ACEF Score Calculator (Age, Creatinine, Ejection Fraction)
ACEF Score Calculator (Age, Creatinine, Ejection Fraction)
Three facts — age divided by ejection fraction, plus one point for a raised creatinine — and the published logistic transform that turns the ratio into a predicted mortality. Derived for elective cardiac surgery only.
ACEF score
3 variables → predicted mortality72 years old, ejection fraction 45 per cent, creatinine 1.1 mg/dL, elective coronary surgery
The score, the transform, and where it has failed
predicted mortality = ey / (1 + ey), y = 1.24 × ACEF − 5.41
- the transform, with its caveat stated
- the two-parameter logistic above was read in two independent sources — the score author’s own follow-up editorial and an independent single-centre validation — and both print the same coefficients, 1.24 and −5.41. Both texts reached this page as PDF extractions that lose exponent formatting, so the printed string is ambiguous about where the exponent ends; (1.24 × ACEF) − 5.41 is the only arithmetically sensible reading and is the one that produces plausible mortalities across the score’s range — 0.82 per cent at an ACEF of 0.5, 1.52 per cent at 1.0, 15.6 per cent at 3.0. That is as far as the sourcing goes and the page says so rather than implying more
- derivation and validation
- 4 557 adults having elective cardiac operations at a single institution, 2001 to 2003, with a validation series of 4 091 subsequent patients at the same institution. Area under the ROC curve 0.808 in validation against the Cleveland Clinic score’s 0.812, and 0.826 in isolated coronary operations against 0.806. The derivation paper’s own full text could not be read for this record; its abstract was, in full
- what it is NOT for
- the score was built and validated in elective operations only. Its author states that it was not validated in urgent or emergency procedures and that it does not account for the type of operation at all — so a redo triple procedure and an isolated bypass in the same patient score identically. This model is for CARDIAC surgery. The RCRI and ARISCAT are for non-cardiac surgery and were derived in cohorts from which cardiac operations were excluded; the two sets of numbers are not interchangeable in either direction.
- the binary creatinine, and the author’s own answer
- one point at a single threshold is a crude way to carry renal function, and the score was criticised in print for it. The author agreed the criticism was reasonable and that the score is “probably improvable” in that respect. A Cockcroft–Gault ACEF variant replaces the binary with one point for every 10 mL/min of creatinine clearance below 60, to a maximum of 6; in a 1 190-patient series of elective operations from 2012 to 2015 its overall accuracy was 51.2 per cent against the original’s 56.1 and EuroSCORE II’s 75.9
- which version you are holding
- ACEF II exists — a 2018 revision adding haematocrit and an emergency term to the same three variables. Its coefficients could not be read in any source reached for this page, so it is not built here; a reader given “an ACEF score” should establish which of the two it is, exactly as MELD, MELD-Na and MELD 3.0 are three different numbers
- why there are no bands
- ACEF publishes no risk strata. The cut-offs in circulation are ROC-derived inside individual validation cohorts and disagree with each other — 1.43 in one series, 1.283 in another whose own discussion then says 1.23 — so banding this page against any of them would present a local optimum as a published threshold. The predicted mortality is given as a number and left as one
Worked example
72 years old, ejection fraction 45 per cent, creatinine 1.1 mg/dL, elective coronary surgery
72 / 45 = 1.600
Creatinine 1.1 mg/dL is not above 2.0, so no point is added — ACEF = 1.600
y = 1.24 × 1.600 − 5.41 = 1.984 − 5.41 = −3.426
e−3.426 / (1 + e−3.426) = 3.15 per cent
Raise the creatinine to 2.1 mg/dL and the score jumps to 2.600, giving 10.10 per cent — a single point on a three-variable score more than triples the predicted mortality, which is what a binary term inside an exponent does
Enter the same creatinine as 186 µmol/L instead and the answer is identical: 186 / 88.4 = 2.104 mg/dL, still above 2.0
Drop the ejection fraction from 45 to 30 and the score rises to 2.400 and the prediction to 8.06 per cent; drop it from 60 to 45, the same fifteen points, and the prediction moves only from 1.94 to 3.15 per cent. The denominator is where this score gets its shape
ACEF’s reported discrimination, cohort by cohort
| Cohort | Patients | Area under the ROC curve |
|---|---|---|
| Derivation institution, validation series, elective cardiac surgery | 4 091 | 0.808 (Cleveland Clinic score 0.812) |
| Same series, isolated coronary operations only | — | 0.826 (Cleveland Clinic score 0.806) |
| Single-centre elective series, January 2011 to February 2013 | 993 | 0.849 (95% CI 0.820–0.876) |
| InCor, HC-FMUSP, elective coronary and valve surgery, May 2007 to July 2009 | 2 565 | 0.625 (0.58–0.67); InsCor 0.744, EuroSCORE 0.763 |
What the three cardiac models ask for
| ACEF | Additive EuroSCORE | EuroSCORE II | |
|---|---|---|---|
| Variables | 3 | 17 | 18 |
| Operation type | Not used at all | One binary (other than isolated CABG) | Four weights, plus thoracic aorta |
| Urgency | Not used — elective only | One binary | Four levels |
| Renal function | Creatinine above 2.0 mg/dL | Creatinine above 200 µmol/L | Cockcroft–Gault clearance, three levels |
| Ejection fraction | Continuous, as a denominator | Two categories | Four categories |
| Population it was fitted in | Elective cardiac surgery, one institution, 2001–2003 | European cardiac surgery, published 1999 | International cardiac surgery, 2010 |
Three variables, and the two things they leave out
The ACEF score is age in years divided by ejection fraction as a percentage, plus one point if the preoperative serum creatinine is above 2.0 mg/dL. That is the whole instrument. It was published in 2009 under a title invoking the law of parsimony, and the claim was not that three variables are better than thirteen but that in the cohort it was built in they were no worse: area under the ROC curve 0.808 against the Cleveland Clinic score’s 0.812 overall, and 0.826 against 0.806 in isolated coronary operations.
The division is doing real work. A points model treats a fifteen-point fall in ejection fraction the same wherever it starts; a ratio does not. From 60 to 45 per cent the predicted mortality here moves from 1.94 to 3.15 per cent; from 45 to 30 per cent, the same fifteen points, it moves from 3.15 to 8.06. That shape is closer to how ventricular function actually behaves than two or four categories are, and it costs nothing in inputs. The binary creatinine term is the opposite case: one point at one threshold, which was criticised in print and which the score’s author agreed was improvable. A Cockcroft–Gault variant grading renal impairment in steps exists and did not perform better.
Two omissions matter more than any of that. The score takes no account of the operation — an isolated bypass and a redo triple procedure in the same patient score identically — and it was built and validated in elective surgery only, with its author stating that it was not validated in urgent or emergency procedures. So it is a model of the patient and not of the operation, which is a legitimate thing to be as long as nobody forgets it. Where it has been tested outside its home institution the results have ranged widely: 0.849 in one 993-patient series, and 0.625 with an inadequate Hosmer–Lemeshow fit in 2 565 patients at InCor in São Paulo, where EuroSCORE reached 0.763 on the same patients.
Use it for what it is good for: a sanity check computable from three numbers that are already in front of you, and a reminder that most of the discrimination in cardiac surgical risk lives in very few variables. For the operation itself, EuroSCORE II carries procedure weight, urgency and sixteen other terms. A model’s predicted probability is not this patient’s probability: it is the rate observed in the cohort the model was fitted to, among patients who shared these inputs and differed in everything the model does not record. Every figure here comes from a named cohort, and cohorts differ in case mix, era, outcome definition and treatment; where your own institution’s protocol differs, it takes precedence. This page reports what a stratum predicted in a named study. It recommends no action.
Frequently asked questions
Is the creatinine threshold 2.0 mg/dL or above, or strictly above?
The sources disagree. The derivation paper’s abstract and three independent validations print “above 2.0 mg/dL”; the score author’s own follow-up editorial prints “2.0 mg/dL or more”. This page uses the strict inequality, which is both the majority reading and the primary one. It changes the answer for exactly one creatinine value, and there it changes it by a whole point — which on this score roughly triples the predicted mortality.
Why is there no risk band or category on this page?
Because ACEF does not publish any. The cut-offs in circulation — 1.43 in one cohort, 1.283 in another whose own discussion then quotes 1.23 — were derived by ROC analysis inside individual validation series and disagree with each other. Presenting one of them as a threshold would dress a local optimum up as a published stratum.
Can I use ACEF for an emergency operation?
It was not validated for one. The derivation and validation cohorts were elective cardiac operations, and the score’s author states in print that it was not validated in urgent or emergency procedures and takes no account of the type of operation. Both of those are exactly what EuroSCORE II’s urgency and weight-of-intervention terms carry.
What is ACEF II and is it the same thing?
No. ACEF II is a 2018 revision that adds haematocrit and an emergency term to the original three variables. Its coefficients could not be read in any source reached for this page, so it is not computed here. If someone hands you “an ACEF score”, establish which version it is before comparing it with anything.
How can three variables match a thirteen-variable model?
Because age, renal function and ventricular function carry most of the separable signal in elective cardiac surgical mortality, and additional variables in a risk model buy progressively less discrimination. What they do buy is calibration in subgroups — which is where ACEF’s results outside its home institution have been weakest, and where its lack of any operation term shows.
Related calculators
References
- Ranucci M, Castelvecchio S, Menicanti L, Frigiola A, Pelissero G. Risk of assessing mortality risk in elective cardiac operations: age, creatinine, ejection fraction, and the law of parsimony. Circulation. 2009;119(24):3053–3061. Abstract read in full through a university repository; the full text could not be reached.
- Ranucci M. The ACEF score one year after: a skeleton waiting for muscles, skin, and internal organs. EuroIntervention. One of the two sources read for the published logistic transform, and the source for the author’s own agreement that the binary creatinine term is improvable.
- A single-centre validation of the ACEF score in 993 elective cardiac surgery patients, January 2011 to February 2013, in the Turkish Journal of Thoracic and Cardiovascular Surgery. The second independent source read for the logistic transform.
- A validation of ACEF, InsCor and EuroSCORE in 2 565 patients having elective coronary or valve surgery at InCor, HC-FMUSP, May 2007 to July 2009, in Arquivos Brasileiros de Cardiologia: ACEF’s area under the curve was 0.625 there.
- Validation of age, creatinine and ejection fraction (ACEF) and Cockcroft–Gault ACEF scores in elective cardiac surgery, Medicina (Buenos Aires) 2017;77(4): 1 190 elective patients, 2012–2015, and the source for the CG-ACEF variant’s definition.
Not medical advice. For healthcare professionals and education. Reference intervals vary by laboratory and assay — always use your own laboratory's. Never base a dose or a treatment decision on this page alone. Full disclaimer at calcengines.com/disclaimer/