TRISS Survival Probability Calculator

The TRISS logistic model from weighted RTS, ISS and an age threshold at 55, with separate blunt and penetrating coefficients and both published revisions side by side. A registry probability is not this patient’s.

Last updated 8 Oct 2026

TRISS survival probability

RTS + ISS + age → Ps

Weighted Revised Trauma Score (0 to 7.8408)The weighted form, not the coded 0 to 12 total. Get it from the Revised Trauma Score calculator, which prints both. Entering a coded total here is the commonest way this model is misused: it puts a value of up to 12 into a term whose domain stops at 7.8408. Anything above 7.8408 is refused here rather than answered.

Injury Severity Score (1 to 75)The ISS, not the NISS. The coefficients were fitted to ISS values, and the NISS runs higher for the same patient, so substituting one for the other lowers the modelled survival without any change in the patient. Use the ISS calculator.

Age (years)Age enters as a threshold, not a continuum: the age index is 0 below 55 years and 1 at 55 and above, and nothing changes between 55 and 95. A 54th and a 55th birthday are the only ages that matter.

MechanismSeparate coefficient sets, not a single model with a mechanism term. Paediatric practice in at least one published implementation applies the blunt model to both mechanisms under 15 years.

Coefficient setTRISS has been recoefficiented. Both published sets are offered because they disagree materially on the same patient, and a TRISS quoted without its coefficient set is an incomplete number.

85.2% modelled survivalExample

Weighted RTS 5.8806, ISS 34, age 42, blunt mechanism, original MTOS coefficients

Formula

Ps = 1 / (1 + e^−b)
b = b₀ + b₁ × RTS + b₂ × ISS + b₃ × AgeIndex
AgeIndex = 0 for age 54 or under, 1 for age 55 or over

original MTOS set, blunt: b = −1.2470 + 0.9544(RTS) − 0.0768(ISS) − 1.9052(AgeIndex). Read in a 2026 International Surgery Journal validation study, which prints the equation in full, and independently in an open-source trauma calculator’s source. The two agree on all four numbers
original MTOS set, penetrating: b = −0.6029 + 1.1430(RTS) − 0.1516(ISS) − 2.6676(AgeIndex). Same two sources, same agreement. Note how different the penetrating set is: the ISS coefficient is twice as steep and the age penalty is 40 per cent larger
later MTOS revision, blunt and penetrating: blunt b = −0.4499 + 0.8085(RTS) − 0.0835(ISS) − 1.7430(AgeIndex); penetrating b = −2.5355 + 0.9934(RTS) − 0.0651(ISS) − 1.1360(AgeIndex). From the Société Française d’Anesthésie et de Réanimation’s TRISS tool, with the blunt set confirmed independently by Jung and colleagues in Yonsei Medical Journal, who publish it component-wise: 0.8085 × the three RTS weights gives exactly their 0.2351, 0.5923 and 0.7574
age is a step: 0 below 55, 1 from 55. Nothing between 55 and 95 changes the answer, and a 54-year-old and a 55-year-old with identical injuries get different probabilities. A crude treatment, and a recognised limitation: one source read for this page states flatly that “the cutoff point of 55 years has to be revised”
derivation and drift: the Major Trauma Outcome Study, an American multicentre trauma registry assembled in the 1980s, whose norms let a single centre compare its outcomes against a benchmark. That is also the limitation: the coefficients encode 1980s trauma care. A 2026 prospective validation of 180 adults at a tertiary centre in Maharashtra reported an area under the curve of 0.994 for TRISS with 56.3 per cent sensitivity and 100 per cent specificity for death, and concluded that coefficients should be recalibrated to local conditions
what it will not tell you: A figure from a registry model is a property of the cohort it was fitted to, not a probability for the patient in front of you. This page computes the published number and states what it predicted in that cohort; it renders no decision, no prescription and no prognosis.

Worked example

Weighted RTS 5.8806, ISS 34, age 42, blunt mechanism, original MTOS coefficients
AgeIndex = 0, because 42 is under 55
b = −1.2470 + 0.9544(5.8806) − 0.0768(34) − 1.9052(0)
b = −1.2470 + 5.6125 − 2.6112 = 1.7543
Ps = 1 / (1 + e^−1.7543) = 1 / 1.1730 = 0.8525, 85.2 per cent modelled survival
Change the age to 55 and nothing else: b falls by 1.9052 to −0.1509 and Ps becomes 46.2 per cent. One birthday, 39 percentage points
Change the mechanism to penetrating with the same inputs and Ps becomes 72.4 per cent — the ISS coefficient is twice as steep in the penetrating model, and an ISS of 34 is punished for it
Switch to the later coefficient revision and the blunt figure moves to 81.2 per cent. The patient has not changed; only the revision has, which is why a TRISS quoted without its coefficient set is incomplete
Finally, try the coded RTS of 9 instead of the weighted 5.8806. This page refuses it, because 9 is outside the 0 to 7.8408 domain of the term — but the arithmetic, if it were allowed to run, would return 99.1 per cent instead of 85.2. That is the error the RTS page and this range check both exist to prevent

Published coefficient sets

Both sets are printed because both are in published use and they disagree on the same patient. Each was read in two independent sources that agree on every digit, and the revision’s blunt RTS coefficient of 0.8085 was confirmed a third way: multiplied by the three Revised Trauma Score weights it reproduces Jung and colleagues’ component coefficients 0.2351, 0.5923 and 0.7574 to four decimals. An internally consistent wrong coefficient set passes every structural check, so a cross-source arithmetic identity is the only thing that catches a mistyped digit.
Set and mechanism	b₀	b₁ (RTS)	b₂ (ISS)	b₃ (AgeIndex)
Original MTOS, blunt	−1.2470	0.9544	−0.0768	−1.9052
Original MTOS, penetrating	−0.6029	1.1430	−0.1516	−2.6676
Later MTOS revision, blunt	−0.4499	0.8085	−0.0835	−1.7430
Later MTOS revision, penetrating	−2.5355	0.9934	−0.0651	−1.1360

The same patient, four ways

Every figure is this calculator’s own arithmetic. The mechanism switch moves the answer by 12.9 percentage points on the first row, and the coefficient revision by 4.0 — so the two selectors above are not cosmetic. Note the last row: both sets print 0.0 per cent to one decimal and neither is truly zero — a logistic function never reaches it, and the original set returns 0.013 per cent for that patient.
Inputs	Original MTOS	Later revision
RTS 5.8806, ISS 34, age 42, blunt	85.2%	81.2%
RTS 5.8806, ISS 34, age 55, blunt	46.2%	43.1%
RTS 5.8806, ISS 34, age 42, penetrating	72.4%	74.9%
RTS 7.8408, ISS 9, age 30, blunt	99.6%	99.4%
RTS 0, ISS 75, age 80, blunt	0.0%	0.0%

A 1980s registry model, and the three ways it is fed wrong

TRISS combines what the patient’s physiology looked like at first contact, how badly they were injured anatomically, and whether they were old, into a single logistic model of survival. The physiology comes in as the weighted Revised Trauma Score, the anatomy as the Injury Severity Score, and age as a binary index at 55 years. Separate coefficient sets are fitted for blunt and penetrating mechanisms, because the way injury severity maps onto death differs between them: in the original set, the ISS coefficient is twice as steep for penetrating trauma.

The model was built on the Major Trauma Outcome Study, an American multicentre registry assembled in the 1980s, and its purpose was comparison rather than prediction — a centre could score its own patients and ask whether its death rate was better or worse than the benchmark. That origin is also the standing limitation. The coefficients encode the trauma care of their era, before damage-control resuscitation, balanced transfusion, tranexamic acid, interventional haemorrhage control and whole-body CT. The coefficients have been revised since, and the two published sets disagree by up to about fourteen percentage points on the same patient, so a TRISS reported without saying which set produced it is an incomplete number. A 2026 prospective single-centre validation reported an area under the curve of 0.994 but with a sensitivity for death of only 56 per cent, and argued for local recalibration rather than for the published coefficients.

Three input errors account for most wrong TRISS values. The first is entering the coded Revised Trauma Score, which runs to 12, in place of the weighted value, which stops at 7.8408; on this page’s worked example that moves the answer from 85 per cent to 99.5. The second is entering a New Injury Severity Score where an ISS belongs: the NISS is never lower than the ISS, so the substitution silently lowers the modelled survival. The third is treating age as a continuous variable. It is a step at 55 and nothing else, which means a 54-year-old and a 55-year-old with identical injuries receive different probabilities and a 55-year-old and a 95-year-old receive the same one.

The output is a property of the registry the model was fitted to. It is not this patient’s chance of living, and it does not belong in a conversation about whether to continue treatment. Where TRISS has a legitimate bedside-adjacent use it is retrospective: a death in a patient whose modelled survival was high is flagged for peer review, which is a question about the service rather than a judgement about the patient.

Frequently asked questions

What does a TRISS survival probability actually mean?

It is the proportion of patients with this combination of weighted RTS, ISS, mechanism and age index who survived in the registry the coefficients were fitted to. It is a property of that cohort, not a probability for the person in front of you, and it should not be used to decide whether to continue treatment.

Which RTS does TRISS need?

The weighted Revised Trauma Score, 0 to 7.8408 — not the coded total of 0 to 12. The coefficient was fitted to the weighted term, so a coded total inflates the result badly — on the worked example, from 85.2 per cent to 99.1. This page refuses any RTS above 7.8408 for that reason, so the substitution returns no answer here rather than a wrong one.

Why does age enter as a threshold at 55?

Because that is how the model was specified: an age index of 0 below 55 and 1 from 55 upward. It is a crude treatment and a known limitation — one source read for this page states that the cutoff has to be revised. In practice it means a 54-year-old and a 55-year-old with identical injuries get different numbers, while a 55-year-old and a 95-year-old get the same one.

Which TRISS coefficients should I use?

Say which you used. Both sets on this page are published and in use: the original Major Trauma Outcome Study set, which is what most reproductions of TRISS mean, and a later MTOS revision dated 1995 in the source read here. They differ by about four percentage points on the worked example and more at the extremes.

Is TRISS still accurate?

Its discrimination remains high in validation studies, but its calibration against modern trauma care has drifted, because the coefficients come from a 1980s registry. A 2026 validation reporting an area under the curve of 0.994 still concluded that predictive accuracy shifts between populations and recommended recalibrating the coefficients locally.

Related calculators

References

Wagh A, Subhedar A, Mujawar P. Evaluation of trauma outcomes using the trauma and injury severity score at a tertiary care centre: a prospective observational study. Int Surg J. 2026;13(10):1947–51.
Société Française d’Anesthésie et de Réanimation. TRISS (scores et utilitaires) — the later MTOS coefficient set. sfar.org/scores2/triss2.php
Jung K, Huh Y, Lee JC, et al. The applicability of trauma and injury severity score for a blunt trauma population in Korea and a proposal of new models using score predictors. Yonsei Med J. 2016;57(3):728–34.
North Carolina Office of Emergency Medical Services. North Carolina Trauma Registry Data Dictionary — the weighted RTS field. info.ncdhhs.gov/dhsr/EMS/trauma/pdf/datadictionary.pdf

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/