Original Article
Charlson scores based on ICD-10 administrative data were valid in assessing comorbidity in patients undergoing urological cancer surgery

https://doi.org/10.1016/j.jclinepi.2005.07.015Get rights and content

Abstract

Background and Objectives

Adjustment for comorbidity is an essential component of any observational study comparing outcomes. We evaluated the validity of the Charlson comorbidity score based on ICD-10 codes in patients undergoing urological cancer surgery within an English administrative database.

Study Design and Setting

Patients who underwent radical urological cancer surgery between 1998 and 2002 in the English National Health Service were identified from the Hospital Episode Statistics database (N = 20,138). ICD-9-CM codes defining comorbid diseases according to the Deyo and Dartmouth–Manitoba adaptations of the Charlson comorbidity score were translated into ICD-10 codes.

Results

Charlson scores derived by the ICD-10 translation of the Deyo and Dartmouth–Manitoba adaptations were identical in 16,623 patients (83%; κ = .63). For both adaptations, ICD-10 scores increased with age, were higher in patients admitted on an emergency basis, and predicted short-term outcome. Addition of either the ICD-10 Charlson Deyo or Dartmouth–Manitoba score to risk models containing age and sex to predict in-hospital mortality resulted in a better model fit but only in small improvements of the predictive power.

Conclusion

The ICD-10 translations of the Deyo and Dartmouth–Manitoba adaptations performed similarly in risk models predicting hospital mortality following urological cancer surgery. Adjustment for comorbidity over and above age and sex alone does not seem to provide a large improvement.

Introduction

Case-mix adjustment is an essential component of any study that aims to evaluate outcomes of care, in that differences in baseline patient characteristics may account for many of the observed differences [1], [2]. These characteristics include demographic features such as patient age, sex, and socioeconomic status, but also include differences in both the presence and severity of comorbid disease and the severity of the primary disease for which medical and surgical treatment is taking place. Practical constraints, however, often limit the extent of case-mix adjustment within administrative databases [3], [4], [5], [6].

Comorbid disease may be defined as preexisting disease or illness that affects a patient in addition to but not as a result of a primary diagnosis [2]. In an attempt to adjust for the presence or absence of comorbid disease, several comorbidity scoring systems have been previously developed [7]. One of the most widely used and validated systems was developed in 1987 by Charlson and coworkers [8] from risk factors that predicted 1-year survival in a cohort of medical inpatients and then validated in a population of patients undergoing surgery for breast cancer. This scoring system was subsequently adapted for use with administrative databases using diagnostic codes from the International Classification of Diseases, 9th revision, Clinical Modification (ICD-9-CM) by Deyo et al. and then further adapted by a collaboration of workers at Dartmouth and Manitoba Universities [9], [10], [11], [12], [13], [14]. The Charlson Score consists of 19 different disease comorbidity categories, each allocated a weight of 1 to 6 based on the adjusted relative risk of 1-year mortality and summed to provide a total score [8]. A number of previous studies using administrative data to examine outcomes following surgery have derived comorbidity scores not only from records of the hospital admission in which surgery took place but also considering records of admissions preceding the admission for surgery, in an attempt to capture more complete comorbidity data [15], [16], [17].

Administrative databases are increasingly being used to study patient outcomes following medical and surgical care [2], [18], [19]. Advantages of the use of such databases include ready availability, relatively low cost, large population coverage, and ability to assess trends over time. Disadvantages include coding errors or omissions and a lack of detailed clinical information [20], [21].

The Hospital Episode Statistics (HES) database of the Department of Health in England records medical, demographic and administrative data relating to all inpatient admissions to hospitals in England and was established in 1989 [22], [23]. More than 12 million records are collected each year. This database was established to support policy development, to identify variations in healthcare delivery with time and between geographic areas, to be used in medical research, and to assess performance [24]. Until 2002, the HES database contained seven diagnosis fields, using codes defined from the 10th revision of the International Classification of Diseases (ICD-10), and four operative procedure fields, using codes defined from the U.K. Tabular List of the Classification of Surgical Operations and Procedures, version 4, of the U.K. Offices of Populations, Censuses, and Surveys (OPCS-4) [25], [26]. The number of diagnosis and procedure fields increased to 14 and 12, respectively, after 2002.

We translated the Deyo and Dartmouth–Manitoba ICD-9-CM adaptations of the Charlson score for use with ICD-10 administrative databases, such as the English HES database (Fig. 1). For a comorbidity index to be clinically valid, it would be expected that increasing comorbidity scores would be related to risk factors for comorbidity as well as to clinical outcomes. Our objective was therefore to evaluate the validity of the ICD-10 Charlson scores within the English HES database based on these expectations.

We considered patients undergoing radical urological cancer surgery, as part of a program of work investigating the quality of care for these patients. We used data from both the index surgical admission and also from admissions over the year preceding surgery. Although previous groups have adapted the Deyo version of the Charlson score for use with ICD-10 data, to the best of our knowledge, this is the first time that in addition the validity of the Dartmouth–Manitoba version of the Charlson score has been evaluated for ICD-10 administrative data [27], [28]. This has considerable practical significance, given the expected implementation of the ICD-10-CM coding system within the United States [29].

Section snippets

Data

Data were extracted from the HES database for the data years 1998/1999 to 2001/2002 for all patients recorded as having undergone a radical prostatectomy (RP), radical cystectomy (RC), or radical nephrectomy (RN). Patients were included in the study if, first, an ICD-10 code representing the cancer as reason for treatment was present in any of the seven diagnosis fields and, second, an OPCS-4 code representing the corresponding surgical procedure was present in any of the four operative

Results

Demographic characteristics, method of admission, length of hospital stay, in-hospital mortality rates, and the number of days spent in hospital over the year prior to surgery are presented in Table 1 for each of the three cohorts of patients.

Table 2 shows the prevalence of each of the 17 comorbid disease categories using the ICD-10/OPCS-4 translations of the Deyo and Dartmouth–Manitoba ICD-9-CM codes. The observed disagreement between the two methods of coding for each comorbidity varied from

Discussion

The Deyo and Dartmouth–Manitoba ICD-9-CM adaptations of the Charlson score were translated into ICD-10 codes and subsequently applied to a cohort of patients undergoing radical urological cancer surgery within a large English administrative database. The obtained Charlson scores were higher in older patients, in men, in those admitted to hospital on an emergency basis; the scores were also significant predictors of short-term outcome. Addition of either adaptation of the Charlson score to

Conclusions

Charlson scores derived from ICD-10 administrative data represent a valid approach to adjust for comorbidity. Charlson Deyo and Dartmouth–Manitoba scores derived from ICD-10 data performed similarly in risk models predicting hospital mortality following urological cancer surgery. Both scores can continue to be used to adjust for comorbidity following the anticipated implementation of the ICD-10 coding system in the United States. However, adjusting for comorbidity does not seem to have a large

Acknowledgments

M.N. was supported by the Bob Young Research Fellowship and the Research Fellowship Scheme of The Royal College of Surgeons of England. J. vd M. received a National Health Service Public Health Career Scientist Award. The funding sources had no involvement in the production of or the decision to submit this manuscript for publication. We would like to thank the Department of Health in England for providing the extract from the Hospital Episode Statistics database used in this study.

References (53)

  • V. Sundararajan et al.

    New ICD-10 version of the Charlson Comorbidity Index predicted in-hospital mortality

    J Clin Epidemiol

    (2004)
  • M.A. Cleves 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)
  • G.E. Rosenthal et al.

    Cleveland health quality choice: a model for collaborative community-based outcomes assessment

    Jt Comm J Qual Improv

    (1994)
  • R. Singh et al.

    Comorbidity assessment in men with localised prostate cancer: a review of currently available techniques

    Eur Urol

    (2004)
  • R. Chahal et al.

    A study of the morbidity, mortality and long-term survival following radical cystectomy and radical radiotherapy in the treatment of invasive bladder cancer in Yorkshire

    Eur Urol

    (2003)
  • H. Powell et al.

    Accuracy of administrative data to assess comorbidity in patients with heart disease: an Australian perspective

    J Clin Epidemiol

    (2001)
  • K.H. Humphries et al.

    Co-morbidity data in outcomes research: are clinical data derived from administrative databases a reliable alternative to chart review

    J Clin Epidemiol

    (2000)
  • A.K. Mukherjee et al.

    The completeness and accuracy of health authority and cancer registry records according to a study of ovarian neoplasms

    Public Health

    (1991)
  • R.E. Park et al.

    Explaining variations in hospital death rates

    JAMA

    (1990)
  • A. Elixhauser et al.

    Comorbidity measures for use with administrative data

    Med Care

    (1998)
  • L.I. Iezzoni

    Assessing quality using administrative data

    Ann Intern Med

    (1997)
  • L.I. Iezzoni

    Severity of illness measures: comments and caveats

    Med Care

    (1990)
  • M.S. Blumberg

    Risk adjusting health care outcomes: a methodologic review

    Med Care Rev

    (1986)
  • Department of Health and Human Services

    The international classification of diseases. 9th rev. ed. Clinical modification, ICD-9-CM

    (1998)
  • World Health Organization (WHO)

    International classification of diseases

    (1977)
  • J.D. Birkmeyer et al.

    Surgeon volume and operative mortality in the United States

    N Engl J Med

    (2003)
  • Cited by (177)

    • Psychopharmacological treatment patterns prior to a schizophrenia diagnosis: A Danish nationwide study

      2022, Schizophrenia Research
      Citation Excerpt :

      To compare with psychotropic drug use in the general population, we identified, for each individual with schizophrenia, five reference individuals matched on sex and birthyear who were without schizophrenia on the index date. We identified the following variables at the index date: age; sex; calendar year; previous mental disorder diagnoses; the presence of a chronic medical disease as defined in the Charlson comorbidity score (Charlson et al., 1987; Nuttall et al., 2006); occupational status; marital status; highest obtained education; lifestyle-related diagnoses (i.e., smoking- and alcohol-related disorders). The Danish National Prescription Registry has registered information on redeemed drugs since January 1, 1995 (i.e., four years before the first possible date of entering the study), while the other registers have registered data before 1995.

    View all citing articles on Scopus
    View full text