Medical Research, Alberta, Canada. W.A.G. is holder of a Government of Canada Research Chair in Health Services Research, and a Health Scholar Award from the Alberta Heritage Foundation for Medical Research.
References (32)
- et al.
A new method of classifying prognostic comorbidity in longitudinal studies: development and validation
J Chronic Dis
(1987) - et al.
A comparison of two comorbidity instruments in arthritis
J Clin Epidemiol
(1999) - et al.
Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases
J Clin Epidemiol
(1992) - et al.
Adapting a clinical comorbidity index for use with ICD-9-CM administrative data: differing perspectives
J Clin Epidemiol
(1993) - et al.
Searching for an improved clinical comorbidity index for use with ICD-9-CM administrative data
J Clin Epidemiol
(1996) - et al.
Evaluation of two competing methods for calculating Charlson's comorbidity index when analyzing short-term mortality using administrative data
J Clin Epidemiol
(1997) - et al.
A validation of the functional independence measurement and its performance among rehabilitation inpatients
Arch Phys Med Rehabil
(1993) - et al.
Selecting a patient characteristics index for the prediction of medical outcomes using administrative claims data
J Clin Epidemiol
(1995) - et al.
Ischemic Heart Disease Patient Outcomes Research Team. A comparison of administrative versus clinical data: coronary artery bypass surgery as an example
J Clin Epidemiol
(1994) - et al.
The performance of different lookback periods and sources of information for Charlson comorbidity adjustment in Medicare claims
Med Care
(1999)
Risk adjustment in outcome assessment: the Charlson comorbidity index
Methods Inf Med
International statistical classification of disease and related health problems, Tenth Revision (ICD-10)
Annotated ICD-9-CM international classification of diseases, 9th revision, clinical modification
Implementation of the International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10)
Epidemiol Bull
A survey to identify the clinical coding and classification systems currently in use across Europe
Medinfo
Where in the world is ICD-10?
J AHIMA
Cited by (1734)
Association between depression, antidepressant use, and the incidence of atherosclerotic cardiovascular diseases
2024, Journal of Affective DisordersAssociation of Healthcare Fragmentation and the Survival of Patients with Colorectal Cancer in Colombia
2024, Value in Health Regional IssuesCannabis use disorders and outcome of admission to intensive care: A retrospective multi-centre cohort study
2024, Journal of Critical CareThe impact of the implementation of a national policy on dementia on healthcare costs in older patients with dementia in Korea: a quasi-experimental difference-in-difference study
2024, The Lancet Regional Health - Western Pacific