NCI Comorbidity Index Overview

History of the NCI Comorbidity Index

The Charlson Comorbidity Index was first developed in 1987 by Mary Charlson and colleagues as a weighted index to predict risk of death within 1 year of hospitalization for patients with specific comorbid conditions. Nineteen conditions were included in the index. Each condition was assigned a weight from 1 to 6, based on the estimated 1-year mortality hazard ratio from a Cox proportional hazards model. These weights were summed to produce the Charlson comorbidity score. In 1992 and 1993, Richard Deyo and Patrick Romano separately adapted the Charlson index to ICD-9-CM diagnosis and procedure codes and CPT-4 codes so that the index could be calculated using administrative data.

In 2000 Carrie Klabunde and colleagues developed a cancer-specific NCI Comorbidity Index using breast and prostate cancer cases included in the SEER-Medicare data. Klabunde et al used codes on the Medicare data to identify the conditions identified by Charlson et al. The NCI Comorbidity Index excludes solid tumors, leukemias, and lymphomas as comorbid conditions, given that the NCI Comorbidity Index was developed from a cohort of cancer patients. The remaining 16 Charlson index conditions were included in the NCI Comorbidity Index, with further consolidation to 14 conditions. The approach for creating the NCI Comorbidity Index included review of Medicare claims for the year prior to the date of diagnosis, excluding the month of diagnosis. Separate comorbidity indexes were created for hospitalizations (Medicare Part A) and physician claims (Medicare Part B). Conditions ascertained by physician claims were required to occur more than once in a period greater than 30 days to reduce the risk of false positives. These indexes were later combined into a single NCI Comorbidity Index, which allowed for greater efficiency in analysis.

Changes to the NCI Comorbidity Index

In 2014, staff members from the Healthcare Delivery Research Program (formerly the Applied Research Program) and Surveillance Research Program undertook a thorough evaluation of the codes used to define each of the 16 conditions in the NCI comorbidity index as well as the weights associated with each condition. Professional coders reviewed all original codes to confirm that the codes were accurate and relevant to the conditions and procedures listed in the original macro. New codes were added as appropriate. The original and new codes can be found here. A separate technical report (PDF, 344 KB) summarizes changes made to the weights included in the NCI Comorbidity Index.

A pilot study was performed using the SEER-Medicare data to assess if revising the codes in the NCI Comorbidity Index changed the estimate of the prevalence of comorbid conditions in cancer patients. SEER patients diagnosed with cancer from 1997-2007 were selected if they were age 66 and over, were covered by Medicare Parts A and B without HMO enrollment during the year prior to their cancer diagnosis. Patients were excluded if they were diagnosed by death certificate and autopsy or had a prior cancer. Non-malignant tumors were excluded. Patients were stratified by cancer site, lung, colorectal, prostate, and breast, and other. Comorbid conditions were extracted from the Medicare claims one year prior to diagnosis. Sources of medical claims included the hospital (Medpar), outpatient facility and physician supplier (carrier) Medicare data. ICD-9-CM diagnosis, ICD-9-CM procedure, and CPT procedure codes were used to identify comorbid conditions. Special focus was devoted to CPT-4 codes because these codes change more frequently than do ICD-9 codes and excluding them from the algorithm could improve stability of the definitions over time. Breast cancer patients from 1997-2007 were used to estimate the prevalence of comorbid conditions using two algorithms with old codes and new codes.

As can be seen on the table below, dropping CPT-4 codes from the index had minimal effect on ascertainment of comorbid conditions. Adding new ICD-9 diagnostic and procedure codes increases disease ascertainment, especially for non-surgical PVD and dementia and these results were similar across cancer sites.

*Example calculations for Moderate/ severe liver disease, “revised with CPT”
Prevalence: 114 / 105465 = 0.0011 0.0011*100 = 0.11%
Percent Change: (114 – 99) / 99 = 0.1515 0.1515*100 = 15.15%

†PVD surgical and diagnosis conditions were combined into one category for PVD in the 2014 version of the macro. Prevalence rates for PVD combined were similar to PVD diagnosis.

Prior to October 2015, all ICD diagnosis and procedure codes on Medicare claims were reported using version ICD-9-CM. The Department of Health and Human Services mandated the use of ICD-10-CM beginning in October 2015. So, in 2021, the macro was updated to include ICD-10-CM codes per Quan et al. Other changes included in the 2021 version were to exclude procedure codes and change the “rule-out” option to look for more than one claim for a condition (rather than for a specific code) at least 30 days apart.

Accessing NCI Comorbidity Index Macros

Software to implement the NCI comorbidity macros are available for download here.

References

1. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis 1987;40(5):373-83. [View Abstract]
2. Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol 1992 Jun;45(6):613-9. [View Abstract]
3. Romano PS, Roos LL, Jollis JG. Adapting a clinical comorbidity index for use with ICD-9-CM administrative data: differing perspectives. J Clin Epidemiol 1993 Oct;46(10):1075-9; discussion 1081-90. [Look up in PubMed]
4. Klabunde CN, Potosky AL, Legler JM, Warren JL. Development of a comorbidity index using physician claims data. J Clin Epidemiol 2000 Dec;53(12):1258-67. [View Abstract]
5. Klabunde CN, Legler JM, Warren JL, Baldwin LM, Schrag D. A refined comorbidity measurement algorithm for claims-based studies of breast, prostate, colorectal, and lung cancer patients. Ann Epidemiol 2007 Aug;17(8):584-90. [View Abstract]
6. Quan H, Sundararajan V, Halfon P, Fong A, Burnand B, Luthi J-C, Saunders LD, Beck CA, Feasby TE, Ghali WA. Coding Algorithms for Defining Comorbidities in ICD-9-CM and ICD-10 Administrative Data. Med Care 2005;43: 1130–1139. [View Abstract]