MAGENT

IMPACT TBI score: how the core, extended and lab models work

The IMPACT score for traumatic brain injury (TBI) adds points for findings at hospital admission, then turns the sum into two estimates: the percentage chance of death and of an unfavorable outcome at 6 months. The core model uses age, the motor score and pupil reactions. The extended model adds CT and early findings, and the lab model adds glucose and hemoglobin.

Not a medical device. Not a diagnosis or dosing recommendation.

Use this calculator via API

What is the IMPACT score?

IMPACT stands for International Mission for Prognosis and Analysis of Clinical Trials in TBI. Steyerberg and colleagues published the models in PLoS Medicine in 2008. Figure 1 of the paper gives a points chart for each model and the equations that turn a points sum into a 6-month estimate.

Magent's calculator returns the points sum, the two linear predictors and the two percentages to 1 decimal. It's a prognostic estimate and nothing more.

This page explains the arithmetic. It doesn't say what an estimate means for a particular person.

The IMPACT formula

core     = age points + motor points + pupil points
extended = core + hypoxia + hypotension + Marshall CT + tSAH + EDH
lab      = extended + glucose points + hemoglobin points

LP (linear predictor) = intercept + slope × points sum
                 mortality            unfavorable outcome
core             -2.55 + 0.275 × sum   -1.62 + 0.299 × sum
extended         -2.98 + 0.256 × sum   -2.10 + 0.276 × sum
lab              -3.42 + 0.216 × sum   -2.82 + 0.257 × sum

6-month probability = 1 / (1 + e^(-LP)), shown as a percent to 1 decimal

Exact expression the calculator uses:

IMPACT (Steyerberg 2008 Fig 1): core = age(<=30:0, 31-39:1, 40-49:2, 50-59:3, 60-69:4, >=70:5) + motor(1-2:6, 3:4, 4:2, 5-6:0, 9:3) + pupils(both:0, one:2, none:4); extended = core + hypoxia(yes/suspected:1) + hypotension(yes/suspected:2) + marshall(I:-2, II:0, III-VI:2) + tSAH(yes:2) + EDH(yes:-2); lab = extended + glucose_mmol_l(<6:0, 6-8.9:1, 9-11.9:2, 12-14.9:3, >=15:4) + hb_g_dl(<9:3, 9-11.9:2, 12-14.9:1, >=15:0); p = 1/(1+e^(-LP)); LPcore mort=-2.55+0.275*sum; LPcore unfav=-1.62+0.299*sum; LPext mort=-2.98+0.256*sum; LPext unfav=-2.10+0.276*sum; LPlab mort=-3.42+0.216*sum; LPlab unfav=-2.82+0.257*sum; 6-month percent 1 decimal; not MRC CRASH

Inputs

InputModelPoints
age (years, 14 to 90)all30 or under 0 · 31 to 39 1 · 40 to 49 2 · 50 to 59 3 · 60 to 69 4 · 70 or over 5
motor_score (GCS motor 1 to 6, or 9 if untestable)all1 or 2 → 6 · 3 → 4 · 4 → 2 · 5 or 6 → 0 · 9 → 3
pupillary_reactivityallboth react 0 · one reacts 2 · neither reacts 4
hypoxiaextended, labyes or suspected 1
hypotensionextended, labyes or suspected 2
marshall_ct (I to VI)extended, labI → -2 · II → 0 · III to VI → 2
tsahextended, labpresent 2
edhextended, labpresent -2
glucose_mmol_llabunder 6 → 0 · 6 to 8.9 → 1 · 9 to 11.9 → 2 · 12 to 14.9 → 3 · 15 or over → 4
hemoglobin_g_dllabunder 9 → 3 · 9 to 11.9 → 2 · 12 to 14.9 → 1 · 15 or over → 0

How the calculation works

  1. Pick the model: core, extended or lab. The calculator doesn't choose one for you.
  2. Add the points for that model's items. The maximum is 15 for core, 22 for extended and 29 for lab.
  3. Linear predictors: apply the model's intercept and slope to the sum, once for mortality and once for unfavorable outcome. Each is rounded to 3 decimals.
  4. Percentages: put each rounded linear predictor through 1 / (1 + e^(-LP)) and show the result as a percent, rounded to 1 decimal.

Units

Glucose is in mmol/L and hemoglobin in g/dL. The motor score is the motor part of the Glasgow Coma Scale; enter 9 when it can't be tested. A missing value, one outside its range or a missing item for the chosen model is rejected with an error.

Worked examples

These use Magent's own expected values for the calculator. All four start from the same core findings: motor score 3 and both pupils reacting.

Example 1: core model, age 35

  1. Age 35 → 1. Motor 3 → 4. Both pupils → 0. Sum 5.
  2. Mortality LP: -2.55 + 0.275 × 5 = -1.175. 6-month estimate: 23.6%.
  3. Unfavorable LP: -1.62 + 0.299 × 5 = -0.125. 6-month estimate: 46.9%.

Example 2: extended model, adding hypoxia, Marshall V and tSAH

  1. Core points 5. Hypoxia 1, Marshall V 2, tSAH 2 → CT and early findings add 5. Sum 10.
  2. Mortality LP: -2.98 + 0.256 × 10 = -0.42. 6-month estimate: 39.7%.
  3. Unfavorable LP: -2.10 + 0.276 × 10 = 0.66. 6-month estimate: 65.9%.

Example 3: lab model, adding glucose 10 mmol/L and hemoglobin 11 g/dL

  1. Extended points 10. Glucose 10 → 2. Hemoglobin 11 → 2. Sum 14.
  2. Mortality LP: -3.42 + 0.216 × 14 = -0.396. 6-month estimate: 40.2%.
  3. Unfavorable LP: -2.82 + 0.257 × 14 = 0.778. 6-month estimate: 68.5%.

Example 4: core model, age 75

  1. Age 75 → 5. Motor 3 → 4. Both pupils → 0. Sum 9.
  2. Mortality LP: -0.075. 6-month estimate: 48.1%.
  3. Unfavorable LP: 1.071. 6-month estimate: 74.5%.

History of the IMPACT score

The calculator cites Steyerberg and colleagues, PLoS Medicine, 2008 (volume 5, issue 8, article e165). The abstract describes individual patient data pooled from 11 studies of moderate and severe TBI. Logistic regression models were built for mortality and for unfavorable outcome on the Glasgow Outcome Scale at 6 months, checked by leaving out each study in turn, and then tested externally in the MRC CRASH trial.

When it's used

The models use findings available at hospital admission after a moderate or severe traumatic brain injury in adults.

Where it's used

The core model needs age, the motor score and pupil reactions. The extended model also needs CT findings and whether there was hypoxia or hypotension. The lab model also needs admission glucose and hemoglobin.

Who uses it

The person using the calculator picks the model and enters the findings. The calculator adds the points and works out both linear predictors and percentages.

Limits of the calculation

  • The calculator doesn't examine the patient, read the CT scan or measure glucose or hemoglobin. Every finding comes from the person entering it.
  • The percentages are model estimates for a group of patients. They are prognostic figures and don't stand in for a decision about care.
  • It's the IMPACT model. It isn't the MRC CRASH prognostic model.
  • This page doesn't say what to do with an estimate.

Clinical decisions belong to a qualified clinician who knows the patient.

  • Glasgow Coma Scale sums eye, verbal and motor responses. IMPACT uses the motor part alone, with its own points.
  • GCS-Pupils subtracts a pupil score from the GCS total and gives no percentage.
  • FOUR score rates eye, motor, brainstem and breathing responses.

Glossary

  • Linear predictor (LP): the model's intercept plus slope times the points sum, before conversion to a percentage.
  • Unfavorable outcome: the poorer categories of the Glasgow Outcome Scale, death included.
  • Marshall CT class: a grading of the head CT, from I to VI.
  • tSAH: traumatic subarachnoid hemorrhage.
  • EDH: epidural hematoma.

Common mistakes

  • Mixing models. Each model has its own intercepts and slopes. Use the pair that matches the points you added.
  • Entering glucose in mg/dL. The lab model expects mmol/L.
  • Scoring age 30 as 1. 30 and under score 0; the 1-point band starts at 31.
  • Leaving out an untestable motor score. Enter 9, which scores 3.

Source

Frequently asked questions

How is the IMPACT score calculated?

Add the points for the chosen model, apply that model's intercept and slope to get a linear predictor, then convert it to a 6-month percentage.

What is the difference between the core, extended and lab models?

Core uses age, motor score and pupils. Extended adds hypoxia, hypotension, Marshall CT class, tSAH and EDH. Lab adds glucose and hemoglobin.

What does a motor score of 9 mean?

The motor response couldn't be tested. It scores 3.

Why can an item subtract points?

Marshall class I and an epidural hematoma carry -2 in the published chart.

Is IMPACT the same as the CRASH model?

No. This calculator uses the IMPACT models.