Charlson Comorbidities in ICD10 codes

Details with Explanation -- Yes this is LONG, but it's to try and answer all the various questions that will arise

• The original description by Charlson of what has become the Charlson Comorbidity Index [CCI] (J. Chr. Dis. 40(5):373-383, 1987) was oriented towards identifying the cumulative burden of pre-existing medical conditions for the purpose of taking account of comorbid conditions that might alter mortality in longitudinal research studies. She identified, by manual chart review, a set of 17 conditions to which were applied weights, resulting in a composite "Charlson Score". Furthermore, in her original schema, these weights were modified by age. She found that in a survival analysis of time to death among people with breast cancer, that this comorbidity index was a significant predictor of the hazard of death.
  • The 17 conditions are: Myocardial infarction; Congestive heart failure; Peripheral vascular disease; Cerebrovascular disease; Dementia; Chronic pulmonary disease; Rheumatic disease; Peptic ulcer disease; Mild liver disease; Moderate or severe liver disease; Diabetes without chronic complications; Diabetes with chronic complications; Hemiplegia or paraplegia; Renal disease; Any malignancy, including lymphoma and leukemia, except of skin; Metastatic solid tumor; AIDS. Some notes about these 17 conditions:
    • Identifying them from chart review requires interpretation -- i.e. it's not always a "slam-dunk". For example, WHICH complications of diabetes should be included? Another example is how should we decide if the liver disease is severe or not? None of these questions of interpretation were defined in Charlson's original formulation.
    • Among the 17 are two sets of mutually exclusive codes -- the 2 on DM and the 2 on liver disease: e.g. if somebody has DM with chronic complications, you're supposed to give them the points for THAT, but NOT also for DM without chronic complications.
• One might ask whether it's reasonable to include only 17 (or 31 or 100 for that matter) specific disorders in trying to measure the cumulative comorbid burden. After all, there are 16,000 ICD-10 codes and the vast majority of them can be pre-existing conditions. The issue here is that there's clearly a balance here between: (i) being practical in not including too many conditions while still maintaining generalizable results, and (ii) including enough conditions that the resulting measure is still meaningful.
• Despite the original purpose of the CCI, it started being used as a measure of the cumulative burden of comorbid illness in a wide variety of purposes. The next big step was taken by Deyo (J. Clin. Epidemiol. 45(6):613-619, 1992) who in created a set of ICD-9 codes for using administrative health data (hospital abstracts) to calculate the CCI. And this is WITHOUT the age-related portion of the original description by Charlson herself.
• Partly lost in the process of moving the CCI from a clinical/chart review thing to something that can be identified from administrative data is if/how to identify whether the diagnosis of interest was actually pre-existing. This is relevant because in hospital and outpatient administrative data across countries, the ability to distinguish a pre-existing condition from one which was a reason for admission or began after admission is highly variable.
  • While many/most of the Charlson conditions, HAD to have been pre-existing, even if it was not actually recognized as being present pre-admission (e.g. Dementia is never acute, while it may be a reason for admission, it MUST have been present prior to admission), some of the 17 conditions could be either -- an example is MI.
• The most recent seminal paper was a modified version of Deyo's coding by Quan (Medical Care 43(11):1130-1139, 2005) who created sets of ICD-9 and ICD-10 codes for CCI (and also, by the way for a different schema by Elixhauser et al. for identifying 31 specific comorbid conditions to assess the cumulative burden of comorbid disease).
  • Quan assessed the value of this administrative data CCI for predicting in-hospital mortality using the Discharge Abstract (hospital) Database in Calgary. He found that the predictive power of the CCI was high, and it was a bit higher if he used ALL hospital diagnoses (i.e. included by us as Comorbid Diagnoses, Admit Diagnoses and Acquired Diagnoses) than if he just used pre-hospital diagnoses identified by us as Comorbid Diagnoses.
• Other work on comorbidity indices has shown that value in predicting death is higher if: (a) both inpatient and outpatient diagnoses are used, and (b) you use not only diagnoses from the current hospitalization but also from 1-2 years prior.

Summary

• There is no ONE correct version of the Charlson Comorbidity Index (CCI), or of any comorbidity measures.
• The "right" one for your own purpose should be used.
• If that purpose is to best help predict bad outcomes such as death, then using diagnoses in any of our Comorbid Diagnoses, Admit Diagnoses and Acquired Diagnoses is best -- but this may mix some acute diagnoses with pre-existing conditions.
• If the purpose is to truly identify the burden of pre-existing illness, then only Comorbid Diagnoses should be used.
• BOTH for simplicity AND to have a measure that is consistent with the way it is usually computed, we have chosen to use the Quan's specific ICD-10 codes, regardless of whether they're listed Comorbid Diagnoses, Admit Diagnoses or Acquired Diagnoses.

Our implementation

This is a list of the ICD10 diagnoses that contribute to the Charlson Comorbidity Index.

Points are not included in the listings because the score depends on the combination of diagnoses. For example, if a patient has two diagnoses from the same Charlson component, they are only counted once. Also, if a patient has diagnoses from a lesser and more severe component of the same type (e.g. Diabetes without and with complications), only the higher one counts.

See Charlson Comorbidity scoring in ICD10 codes for more info.

See #How this page works for info on how these codes are stored on the wiki.

Listing of ICD10 comorbid codes

There are 0 of these:

Semantic annotations

This will appear empty but here is where you would edit the content of the table above.

populated using query s_ICD10_Charlson_Comorbids in CCMDB.accdb. If you edit here, make sure you update all related data in the database!

How this page works

#Listing of ICD10 comorbid codes contains a semantic query that lists the data.

#Semantic annotations contains calls to Template:ICD10 Charlson which populates the semantic data that drive the listing. That part can be populated directly from column "wikiline" of the S ICD10 Charlson Comorbids query.