Charlson or Elixhauser: choosing a comorbidity index

The Charlson index was derived in a cohort of 559 medical inpatients to predict mortality at one year. It weights 16 comorbidity categories from 1 to 6 points, and an age adjustment of up to 4 points was added later to give the combined age–comorbidity index used in most orthopaedic papers.

The Elixhauser measures were created to adjust for comorbidity in hospital administrative data, and van Walraven and colleagues later converted them into a single point score for in-hospital death. Thirty conditions each carry an integer weight, and the weights run from −7 to +12.

In a 2021 analysis of every adult general-hospital inpatient in Switzerland between 2012 and 2017 — more than six million cases — a model built from age, sex and hospital type predicted in-hospital mortality with a c-statistic of 0.757. Adding the Charlson weights raised it to 0.850, and adding the van Walraven-weighted Elixhauser score raised it to 0.863.

That is a modest advantage for the Elixhauser weights on that outcome in that dataset. It does not show that the Elixhauser score is better at predicting survival over years, which is the question the Charlson index was built for, and it does not transfer automatically to a different country or setting.

Six Elixhauser comorbidities carry negative weights in the van Walraven system: valvular disease, obesity, blood loss anaemia, deficiency anaemias, drug abuse and depression. These describe an association in the derivation data — after adjusting for the other comorbidities, patients coded with these conditions had lower in-hospital mortality — and not a protective effect. A negative contribution is not a reason to expect a better outcome for an individual patient.

Nine further conditions, including hypertension and diabetes, carry a weight of zero and cannot change the total.

For adjusting for case mix in a registry or administrative-database study where diagnosis codes are available, and the outcome is in-hospital death, the Elixhauser score is a reasonable default. For a question about longer-term survival, the Charlson index is the instrument built for it, with the caveat that its survival formula was fitted to patients treated decades ago and overstates the risk under current care.

Neither answers fitness for anaesthesia, which is what the ASA physical status class describes. And neither is designed to counsel one patient about their prognosis: both describe groups.