Written and maintained by CASRAI Editorial Board
Last updated
The Elixhauser Comorbidity Index is a risk-adjustment tool that quantifies a patient’s pre-existing chronic disease burden from administrative diagnosis codes — the same purpose the Charlson Comorbidity Index serves, but built differently and, in several published comparisons, a stronger predictor of short-term in-hospital mortality. It was developed by Elixhauser A, Steiner C, Harris DR, and Coffey RM (“Comorbidity measures for use with administrative data,” Medical Care, 1998;36(1):8–27) as a set of 30 comorbidity measures identified directly from ICD diagnosis codes on a hospital discharge record, rather than a single clinician-assigned severity judgment. The Agency for Healthcare Research and Quality (AHRQ) now maintains the index as free, versioned software — the Elixhauser Comorbidity Software Refined for ICD-10-CM — which is the version most US hospitals and researchers actually run today.
What the index actually measures, and how that differs from Charlson
Elixhauser and Charlson answer the same underlying question — how sick was this patient before this admission, independent of what happened during it — but they get there differently. Charlson assigns a single weighted point value (1, 2, 3, or 6) to 19 conditions and sums them into one number. The original 1998 Elixhauser measures were deliberately unweighted: each of the 30 categories is a binary flag (present or absent), with no single summary score in the source paper. That design choice was intentional — the original authors argued that forcing every comorbidity into one weighted scale discards information a regression model can use directly as 30 separate covariates. The tradeoff is that a flag list isn’t a single number a reader can eyeball the way a Charlson score is, which is the practical reason weighted Elixhauser variants were built afterward (see below).
Published comparisons have found Elixhauser’s larger, more granular category list predicts short-term outcomes — in-hospital mortality in particular — at least as well as, and in several studies better than, the older Charlson index, while Charlson remains more established for longer-horizon (1-year+) survival prediction and is far more widely used as a simple reported number in clinical papers. Neither supersedes the other; which one a study or registry should use depends on the outcome being modeled, and some risk-adjustment protocols report both.
From 30 measures to 38: the AHRQ ICD-10-CM update
The original 1998 paper defined 30 comorbidity categories, derived and validated against ICD-9-CM discharge data. As ICD-10-CM adoption spread and coding granularity increased, AHRQ’s Healthcare Cost and Utilization Project (HCUP) built and now versions the Elixhauser Comorbidity Software Refined for ICD-10-CM, which currently defines 38 comorbidity categories — broad groupings spanning cardiovascular disease (hypertension, uncomplicated and complicated, congestive heart failure, cardiac arrhythmias, valvular disease, peripheral vascular disease), endocrine and metabolic disease (diabetes with and without complications, obesity, weight loss, fluid and electrolyte disorders), pulmonary disease, renal failure, liver disease, neurological disorders, psychiatric conditions (depression, psychoses), substance use disorders, coagulopathy, anemia, and malignancy (solid tumor without and with metastasis, lymphoma). AHRQ’s HCUP tools page is the authoritative source for the exact current category list and ICD-10-CM code mappings — those mappings are versioned and updated, so a research team building or auditing a risk-adjustment model should pull the current release directly from HCUP rather than transcribing an older list.
From a flag list to a score: van Walraven and the AHRQ summary indices
Because the original Elixhauser measures produce 30 (now 38) separate binary flags rather than one number, several groups built weighted summary scores on top of them so the index could be reported and compared the way a Charlson score is. The most widely cited is van Walraven C, Austin PC, Jennings A, Quan H, and Forster AJ, “A modification of the Elixhauser comorbidity measures into a point system for hospital death using administrative data,” Medical Care, 2009;47(6):626–633, which assigns each comorbidity category a weight (some negative, reflecting comorbidities associated with lower mortality risk in that model) and sums them into a single point value calibrated specifically to predict in-hospital death. AHRQ’s own Refined software goes a step further and computes two separate summary scores from the same 38 flags — one set of weights calibrated to predict in-hospital mortality, and a second, differently weighted set calibrated to predict 30-day all-cause readmission — because the comorbidities that predict dying during this admission are not the same ones that predict bouncing back within 30 days. A team citing “an Elixhauser score” without specifying which weighting scheme and software version produced it has not actually specified a reproducible measure; the weighting scheme is not incidental, it is the number.
Its role in hospital quality reporting and risk-adjusted research
Elixhauser comorbidities feed the same class of risk-adjustment work covered in our Observed-to-Expected (O/E) Ratio and Risk Adjustment guide and in outcome models behind public reporting programs including the Hospital Readmissions Reduction Program, the CMS Overall Hospital Star Rating Methodology, and the AHRQ Patient Safety Indicators AHRQ itself publishes. Because Elixhauser is derived entirely from coded secondary diagnoses already present on the discharge abstract — no chart abstraction, no additional clinician documentation burden — it is cheap to compute at scale across an entire hospital’s or health system’s discharge data, which is exactly why it shows up as a standard covariate in large administrative-data studies and multi-hospital comparisons rather than single-patient bedside use. For a quality-registry or research-administration team validating a risk-adjustment model against a national benchmark, the practical requirement is the same one that applies to Charlson: state explicitly which software version, ICD code set, and (if reporting a single score rather than the raw flags) which weighting scheme — van Walraven, AHRQ mortality, or AHRQ readmission — was used, since the three do not produce the same number for the same patient.
Frequently asked questions
Is the Elixhauser Comorbidity Index the same as the Charlson Comorbidity Index?
No. Both are administrative-data comorbidity measures used for risk adjustment, but they were developed independently, use different (and non-identical) lists of conditions, and score them differently — Charlson sums a small set of pre-assigned weights (1, 2, 3, or 6 points per condition) into one number; the original Elixhauser measures are 30 (now 38, in the AHRQ ICD-10-CM Refined version) unweighted binary flags, with weighted summary scores available separately. A study or registry should specify which one it used, and the two scores are not interchangeable or directly convertible.
How many comorbidity categories does the Elixhauser Comorbidity Index include?
The original 1998 paper defined 30 categories from ICD-9-CM data. The current AHRQ Elixhauser Comorbidity Software Refined for ICD-10-CM defines 38 categories — the expansion reflects both increased ICD-10-CM coding granularity and refinements made since the original publication. Always confirm which version and category count a given dataset or published study used before comparing scores across sources.
What’s the difference between the raw Elixhauser flags and a single “Elixhauser score”?
The original Elixhauser measures are not a single score at all — they’re a set of yes/no flags meant to be used as separate covariates in a regression model. When a paper or dashboard reports one “Elixhauser score” as a single number, it’s almost always using a weighted summary index built on top of the flags — most commonly van Walraven’s 2009 point system, or one of AHRQ’s own mortality- or readmission-weighted summary scores — and different summary schemes will produce different numbers for the same patient, so the specific scheme matters.
Do I need special software to calculate an Elixhauser score?
In practice, yes — unlike a bedside instrument scored by hand, Elixhauser comorbidities are derived from administrative/claims diagnosis codes at the population or cohort level, using AHRQ’s free HCUP software (SAS, R, and other implementations exist) that maps ICD-10-CM secondary diagnosis codes to the 38 categories and, optionally, computes the mortality- and readmission-weighted summary scores. It is not typically calculated from a single chart by hand the way a bedside clinical score is.
