Slankamenac et al developed 2 models (full and reduced versions) for predicting acute renal failure after liver resection. The authors are from the University of Zurich.
Patient selection: hepatic resection
Outcome: acute renal failure (occurred in about 15%)
Parameters in full model:
(1) cardiovascular disease
(2) chronic renal failure (CRF)
(3) diabetes
(4) serum ALT in U/L and sex
(5) age in years
(6) serum total bilirubin in µmol/L
(7) sex
Parameters in reduced model: (1), (2), (3), (4)
Parameter |
Finding |
Full Model |
Reduced Model |
cardiovascular disease |
no |
0 |
0 |
|
yes |
5 |
2 |
CRF |
no |
0 |
0 |
|
yes |
4 |
2 |
diabetes |
no |
0 |
0 |
|
yes |
3 |
1 |
serum ALT |
<= 35, female |
0 |
0 |
|
> 35, female |
5 |
2 |
|
<= 50, male |
0 |
0 |
|
> 50, male |
5 |
2 |
age in years |
< 60 years |
0 |
0 |
|
60 to 69 years |
1 |
0 |
|
>= 70 years |
2 |
0 |
total bilirubin |
< 17 µµmol/L |
0 |
0 |
|
>= 17 µmol/L |
1 |
0 |
sex |
male |
0 |
0 |
|
female |
1 |
0 |
total score for full model =
= SUM(points for all 7 parameters)
total score for reduced model =
= SUM(points for all 4 parameters)
Interpretation:
• minimum full or reduced score: 0
• maximum for full model: 22
• maximum for reduced model: 7
• The mortality rate for a patient with acute renal failuare was 22% vs about 1% without.
Score |
ARF Full Model |
ARF Reduced Model |
0 |
4% |
6% |
1 |
5% |
10% |
2 |
6% |
17% |
3 |
8% |
27% |
4 |
10% |
41% |
5 |
12% |
55% |
6 |
15% |
69% |
7 |
18% |
81% |
8 |
21% |
|
9 |
26% |
|
10 |
30% |
|
11 |
35% |
|
12 |
41% |
|
13 |
46% |
|
14 |
52% |
|
15 |
58% |
|
16 |
63% |
|
17 |
69% |
|
18 |
73% |
|
19 |
78% |
|
20 |
81% |
|
21 |
85% |
|
22 |
87% |
|
Performance:
• The area under the ROC curve was 0.80 for the full model and 0.77 for the reduced.
Specialty: Nephrology, Gastroenterology, Surgery, general