Background: The Charlson comorbidity index (CCI) has been one of the most commonly used and validated prognostication tools. In addition to being a valuable resource for health services researchers, the CCI is being used in clinical practice by hospitals to identify patients at risk for poor outcomes. The use of the CCI in older adults has been controversial as it does not account for measures such as functional status and severity of dementia. The objective of this study was to evaluate the Charlson Comorbidity Index (CCI) for the prediction of short-term outcomes in hospitalized older adults.
Methods: A cohort study comparing length of stay (LOS), in-hospital mortality, and 30-day readmissions in hospitalized medical patients 75 years and older with different levels of comorbidity at baseline. Administrative data (ICD-9-CM adaptation) was used in translating documented ICD-9 codes in the electronic medical record (EMR) into the comorbidity score. EMR documentation used to obtain the CCI included any past medical history indicated on the present admission as well as any diagnosis documented on any prior admission. Two CCI was calculated using the traditional Quan index. Logistic regression was used to determine the predictive ability for the CCI in regards to in-hospital mortality and 30-day admissions. The Pearson correlation coefficient was used to determine the association of the CCI with the LOS.
Results: Of the 2,990 patients, the average age was 84.6, 59.7% were female, 78.1% were white, and 9.5% black. The majority of subjects were married (45.9%) and widowed (39.2%). The average LOS was 6.3 days (median 5 days), with 2.0% (59 patients) in-hospital mortality, and 20% were readmitted within 30-days. There was a significant association between the CCI and in-hospital mortality (OR=1.20; 95% CI: 1.08, 1.32) as well as 30-day readmissions (OR=1.17; 95% CI: 1.11, 1.22). However, both models had demonstrated a poor predictive ability (AUC of 0.6319, 95% CI: 0.5622-0.716 and 0.5962, 95% CI: 0.5624-0.6300, respectively). In addition while the CCI showed a significant association with LOS, the model had poor predictive accuracy (rho=0.12, p<0.0001).
Conclusions: The study demonstrates that despite a small but significant association the CCI is not a reliable predictor of short-term outcomes in a cohort of hospitalized medical older adults. Today, Medicare patients account for over 50% of hospital days and over 30% of all hospital discharges in the United States. While improving quality of care for all older adults is essential, tools that can identify patients that are at high risk of poor outcomes are critical in order to better allocate resources. Comprehensive tools that account for more comprehensive measures, such as functional and cognitive status are critical in identifying this vulnerable population.