EXPERIMENTAL: ARDS 60-day mortality v2

Derived variable

Equations

The model uses:

R = PaO₂ / PaCO₂

Define the hinge function:

h(u) = max(0, u)

The Yeo–Johnson transformation for x ≥ 0 is:

YJ_λ(x) = ((x + 1)^λ − 1) / λ when λ ≠ 0.

The model linear predictor is:

η = 0.1803450722 − 0.0239113110 × h(124.8493196754 − YJ_1.1395665474(Age))

+ 4.5690181283 × h(0.5970780861 − YJ_{−1.1051799726}(R))

+ 4.2596007650 × h(YJ_{−1.0302359828}(Lactate) − 0.2843467765)

− 0.5602665090 × h(3.3011808232 − YJ_0.1891305588(Days))

+ 4.0352484115 × h(7.3 − pH)

Predicted probability of death by 60 days:

P = 1 / (1 + e^−η)

Predicted 60-day mortality = 100 × P