Charlson Comorbidity Index (CCI):
Developed in 1987, the CCI was originally designed to predict 1‑year mortality in a cohort of medical patients by assigning weights (ranging from 1 to 6) to 17 comorbid conditions. Its simplicity and historical use have made it very popular. However, it covers a relatively limited set of conditions.

Elixhauser Comorbidity Index (ECI):
Introduced in 1998 by Elixhauser and colleagues, this index includes around 30 conditions (with later modifications expanding or refining these numbers). Initially, it was not summarized into a single weighted score, but subsequent adaptations (for example, by van Walraven et al.) have provided a composite score. Many studies have found that the ECI often has better discriminative ability (especially for in‑hospital mortality, complications, and resource use) than the CCI, likely because it covers a broader range of comorbidities.

Implementation in NIS and HCUP Data:
The Agency for Healthcare Research and Quality (AHRQ) has developed software tools for assigning the Elixhauser comorbidity measures (and derived indices) specifically for ICD‑10‑CM data, which are often applied to NIS datasets. Many investigators favor using the Elixhauser index when working with the NIS because of its comprehensive nature and because it has been validated in large administrative databases (​

Degrees of Freedom and Sample Size:
With very large datasets like the NIS, researchers sometimes prefer to include individual comorbidity indicators instead of—or in addition to—a summary score. Using individual components can allow for more customized adjustment because the weights derived in one population may not perfectly match the current study’s population. However, when the sample size is enormous, as with the NIS, including a larger number of parameters is generally less problematic. In smaller datasets, a summary score can be beneficial to conserve degrees of freedom.

HCUP Resources:

• The Healthcare Cost and Utilization Project (HCUP) provides comprehensive documentation and tools related to the NIS. Their reports, such as the "Method Series Report #2016-02," discuss the application of comorbidity measures, including both the Elixhauser and Charlson indices, especially in the context of transitioning to ICD-10-CM/PCS coding systems.

HCUP Tools: Labels and Formats

• Format Programs
  • DRG Formats Program Creates SAS formats to label the values of each DRG and MDC category
  • HCUP Formats Program  Creates SAS formats to label the values of selected categorical data elements in HCUP files
  • HCUP Diagnosis and Procedure Groups Formats Program  Creates SAS formats to label the values of HCUP Diagnosis and Procedure Groups data elements, including Clinical Classifications Software Refined (CCSR) data elements
  • ICD-9-CM Formats Program Creates SAS formats to label the values of ICD-9-CM Diagnoses and Procedures
  • ICD-10-CM Formats Program  Creates SAS formats to label the values of ICD-10-CM Diagnoses and Procedures
  • Severity Formats Program Creates SAS formats to label the values of data elements in HCUP Severity Files

    Discussion:

    On various research forums (including discussions on NIS or related data analysis forums), many investigators point out:

    • The Elixhauser index’s broader range of conditions tends to provide better risk adjustment for outcomes like mortality.
    • When researchers have a sufficiently large sample (as with NIS), it is often preferable to include individual comorbidity indicators rather than relying solely on a pre‐calculated summary score.
    • Some debates have centered on whether one should use a summary measure or adjust for each condition individually. The consensus tends to be that if your sample size is very large, using the individual components may offer better control for confounding because you are not “borrowing” weights from another cohort (​
      discourse.datamethods.org
      statalist.org

      CONCLUSION:
      Ultimately, the choice between using the Charlson index, the Elixhauser index, or individual comorbidity indicators should be guided by your specific research question, sample size, and the outcomes of interest. For many NIS-based studies, the Elixhauser index (or its individual components) is now preferred for its improved discriminative ability