Article Text

  1. J Tang,
  2. J. Y. Wan,
  3. J. E. Bailey
  1. University of Tennessee Health Science Center, Memphis, TN


Background The Charlson Index (CI) is commonly used in outcome studies to adjust for patient comorbid conditions but has not been specifically validated for use in studies of hypertension. Previous studies have primarily assessed the validity of the CI in predicting mortality but not stroke incidence, another critical outcome for hypertensive individuals. The present study's purpose was to validate CI for death and stroke outcomes in a cohort of hypertensive patients and to compare CI performance to that of four alternative comorbidity indices assessing prior cardiovascular events (CVD), traditional stroke-risk factors (CVA), health care utilization (HCU), and medication utilization (MU) respectively.

Methods This retrospective cohort study identified 49,479 Tennessee Medicaid (TennCare) enrollees aged 18 to 64 with chronic, drug-treated hypertension who met strict eligibility criteria for a 2-year baseline period and a 1- to 5-year follow-up period from 1994 to 2000. Comorbidity measures were calculated for each participant both at baseline and for the 6-month period prior to event or study conclusion. Modified CI scores were calculated using hospital-discharge and outpatient-visit International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) diagnosis codes. Use of CI score was compared with the use of CVD, CVA, HCU, and MU, controlling for demographic variables in Cox proportional hazards models. Five models using baseline data compared the comorbidity measures for each of the two outcomes. Similar models were also compared, using follow-up data for the period prior to event or study conclusion.

Results Models using baseline, time-independent data gave similar results to those of models using follow-up, time-dependent data. Whether baseline or follow-up data were employed, the model using the modified CI had the smallest Akaike Information Criterion (AIC) value among 5 models for death outcome. However, when assessing risk of stroke, the model using MU had the smallest AIC value whether baseline or follow-up data were employed.

Conclusions This study indicates that the modified CI is a valid tool in assessing comorbidity and predicting mortality in patients with hypertension. However, an index of MU incorporating measures of medication-variety exposure, regimen complexity, adherence, and thiazide-medication exposure was better than the modified CI was in predicting stroke in such patients.

Statistics from

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.