Liu et al reported a nomogram for predicting in-hospital mortality for a patient in the ICU. The authors are from Chengde Medical University in China.
Patient selection: intensive care (with patients in the Medical Information Mart for Intensive Care, or MIMIC, III database)
Parameters:
(1) age in years, from 10 to 100
(2) sex
(3) body mass index in kg per sq meter, from 0 to 100
(4) mean serum osmotic pressure, from 240 to 370
(5) SOFA score, from 0 to 24 (unspecified if on admission or later in ICU stay)
(6) mean albumin in g/dL, from 1 to 5.5
points for age =
= (0.402 * (age)) - 4.022
points for serum albumin =
= 73.7 - (13.4 * (albumin))
points for body mass index =
= 97.3 - (0.973 * (BMI))
points for osmolarity =
= (0.671 * (osmolarity)) - 160.985
points for SOFA score =
= 4.167 * (SOFA)
Parameter
|
Finding
|
Poiints
|
sex
|
male
|
0
|
|
female
|
6.6
|
total score =
= SUM(points for all of the parameters)
Interpretation:
• minimum score: 0
• maximum score: 387.6
value of X =
= (0.05723 * (score)) - 13.97
probability of in-hospital mortality =
= 1 / 1 + EXP((-1) * X))
Performance
• The area under the ROC curve was 0.85 in the training and 0.82 in the validation cohorts.